My Links

Nicholas Horton

Professor of Statistics, Amherst College
Seelye Mudd Hall 303
(413) 542–5655
nhorton AT



(selected papers are reprinted with permission from the publishers, paperwork available upon request)

  1. Zaslavsky AM and Horton NJ. Balancing disclosure risk against the loss of nonpublication. Journal of Official Statistics 1998; 14:411-419.

    A nondisclosure policy for tabular data on microdata restricts release of information that could be related to a specific individual. Pannekoel and de Wall (1998) describe a rule that suppresses data release when the number of people in a cell defined by a rare characteristic falls below a fixed floor, and show how empirical Bayes methods can be used to improve the estimation of that number. We argue that the nondisclosure problem can be formulated as a decision problem in which one loss is associated with the possibility of disclosure and another with nonpublication of data. This analysis supports a decision on whether to disclose information in each cell, minimizing the expected sum of the two losses. We present arguments for several loss functions, considering both tabular and microdata releases, and illustrate their application to simple simulated data.
  2. Horton NJ, Bebchuk JD, Jones CL, Lipsitz SR, Catalano PJ, Zahner GEP and Fitzmaurice GM. Goodness of fit for GEE: An example with mental health service utilization, Statistics in Medicine 1999; 18:213-222. (The SAS macro implementing the method is available from our website.)

    Suppose we use generalized estimating equations to estimate a marginal regression model for repeated binary observations. There are no established summary statistics available for assessing the adequacy of the fitted model. In this paper we propose a goodness-of-fit test statistic which has an approximate chi-squared distribution when we have specified the model correctly. The proposoed statistic can be viewed as an extension of the Hosmer and Lemeshow goodness-of-fit statistic for ordinary logistic regression to marginal regression models for repeated binary responses. We illustrate the methods using data from a study of mental health service utilization by children. The repeated responses are a set of binary measures of service use. We fit a marginal logistic regression model to the data using generalized estimating equations, and we apply the proposed goodness-of-fit statistic to assess the adequacy of the fitted model.
  3. Horton NJ, Laird NM and Zahner GEP. Use of multiple informant data as a predictor in psychiatric epidemiology. International Journal of Methods in Psychiatric Research 1999; 8:6-18.

    Multiple informant reports of psychopathology are often collected in studies of psychiatric epidemiology. Previous reports have primarily focused on methods for handling multiple informant information when it is used to measure a psychiatric outcome in a regression model. Here we deal with methods for incorporating multiple informant reports as predictor variables (covariates) in regression modeling. In general, there is no single appropriate analytic strategy in this setting. Approaches seen in the literature include separate analyses by informant, inclusion of both reports in the model, or combination of the reports (using the OR rule or use of concordant reports). Other approaches include use of measurement error models and latent class analysis. We review these approaches, discuss their relative advantages and disadvantages and illustrate them with an example where dichotomous reports of psychopathology are used to predict mental health service utilization in a community-based sample.
  4. Horton NJ and Laird NM. Maximum likelihood analysis of generalized linear models with missing covariates. Statistical Methods in Medical Research 1999; 8:37-50.

    Missing data is a common occurrence in most medical research data collection enterprises. There is an extensive literature concerning missing data, much of which has focused on missing outcomes. Covariates in regression models are often missing, particularly if information is being collected from multiple sources. The method of weights is an implementation of the EM algorithm for general maximum-likelihood analysis of regression models, including generalized linear models (GLMs) with incomplete covariates. In this paper, we will describe the method of weights in detail, illustrate its application with several examples, discuss its advantages and limitations, and review extensions and applications of the method.
  5. Horton NJ and Lipsitz SR. Review of software to fit generalized estimating equation (GEE) regression models. The American Statistician 1999; 53:160-169.

    Researchers are often interested in analyzing data which arise from a longitudinal or clustered design. While there are a variety of standard likelihood-based approaches to analysis when the outcome variables are approximately multivariate normal, models for discrete-type outcomes generally require a different approach. Liang and Zeger formalized an approach to this problem using Generalized Estimating Equations (GEEs) to extend Generalized Linear Models (GLMs) to a regression setting with correlated observations within subjects. In this paper, we briefly review GLM, the GEE methodology, introduce some exampoles, and compare the GEE implementations of several general purpose statistical packages (SAS, Stata, SUDAAN and S-Plus). We focus on the user interface, accuracy, and completeness of implementations of this methodology.
  6. Seidman LJ, Buka SL, Goldstein JM, Horton NJ, Rieder RO and Tsuang MT. The relationship of prenatal and perinatal complications to cognitive functioning at age 7 in the New England Cohorts of the National Collaborative Perinatal Project. Schizophrenia Bulletin 2000; 26(2):309-21.

    Previous literature shows that children who later develop schizophrenia have elevated rates of prenatal and perinatal complications (PPCs) and neuropsychological deficits in childhood. However, little is known about the relationship of these risk factors to each other. We evaluated the relationship between PPCs and neuropsychological functioning at age 7 in a large epidemiological study of pregnancy, birth, and development: the National Collaborative Perinatal Project (NCPP). Thirteen standardized measures of cognitive abilities were acquired on 11,889 children at approximately age 7. Principal components analysis was used to create three neuropsychological measures: academic achievement skills, verbal-conceptual abilities, and perceptual-motor abilities. We measured the relationship between these factors and three measures of PPCs: low birth weight (LBW), probable hypoxicischemic complications, and chronic hypoxia. All three measures of PPCs were significantly associated with lower neuropsychological performance, after controlling for various confounders. LBW had the strongest association with neuropsychological performance, followed by an index of presumed hypoxic insults. The effect sizes between PPCs and cognitive factors at age 7 were consistently largest with perceptual-motor abilities, followed by academic achievement skills and verbal-conceptual abilities. Future studies will evaluate the effects of specific PPCs and genetic risk factors for psychosis on cognitive functioning in childhood.
  7. Goldstein JM, Seidman LJ, Buka SL, Horton NJ, Donatelli JL, Rieder RO and Tsuang MT. Impact of genetic vulnerability and hypoxia on overall intelligence by age 7 in offspring at high risk for schizophrenia compared with affective psychoses. Schizophrenia Bulletin 2000; 26(2):323-34.

    Risk factors for schizophrenia, such as genetic vulnerability and obstetric complications, have been associated with cognitive deficits in schizophrenia. We tested the association of these risk factors with general intellectual ability in offspring at high risk for psychoses and normal control subjects. Offspring of 182 parents with DSM-IV schizophrenia or affective psychose were recruited and diagnosed from the Boston and Providence cohorts of the National Collaborative Perinatal Project (NCPP). Control subjects from the NCPP were selected to be comparable with affected parents based on the parent's age, ethnicity, study site, number of offspring enrolled in the NCPP, and payment status, and on the offspring's age, sex, and history of obstetric complications. Based on data prospectively acquired from pregnancy and events of gestation, labor, delivery, and the neonatal period, we derived a measure of probable hypoxic-ischemic insult. We also report on standardized measures of general intelligence (intelligence quotient [IQ]) collected at age 7. General linear mixed models were used to test for the simultaneous effects of genetic vulnerability, defined as parental diagnosis, and probable hypoxic insult on age 7 IQ. Specificity of the effects for schizophrenia compared with affective psychoses and sex effects were also tested. Low IQ at age 7 was significantly associated with genetic vulnerability to psychoses, in particular with schizophrenia.
  8. Horton NJ and Laird NM. Maximum likelihood analysis of logistic regression models with incomplete covariate data and auxiliary information. Biometrics 2001; 57:34-42.

    This paper presents a new method for maximum likelihood estimation of logistic regression models with incomplete covariate data where auxiliary information is available. This auxiliary information is extraneous to the regression model of interest, but predictive of the covariate with missing data. Ibrahim (Journal of the American Statistical Association, 1990) provides a general method for estimating generalized linear regression models with missing covariates using the EM algorithm that is easily implemented when there is no auxiliary data. Vach (Statistics in Medicine, 1997) describes how the method can be extended when the outcome and auxiliary data are conditionally independent given the covariates in the model. The method allows the incorporation of auxiliary data without making the conditional independence assumption. We suggest tests of conditional independence and compare the performance of several estimators in an example concerning mental health service utilization in children. Using an artificial dataset, we compare the performance of several estimators when auxiliary data are available.
  9. Horton NJ, Laird NM, Murphy JM, Monson RR, Sobol AM and Leighton AH. Multiple informants: mortality associated with psychiatric disorders in the Stirling County Study. American Journal of Epidemiology, 2001; 154(7):649-656.

    This paper applies new statistical procedures for analyzing multiple-source information about the relationship of psychiatric diagnoses to mortality. The data come from the Stirling County Study, a longitudinal community investigation of adults, which collected multiple source reports (self-report and physician-report) about psychiatric disorders. These reports are used as predictors of mortality risk over a 16-year follow-up period. Despite extensive efforts, one or both of these reports were sometimes missing. Missingness of self-report was related to demographic characteristics as well as physician reports of psychiatric diagnosis. The statistical procedures used here draw together into a single frame of reference both informant reports for the initial Stirling survey, and relate these to mortality risk using weighted generalized estimating equation (GEE) regression models for time to event data. This unified method has two advantages over traditional approaches: the relative predictiveness of each informant can be assessed, and all subjects contribute to the analysis. The methods are applicable to other areas of epidemiology where multiple informant reports are utilized. The results for self-reports and physician-reports of disorders were comparable: psychiatric diagnosis was associated with higher mortality, particularly among younger subjects.
  10. Goldstein JM, Seidman LJ, Horton NJ, Makris N, Kennedy DN, Caviness VS Jr, Faraone SV and Tsuang MT. Normal sexual dimorphism of the adult human brain assessed by in vivo magnetic resonance imaging. Cerebral Cortex 2001; 11(6):490-497.

    The etiology and consistency of findings on normal sexual dimorphisms of the adult human brain are unresolved. In this study, we present a comprehensive evaluation of normal sexual dimorphisms of cortical and subcortical brain regions, using in vivo magnetic resonance imaging, in a community sample of 48 normal adults. The men and women were similar in age, education, ethnicity, socioeconomic status, general intelligence and handedness. Forty-five brain regions were assessed based on T(1)-weighted three-dimensional images acquired from a 1.5 T magnet. Sexual dimorphisms of adult brain volumes were more evident in the cortex, with women having larger volumes, relative to cerebrum size, particularly in frontal and medial paralimbic cortices. Men had larger volumes, relative to cerebrum size, in frontomedial cortex, the amygdala and hypothalamus. A permutation test showed that, compared to other brain areas assessed in this study, there was greater sexual dimorphism among brain areas that are homologous with those identified in animal studies showing greater levels of sex steroid receptors during critical periods of brain development. These findings have implications for developmental studies that would directly test hypotheses about mechanisms relating sex steroid hormones to sexual dimorphisms in humans.
  11. Horton NJ and Lipsitz SR. Multiple imputation in practice: comparison of software packages for regression models with missing variables. The American Statistician, 2001; 55(3):244-254.

    Missing data frequently complicates data analysis for scientific investigations. The development of statistical methods to address missing data has been an active area of research in recent decades. Multiple imputation, originally proposed by Rubin in a public use dataset setting, is a general purpose method for analyzing datasets with missing data that is broadly applicable to a variety of missing data settings. We review multiple imputation as an analytic strategy for missing data. We describe and evaluate a number of software packages that implement this procedure, and contrast the interface, features, and results. We compare the packages, and detail shortcomings and useful features. The comparisons are illustrated using examples from an artificial dataset and a study of child psychopathology. We suggest additional features as well as discuss limitations and cautions to consider when using multiple imputation as an analytic strategy for incomplete data settings.
  12. Rees VW, Saitz R, Horton NJ and Samet JH. Association of alcohol consumption with HIV sex and drug risk behaviors among drug users. Journal of Substance Abuse Treatment, 2001; 21(3):129-134.

    The relationship between alcohol use and HIV transmission is well recognized but not fully understood. In particular, the role of alcohol abuse as a mediator of HIV risk behavior among drug users is not well documented. We hypothesized that alcohol use in drug users will result in greater HIV risk-taking behavior. Participants were 354 drug users, of whom 105 were recent injection drug users. Multiple regression models were used to characterize whether measures of sexual and injection drug use HIV risk behavior were related to alcohol consumption, controlling for other potentially associated factors. We found that sexual HIV risk-taking behavior is associated with increased alcohol consumption among women (p = 0.02), with women having more risky sexual behavior than males. However, contrary to our hypothesis, there was no significant association of alcohol consumption with risky injection drug behavior. Addressing alcohol problems among drug users, particularly women, may be an important opportunity to reduce HIV sexual risk behavior among this high-risk population.
  13. Ibrahim JG, Lipsitz SR and Horton NJ. Using auxiliary data for parameter estimation with non-ignorably missing outcomes Journal of the Royal Statistical Society: Series C (Applied Statistics), 2001; 50(3):361-373.

    We propose a method for estimating parameters in generalized linear models when the outcome variable is missing for some subjects and the missing data mechanism is non-ignorable. We assume throughout that the covariates are fully observed. One possible method for estimating the parameters is maximum likelihood with a non-ignorable missing data model. However, caution must be used when fitting non-ignorable missing data models because certain parameters may be inestimable for some models. Instead of fitting a non-ignorable model, we propose the use of auxiliary information in a likelihood approach to reduce the bias, without having to specify a non-ignorable model. The method is applied to a mental health study.
  14. Goldstein JM, Seidman LJ, O'Brien LM, Horton NJ, Kennedy DN, Makris N, Caviness VS, Faraone SV and Tsuang MT. Impact of normal sexual dimorphisms on sex differences in structural brain abnormalities in schizophrenia assessed by magnetic resonance imaging. Archives of General Psychiatry, 2002; 59(2):154-64.

    BACKGROUND: Previous studies suggest that the impact of early insults predisposing to schizophrenia may have differential consequences by sex. We hypothesized that brain regions found to be structurally different in normal men and women (sexual dimorphisms) and abnormal in schizophrenia would show significant sex differences in brain abnormalities, particularly in the cortex, in schizophrenia. METHODS: Forty outpatients diagnosed as having schizophrenia by DSM-III-R were systematically sampled to be comparable within sex with 48 normal comparison subjects on the basis of age, ethnicity, parental socioeconomic status, and handedness. A comprehensive assessment of the entire brain was based on T1-weighted 3-dimensional images acquired from a 1.5-T magnet. Multivariate general linear models for correlated data were used to test for sex-specific effects regarding 22 hypothesized cortical, subcortical, and cerebrospinal fluid brain volumes, adjusted for age and total cerebrum size. Sex x group interactions were also tested on asymmetries of the planum temporale, Heschl's gyrus, and superior temporal gyrus, additionally controlled for handedness. RESULTS: Normal patterns of sexual dimorphisms were disrupted in schizophrenia. Sex-specific effects were primarily evident in the cortex, particularly in the frontomedial cortex, basal forebrain, cingulate and paracingulate gyri, posterior supramarginal gyrus, and planum temporale. Normal asymmetry of the planum was also disrupted differentially in men and women with schizophrenia. There were no significant differential sex effects in subcortical gray matter regions or cerebrospinal fluid. CONCLUSION: Factors that produce normal sexual dimorphisms may be associated with modulating insults producing schizophrenia, particularly in the cortex.
  15. Rees VW, Horton NJ, Hingson RW, Saitz R and Samet JH. Injury among detoxification patients: alcohol users' greater risk. Alcohol: Clinical and Experimental Research, 2002; 26(2):212-7.

    BACKGROUND: Little information exists about serious injury in alcohol or drug abusers who seek detoxification. This study examined injury prevalence, and the impact of alcohol use on injury, among alcohol- and drug-dependent persons. METHOD: We enrolled patients at a detoxification unit into a prospective cohort study and performed follow-up interviews to determine the prevalence of injury. Baseline and follow-up (6, 12, 18, and 24 month) data were examined separately by using multivariate logistic regression to determine factors associated with self-reported episodes of serious injury. The main independent variable was self-reported first and second substances of choice at admission: alcohol only, drug only, or both. RESULTS: Overall, 24% of the 470 subjects reported at least one instance of serious injury over the 6-month period before detoxification. Similarly, approximately 20% of subjects had serious injury during each 6-month follow-up period. Injury in the past 6 months was highest among the 63% of subjects who reported alcohol as a drug of choice (28-29% vs. 16% for drug only), even after we controlled for potential confounders. Analysis of 2-year follow-up data revealed a similar association, after we controlled for baseline injury and alcohol consumption. CONCLUSIONS: Injury is a serious problem for a substantial proportion of patients who undergo detoxification, particularly those with alcohol dependence. This marked risk for serious injury persisted for 24 months after detoxification. Patients at detoxification, particularly those with alcohol problems, represent a high-risk population for injury that may benefit from interventions to reduce these preventable complications.
  16. Goldstein JM, Cohen LS, Horton NJ, Lee H, Andersen S, Tohen M, Crawford AK and Tollefson G. Sex differences in clinical response to olanzapine compared with haloperidol. Psychiatry Research, 2002; 110(1):27-37.

    There is current disagreement over whether men and women respond differently to typical or atypical antipsychotic medications. This study reanalyzed a large international clinical trial of olanzapine (Olz) compared with haloperidol (Hal) to test for sex differences in treatment response, controlling for illness chronicity and menopausal status. We hypothesized that women would show a greater response to either medication than men, particularly among first admission, premenopausal women. DSM-III-R schizophrenia inpatients (700 women and 1295 men) were randomly assigned to a 6-week trial of Olz vs. Hal. Longitudinal random effect models were used to test for interactions of sex with medication, chronicity and menopausal status on treatment response. Findings showed that women on olanzapine had a greater drop in overall symptomatology by week 4 than any other group, and their level of symptomatology remained lower throughout the 6-week trial. The sex differences in treatment response in olanzapine compared with haloperidol were, in part, dependent on chronicity and, in women, menopausal status. That is, first episode women on haloperidol exhibited an increase in symptomatology over the 6-week trial compared to their male counterparts, while multiply hospitalized women had a better treatment response on haloperidol than their male counterparts. Women on olanzapine had a significantly better treatment response than men, regardless of chronicity. Finally, premenopausal women had a significantly better treatment response than postmenopausal women, regardless of treatment and chronicity.
  17. Seidman LJ, Faraone SV, Goldstein JM, Kremen WS, Horton NJ, Makris N, Toomey R, Kennedy D, Caviness VS and Tsuang MT. Left hippocampal volume as a vulnerability indicator for schizophrenia: an MRI morphometric study of non-psychotic first degree relatives. Archives of General Psychiatry, 2002; 59(9):839-849.

    BACKGROUND: Clues to the causes of schizophrenia can be derived from studying first-degree relatives because they are genetically related to an ill family member. Abnormalities observed in nonpsychotic relatives are indicators of possible genetic vulnerability to illness, independent of psychosis. We tested 4 hypotheses: (1) that hippocampal volume is smaller in nonpsychotic relatives than in controls, particularly in the left hemisphere; (2) that hippocampi will be smaller in multiplex relatives as compared with simplex relatives, and both will be smaller than in controls; (3) that hippocampal volumes and verbal declarative memory function will be positively correlated; and (4) that hippocampi will be smaller in patients with schizophrenia than in their nonpsychotic relatives or in controls. METHODS: Subjects were 45 nonpsychotic adult first-degree relatives from families with either 2 people ("multiplex," n = 17) or 1 person ("simplex," n = 28) diagnosed with schizophrenia, 18 schizophrenic relatives, and 48 normal controls. Sixty contiguous 3-mm coronal, T1-weighted 3-dimensional magnetic resonance images of the brain were acquired on a 1.5-T magnet. Volumes of the total cerebrum and the hippocampus were measured. RESULTS: Compared with controls, relatives, particularly from multiplex families, had significantly smaller left hippocampi. Verbal memory and left hippocampal volumes were significantly and positively correlated. Within families, hippocampal volumes did not differ between schizophrenic patients and their nonpsychotic relatives. CONCLUSIONS: Results support the hypothesis that the vulnerability to schizophrenia includes smaller left hippocampi and verbal memory deficits. Findings suggest that smaller left hippocampi and verbal memory deficits are an expression of early neurodevelopmental compromise, reflecting the degree of genetic liability to schizophrenia.
  18. Horton NJ and Fitzmaurice GM. Maximum likelihood estimation of bivariate logistic models for incomplete responses with indicators of ignorable and non-ignorable missingness. Journal of the Royal Statistical Society, Series C (Applied Statistics), 2002; 51(3):281-295.

    Missing observations are a common problem that complicate the analysis of clustered data. In the Connecticut child surveys of childhood psychopathology, it was possible to identify reasons why outcomes were not observed. Of note, some of these causes of missingness may be assumed to be ignorable, whereas others may be non-ignorable. We consider logistic regression models for incomplete bivariate binary outcomes and propose mixture models that permit estimation assuming that there are two distinct types of missingness mechanisms: one that is ignorable; the other non-ignorable. A feature of the mixture modelling approach is that additional analyses to assess the sensitivity to assumptions about the missingness are relatively straightforward to incorporate. The methods were developed for analysing data from the Connecticut child surveys, where there are missing informant reports of child psychopathology and different reasons for missingness can be distinguished.
  19. Liebschutz J, Savetsky JB, Saitz R, Horton NJ, Lloyd-Travaglini C and Samet JH. The relationship between sexual and physical abuse and substance abuse consequences. Journal of Substance Abuse Treatment, 2002; 22(3):121-128.

    This study examines the relationship between a history of physical and sexual abuse (PhySexAbuse) and drug and alcohol related consequences. We performed a cross-sectional analysis of data from 359 male and 111 female subjects recruited from an inpatient detoxification unit. The Inventory of Drug Use Consequences (InDUC), measured negative life consequences of substance use. Eighty-one percent of women and 69% of men report past PhySexAbuse, starting at a median age of 13 and 11, respectively. In bivariate and multivariable analyses, PhySexAbuse was significantly associated with more substance abuse consequences (p < 0.001). For men, age less than or equal to 17 years at first PhySexAbuse was significantly associated with more substance abuse consequences than an older age at first abuse, or no abuse (p = 0.048). For women, the association of PhySexAbuse with substance use consequences was similar across all ages (p = 0.59). Future research should develop interventions to lessen the substance abuse consequences of physical and sexual abuse.
  20. Rosenberg L, Palmer JR, Wise LA, Horton NJ and Corwin MJ. Perceptions of racial discrimination in relation to preterm birth. Epidemiology, 2002; 13(6):646-652.

    BACKGROUND Because racial discrimination might contribute to their excess of preterm births, we assessed experiences of racism in relation to preterm birth among African-American women.METHODS We used data from the Black Women's Health Study, a follow-up study of African-American women begun in 1995. Data on subsequent singleton births were obtained using follow-up questionnaires in 1997 and 1999; nine questions about experiences of racism were asked in 1997. We compared mothers of 422 babies born 3 or more weeks early (because of premature labor for unknown reasons or rupture of membranes) with mothers of 4544 babies of longer gestation. We used generalized estimating equation models to estimate odds ratios (ORs) for preterm birth, controlling potential confounders. RESULTS The adjusted ORs for preterm birth were 1.3 (95% confidence interval [CI] = 1.1-1.6) for women who reported unfair treatment on the job and 1.4 (1.0-1.9) for women who reported that people acted afraid of them at least once a week. Overall ORs for the seven other racism questions were close to 1.0. Among 491 women with less than or equal to 12 years of education, ORs were 2.0 or greater for four racism variables. CONCLUSIONS These data provide some evidence for an increase in preterm birth among women who report experiences of racism, particularly women with lower levels of education.
  21. Demissie S, LaValley MP, Horton NJ, Glynn RJ and Cupples LA. Bias due to missing covariate information using complete-case analysis in the proportional hazards model. Statistics in Medicine, 2003; 22(4):545-57.

    We studied bias due to missing exposure data in the proportional hazards regression model when using complete-case analysis (CCA). Eleven missing data scenarios were considered: one with missing completely at random (MCAR), four missing at random (MAR), and six non-ignorable missingness scenarios, with a variety of hazard ratios, censoring fractions, missingness fractions and sample sizes. When missingness was MCAR or dependent only on the exposure, there was negligible bias (2-3 per cent) that was similar to the difference between the estimate in the full data set with no missing data and the true parameter. In contrast, substantial bias occurred when missingness was dependent on outcome or both outcome and exposure. For models with hazard ratio of 3.5, a sample size of 400, 20 per cent censoring and 40 per cent missing data, the relative bias for the hazard ratio ranged between 7 per cent and 64 per cent. We observed important differences in the direction and magnitude of biases under the various missing data mechanisms. For example, in scenarios where missingness was associated with longer or shorter follow-up, the biases were notably different, although both mechanisms are MAR. The hazard ratio was underestimated (with larger bias) when missingness was associated with longer follow-up and overestimated (with smaller bias) when associated with shorter follow-up. If it is known that missingness is associated with a less frequently observed outcome or with both the outcome and exposure, CCA may result in an invalid inference and other methods for handling missing data should be considered.
  22. Lash TL, Thwin SS, Horton NJ, Guadagnoli E and Silliman RA. Multiple informants: a new method to assess breast cancer patients' comorbidity. American Journal of Epidemiology, 2003; 157(3):249-257.

    Past assessments of comorbidity indices have sought to recommend a single index that performs better than others. The authors used a multiple informants approach as an alternative method to simoultaneously assess five indices of comorbidity. This approach provides a single estimate of the overall effect of comorbidity and evaluates the relation any individual index has to the outcomes of interest. Association of comorbidity with definitive primary therapy, discussion of tamoxifen, and receipt of tamoxifen was evaluated in a cohort of 830 older breast cancer patients enrolled at four geographically distinct centers in the United States from 1996 to 1999. The estimated adjusted effect of a unit increase in comorbidity on the odds of discussing tamoxifen therapy was 0.70 (95% confidence interval: 0.56, 0.88). An increase in comorbidity was not associated with receipt of definitive primary therapy (odds ratio = 0.94, 95% confidence interval: 0.79, 1.13) or receipt of tamoxifen (odds ratio = 0.96, 95% confidence interval: 0.72, 1.27). The multiple informants regression proved superior to separate regression models that included only one index. In analyses that require comorbidity adjustment and for which no single index ios expected to be ideal, the multiple informants approach is an attractive alternative to selecting a single index and to other methods of using multiple indices.
  23. Saitz R, Horton NJ and Samet JH. Alcohol and medication interactions in primary care patients: common and unrecognized. American Journal of Medicine, 2003; 114(5):407-410.

    Alcohol use disorders are common and costly. Primary care settings are well suited for alcohol screening and intervention. There are valid, brief, practical screening tools; brief interventions can reduce drinking and improve health outcomes; and national practice guidelines recommend these activities. However, alcohol problems in primary care are often unrecognized and untreated. Interactions between alcohol and medications may be another reason to identify and address alcohol problems in medical settings. Physicians are advised routinely to check for drug interactions between medications, but it is not known hoow frequently physicians check for alcohol and medication interactions. We sought to determine the prevalence of potential common alcohol and medication interactions in primary care patients with at-risk, harmful, or dependent (hereafter called "hazardous") drinking, and whether users of medications that interact with alcohol were counseled by physicians about drinking.
  24. Palmer JR, Rosenberg L, Wise LA, Horton NJ and Adams-Campbell LL. Onset of natural menopause in African-American women. American Journal of Public Health, 2003; 93(2):299-306.

    OBJECTIVES: This study assessed predictors of the onset of natural menopause in African American women. METHODS: We used mailed questionnaires to collect data at baseline in 1995 and during follow-up from Black Women's Health Study participants. Cox proportional hazards regression was used to assess potential predictors-including experiences of racism-of the onset of natural menopause among 17 070 women aged 35 to 55 years and premenopausal in 1995. RESULTS: The hazard ratio (HR) was 1.43 for current smokers (95% confidence interval [CI] = 1.24, 1.66) and 1.21 (95% CI = 1.06, 1.38) for ex-smokers and significantly less for obese women and oral contraceptive users. Hazard ratios for most questions about racism were elevated by 10% to 30% but were not statistically significant. CONCLUSIONS:Earlier onset of natural menopause among African American women is strongly associated with smoking and inversely associated with body mass index and oral contraceptive use.
  25. Samet JH, Larson MJ, Horton NJ, Doyle K, Winter M and Saitz R. Linking alcohol and drug dependent adults to primary medical care: A randomized controlled trial of a multidisciplinary health intervention in a detoxification unit. Addiction, 2003; 98(4):509-516.

    AIM: Pragmatic approaches to integration of medical care and substance abuse treatment are desired. We assessed the effectiveness of a novel multi-disciplinary clinic for linking patients in a residential detoxification program to primoary medical care. PARTICIPANTS: We enrolled patients undergoing in-patient detoxification from alcohol, heroin or cocaine who had no primary care physician into a randomized controlled trial. The intervention consisted of a clinical evaluation at the detoxification unit in the health evaluation and linkage to primary care (HELP) clinic by a nurse, social worker and physician and facilitated referral to an off-site primary care clinic. The primary outcome of interest was attendance at a primary care appointment within 12 months. Secondary outcomes assessed over 24 months were addiction severity, health-related quality of life, utilization of medical and addiction services and HIV risk behaviors. FINDINGS: Of the 470 subjects enrolled, 235 were randomized to the HELP clinic intervention. Linkage to primary medical care occurred in 69% of the intervention group compared to 53% in the control group (P = 0.0003). The clinic was similarly effective for subjects with alcohol and illicit drug problems. Randomization to the HELP clinic resulted in no significant differences in secondary outcomes. CONCLUSIONS: The HELP clinic, a multi-disciplinary clinic located in a detoxification unit, effectively linked alcohol- and drug-dependent individuals to primary medical care. This intervention utilized a 'reachable moment', the period of addiction care, as a window of opportunity for linking substance abusers to medical care.
  26. Saitz R, Horton NJ, Sullivan LM, Moskowitz MA and Samet JH. Addressing alcohol problems in primary care: A cluster randomized, controlled trial of a systems intervention (the Screening and Intervention in Primary care [SIP] Study). Annals of Internal Medicine, 2003; 138(5):372-382.

    BACKGROUND: Screening and intervention for alcohol problems can reduce drinking and its consequences but are often not implemented. OBJECTIVE: To test whether providing physicians with patients' alcohol screening results and simple individualized recommendations would affect the likelihood of a physician's having a discussion with patients about alcohol during a primary care visit and would affect subsequent alcohol use. DESIGN: Cluster randomized, controlled trial. SETTING: Urban academic primary care practice. PARTICIPANTS: 41 faculty and resident primary care physicians and 312 patients with hazardous drinking. INTERVENTIONS: Providing physicians with alcohol screening results (CAGE questionnaire responses, alcohol consumption, and readiness to change) and recommendations for their patients at a visit. MEASUREMENTS: Patient self-report of discussions about alcohol use immediately oafter the physician visit and alcohol use 6 months later. RESULTS: Of 312 patients, 240 visited faculty physicians, 301 (97%) completed the outcome assessment after the office visit, and 236 (76%) were followed for 6 months. Faculty physicians in the intervention group tended to be more likely than faculty physicians in the control group to give patients advice about drinking (adjusted proportion, 64% [95% CI, 47% to 79%] vs. 42% [CI, 33% to 53%]) and to discuss problems associated with alcohol use (74% [CI, 59% to 85%] vs. 51% [CI, 39% to 62%]). Resident physicians' advice and discussions did not differ between groups. Six months later, patients who saw resident physicians in the intervention group had fewer drinks per drinking day (adjusted mean number of drinks, 3.8 [CI, 1.9 to 5.7] versus 11.6 [CI, 5.4 to 17.7]). CONCLUSIONS: Although effects seem to differ by physician level of training, prompting physicians with alcohol screening results and recommendations for action can modestly increase discussions about alcohol use and advice to patients and may decrease alcohol consumption.
  27. Kertesz SG, Horton NJ, Friedmann PD, Saitz R, and Samet JH. Slowing the revolving door: stabilization programs reduce homeless persons' substance use after detoxification. Journal of Substance Abuse Treatment, 2003; 24(3):197-207.

    This study examined whether homelessness predicted earlier resumption of substance use after detoxification, and sought evidence concerning the impact of post-detoxification stabilization programs among homeless and nonhomeless individouals. Kaplan-Meier plots and proportional hazards models were used to determine the association between homelessness, stabilization program use, and recurrent substance use in a prospective cohort of persons entering inpatient detoxification (n=470). Among 254 persons available at 6 months, 76% reported recurrent substance use. Homeless persons not using stabilization programs experienced the highest hazard of return to substance use after detoxification, Hazard Ratio (HR) 1.26, 95% CI (0.88, 1.80). Homeless persons using these programs had the lowest rate of return to substance use: HR 0.61, 95% CI (0.40, 0.94). A similar impact of stabilization programs was not seen among nonhomeless subjects. Post-detoxification stabilization programs were associated with improved outcomes for homeless addicted persons. This treatment modality may slow the "revolving door" phenomenon of relapse after detoxification among homeless persons.
  28. Liebschutz J, Finley EP, Braslins PG, Christiansen D, Horton NJ, and Samet JH. Screening for sexually transmitted infections in substance abuse treatment programs. Drug and Alcohol Dependence, 2003; 70(1):93-99.

    Objectives: We evaluated the prevalence of the sexually transmitted infections (STIs) chlamydia and gonorrhea in clients at a methadone maintenance program and a residential detoxification program. Methods: We collected urine specimens for chlamydia and gonorrhea ligase chain reaction testing and assessed sexual, substance abuse and STI histories. Results: Of 700 subject assessments, 490 occurred among detoxification clients and 210 in methadone maintenance. Chlamydia trachomatis was detected in 5/700 (0.9, 95% CI=0.1š1.8%) and Neisseria gonorrhoeae in none. All chlamydia infected subjects were recruited from the detoxification program. Subjects reported high risk sexual behavior: 17% reported commercial sex exchange, and 22% reported inconsistent condom use with multiple sexual partners during the prior 2 months. Conclusion: Based on prevalence in Boston, MA, universal screening for STI in substance abuse treatments programs is not warranted. However, routine screening for younger substance abusers and in communities with high prevalence should be considered.

  29. Palmer JR, Wise LA, Horton NJ, Adams-Campbell LL and Rosenberg L. Dual effect of parity on breast cancer risk in African-American women. Journal of the National Cancer Institute, 2003; 95(6):478-83.

    BACKGROUND: In the United States, breast cancer incidence is higher among African-American women than among white women before age 45 but lower at older ages. To explore whether differences in childbearing patterns can explain this observation, we assessed the relation of several childbearing variables to breast cancer risk in a large prospective cohort study of U.S. African-American women. METHODS: Black Women's Health Study participants were enrolled in 1995 and were followed by mailed questionnaires every 2 years (in 1997 and 1999). Of the 64 500 women enrolled, 56 725 (88%) completed at least one of the follow-up quesotionnaires. During 214 862 person-years of follow-up, participants reported 349 breast cancers, of which 128 were among women younger than 45 years and 221 were among women aged 45-70 yy"ears. Incidence rate ratios (IRRs) with 95% confidence intervals (CIs) were derived from age-stratified Cox regression models that adjusted for each of the childbearing variables (parity, age at first birth, and time since last birth). RESULTS: Compared with primiparity, high parity was associated with an increased risk of breast cancer among women younger than 45 years (IRR for four or more births = 2.4, 95% CI = 1.1 to 5.1) and a decreased risk among women aged 45 years and older (IRR = 0.5, 95% CI = 0.3 to 0.9). The IRR for late age at first birth compared with early age was 2.5 (95% CI = 1.1 to 5.8) among the younger women and was not elevated among older women. We found no statistically significant association of time since last birth with breast cancer risk among either younger or older women. CONCLUSIONS: Parity has a dual association with breast cancer risk in African-American women; among women younger than 45 years, parity is associated with an increased risk and among women 45 years and older it is associated with a decreased risk. This dual effect may explain some of the observed differences in breast cancer incidence rates among African-American and white women.
  30. Horton NJ, Saitz R, Laird NM and Samet JH. A method for modeling utilization data from multiple sources: application in a study of linkage to primary care. Health Services and Outcomes Research Methodology, 2002; 3:211-223.

    In many studies designed to measure health outcomes, information about subject utilization of health services is often obtained from multiple sources (or informants). Key methodological challenges in analyzing such data concern how they should best be represented and interpreted in statistical models. In the HELP (Health Evaluation and Linkage to Primary care) study, subjects without primary medical care undergoing alcohol or drug detoxification enrolled into a randomized controlled trial of a health evaluation intervention to link them with primary care. The outcome of interest was attendance at a primary care appointment (linkage to primary care) after discharge from the detoxification unit. Both self-report and administrative sources of linkage were collected. We apply methodology developed by Fitzmaurice et al. (American Journal of Epidemiology, 1995) to fit a single regression that allows inclusion of all multiple-source outcomes in a single multivariate regression analysis. This model allows testing for source differences in outcome, and estimation of different source effects where necessary, and includes data from subjects with partially observed source observations. These methods were applied to the analyosis of the HELP study using correlated survival regression models to assess the magnitude and significance of the relationship between predictor variables and linkage.
  31. Murphy JM, Horton NJ, Monson RR, Laird NM, Sobol AM, and Leighton AH. Cigarette smoking in relation to depression: historical trends from the Stirling County Study. American Journal of Psychiatry; 2003; 160(9):1663-1669.

    OBJECTIVE: Building on findings about the prevalence and incidence of depression over a 40-year period, the authors provide data on trends in cigarette smoking and associations with depression. METHOD: Data come from interviews with adult population samples (1952, 1970, and 1992) and followed cohorts (1952-1970 and 1970-1992). Logistic regression models and survival regressions were used to analyze the data. RESULTS: The associations between smoking and depression were small and nonsignificant in 1952 and 1970. In 1992, however, the odds that a smoker would be depressed were three times the odds that a nonsmoker would be depressed. The interaction between smoking and study year was significant, indicating that the association was limited to the most recent sample. In the cohort analysis, smoking at baseline did not predict the onset of depression, but subjects who became depressed were more likely to start or continue smoking and less likely to quit than those who never had a depression. CONCLUSIONS: In terms of population trends, the association between depression and cigarette smoking became prominent as the use of tobacco declined because of awareness of the risks involved. The findings about individuals followed over time suggest that those who became depressed were more involved with nicotine than those who never had a depression. The authors discuss hypotheses involving "self-medication," risk-taking, and changes in the social climate but conclude that the relationships between smoking and depression are probably multiple and complex.
  32. Palepu A, Horton NJ, Tibbetts N, Dukes K, Meli S and Samet JH. Substance abuse treatment and emergency department utilization among cohort of HIV-infected persons with alcohol problems. Journal of Substance Abuse Treatment, 2003; 25:37-42.

    We examined the association between substance abuse treatment services and emergency department (ED) utilization among participants in the HIV-Alcohol Longitudinal Cohort study of HIV-infected persons with a history of alcohol problems. A standardized questionnaire was administered to 350 subjects inquiring about demographics, substance use, use of substance abuse treatment services, and ED attendance. We operationally defined substance abuse treatment services as: stable, unstable, or none. The outcome, ED attendance, defined as presenting to an ED without subsequent hospitalization, occurred in 101 (28.8%) subjects in the past 6 months. Stable substance abuse treatment was significantly associated with lower odds of ED attendance (adjusted odds ratio; 95% CI): (0.5; 0.3-0.9). The relationship between substance abuse treatment and HIV-infected individuals' use of health care services merits additional attention. Substance abuse treatment may present an opportunity to impact health care utilization patterns of HIV-infected persons with a history of alcohol problems.
  33. Samet JH, Horton NJ, Traphagen ET, Lyon SM and Freedberg KA. Alcohol consumption and HIV progression: are they related? Alcoholism: Clinical and Experimental Research; 2003; 27(5):862-867.

    BACKGROUND: The relationship between alcohol consumption and HIV disease progression has received limited attention in the era of highly active antiretroviral therapy (HAART). METHODS: We assessed CD4 cell count, HIV RNA levels, and alcohol consumption in the past month, defined as none, moderate, and at risk, in 349 HIV-infected people with a history of alcohol problems. We investigated the relationship between alcohol consumption and HIV disease markers CD4 cell count and HIV RNA level, stratified by HAART use, using multivariable regression. RESULTS: No significant differences in CD4 cell count or HIV RNA level were found across the categories of alcohol consumption for patients who were not receiving HAART. However, among patients who were receiving HAART, log HIV RNA levels were significantly higher in those with moderate (2.17 copies/ml) and at-risk (2.73 copies/ml) alcohol use compared with none (1.73 copies/ml; p = 0.006). CD4 cell counts in those with moderate (368 cells/microl) and at-risk (360 cells/microl) alcohol use were lower than for subjects who reported none (426 cells/microl; p = 0.07). CONCLUSION: Among patients who have a history of alcohol problems and are receiving antiretroviral treatment, alcohol consumption was associated with higher HIV RNA levels and lower CD4 counts. No comparable association was found for similar patients who were not receiving HAART. Addressing alcohol use in HIV-infected patients, especially those who are receiving HAART, may have a substantial impact on HIV disease progression.
  34. Horton NJ, Lipsitz SR, and Parzen M. A potential for bias when rounding in multiple imputation. The American Statistician; 2003; 57(4):229-232.

    With the advent of general purpose packages that support multiple imputation for analyzing datasets with missing data (e.g., Solas, SAS PROC MI, and S-Plus 6.0), we expect much greater use of multiple imputation in the future. For simplicity, some imputation packages assume the joint distribution of the variables in the multiple imputation model is multivariate normal, and impute the missing data from the conditional normal distribution for the missing data given the observed data. If the possibly missing data are not multivariate normal (say, binary), imputing a normal random variable can yield implausible values. To circumvent this problem, a number of methods have been developed, including rounding the imputed normal to the closest observed value in the dataset. We show that this rounding can cause biased estimates of parameters, whereas if the imputed value is not rounded, no bias would occur. This article shows that rounding should not be used indiscriminately, and thus some caution should be exercised when rounding imputed values, particularly for dichotomous variables.
  35. Rosenberg L, Palmer JR, Wise LA, Horton NJ, Kumanyika SK, and Adams-Campbell LL. A prospective study of the effect of childbearing on weight gain in African-American women. Obesity Research; 2003; 11(12):1526-1535.

    OBJECTIVE: To prospectively assess the influence of bearing a first, second, or later child on weight gain among African-American women in the context of other risk factors. RESEARCH METHODS AND PROCEDURES: Data were obtained in a prospective follow-up study of African-American women from across the U.S. who are participants in the Black Women's Health Study. Postal questionnaires were used to collect baseline data in 1995 and follow-up data in 1997 and 1999. Parous and nulliparous women (11,196) (21 to 39 years old at baseline), of whom 1230 had a singleton birth during follow-up, are the subjects of the present analyses. We assessed change in BMI (kilograms per meter squared) in relation to childbearing during 4 years of follow-up, with use of multivariable linear regression to control for important risk factors. RESULTS: During 4 years of follow-up, the BMI of participants increased by an average of 1.6 kg/m2, equivalent to a weight gain of approximately 4.4 kg. Women who had a child during follow-up gained more weight than women who remained nulliparous, and those who had a first child gained more than those who had a second or later child. The weight gain associated with childbearing increased with increasing baseline BMI and was appreciable among heavier women. For example, among women with a baseline index of 36, the increase in BMI for women who bore a first child was 1.1 kg/m2 more than that of nulliparous women, equivalent to a difference in weight gain of approximately 3.0 kg. DISCUSSION: Childbearing is an important contributor to weight gain among African-American women.
  36. Seidman LJ, Pantelis C, Keshavan M, Faraone SV, Goldstein JM, Horton NJ, Makris N, Falkai P, Caviness VS and Tsuang MT. A review and a new report of medial temporal lobe dysfunction as a vulnerability indicator for schizophrenia: a MRI morphometric family study of the parahippocampal gyrus. Schizophrenia Bulletin; 2003; 9(4):803-30.

    A central question in schizophrenia research is which brain abnormalities are independent of psychosis and which evolve before and after psychosis begins. This question can be addressed by longitudinal neuroimaging studies beginning in the prodrome, but at present there is only one published study. We reviewed the literature on structural brain imaging in persons with chronic and first episode schizophrenia, nonpsychotic persons at genetic high risk, and persons thought to be at risk for imminent psychosis ("prodromal" persons). Medial temporal lobe (MTL), especially hippocampal, volume alterations are among the most robust brain vulnerabilities for schizophrenia. Because verbal declarative memory (VDM) deficits are prominent and the parahippocampal gyrus (PHG) is considered to be centrally involved with the hippocampus in VDM processing, we analyzed PHG data from a family study of schizophrenia. Patients with schizophrenia and nonpsychotic relatives from "multiplex" families (families with multiple persons with schizophrenia) had significantly smaller right parahippocampal anterior (PHa) volumes than controls. Marginally significant findings were observed for the left PHa. Unexpectedly, relatives from "simplex" families (families with only one person with schizophrenia) had significantly larger PH posterior volumes than controls and did not differ from controls on PHa. Results provide some support for the hypothesis that the vulnerability to schizophrenia includes abnormal volumes of the PHG. These data provide additional support for the hypothesis that some MTL abnormalities in schizophrenia are independent of psychosis, at least in families with presumably high genetic loading. Implications of genetic risk studies for prodromal research are discussed.
  37. Samet JH, Phillips SJ, Horton NJ, Traphagen ET and Freedberg KA. Detecting alcohol problems in HIV-infected patients: use of the CAGE questionnaire. AIDS Research and Human Retroviruses; 2004; 20(2):151-5.

    The HIV epidemic has been consistently associated with injection drug use and crack cocaine, but alcohol problems in HIV-infected persons are less well described. Our objectives were 2-fold: (1) to assess the prevalence of alcohol problems in HIV-infected patients initiating medical care; and (2) to determine the positive predictive value of the CAGE questionnaire for alcohol abuse or dependence in HIV-infected patients. Between July 1997 and October 2000, we assessed a consecutive series of patients who were establishing primary care for HIV infection (clinic sample), using an established alcohol screening test, the CAGE questionnaire. In addition, we enrolled other HIV-infected patients, including some of the clinic sample, who had two or more positive responses to the CAGE questionnaire into a longitudinal cohort (cohort sample), performed a diagnostic interview for lifetime history of alcohol abuse and dependence, and determined the positive predictive value of CAGE for alcohol diagnoses. In the clinic sample (n = 664), 42% (276 of 664) had two or more positive responses to the four CAGE questions. In the cohort sample (n = 141), 95% (134 of 141) met DSM-IV criteria for diagnosis of lifetime alcohol abuse or dependence. For patients initiating HIV primary care, a history of alcohol problems is very common. The CAGE questionnaire identifies a lifetime history of alcohol abuse or dependence in HIV-infected patients. Routine screening for alcohol problems should be performed in all patients entering HIV medical care and the CAGE questions are useful in this setting.
  38. Liddicoat RV, Horton NJ, Urban R, Maier E, Christiansen D, and Samet JH. Assessing missed opportunities for HIV testing in medical settings. Journal of General Internal Medicine; 2004; 19(4):349-356.

    BACKGROUND: Many HIV-infected persons learn about their diagnosis years after initial infection. The extent to which missed opportunities for HIV testing occur in medical evaluations prior to one's HIV diagnosis is not known. DESIGN: We performed a 10-year retrospective chart review of patients seen at an HIV intake clinic between January 1994 and June 2001 who 1). tested positive for HIV during the 12 months prior to their presentation at the intake clinic and 2). had at least one encounter recorded in the medical record prior to their HIV-positive status. Data collection included demographics, clinical presentation, and whether HIV testing was recommended to the patient or addressed in any way in the clinical note. Prespecified triggers for physicians to recommend HIV testing, such as specific patient characteristics, symptoms, and physical findings, were recorded for each visit. Multivariable logistic regression was used to identify factors associated with missed opportunities for discussion of HIV testing. Generalized estimating equations were used to account for multiple visits per subject. RESULTS: Among the 221 patients meeting eligibility criteria, all had triggers for HIV testing found in an encounter note. Triggers were found in 50% (1702/3424) of these 221 patients' medical visits. The median number of visits per patient prior to HIV diagnosis to this single institution was 5; 40% of these visits were to either the emergency department or urgent care clinic. HIV was addressed in 27% of visits in which triggers were identified. The multivariable regression model indicated that patients were more likely to have testing addressed in urgent care clinic (39%), sexually transmitted disease clinic (78%), primary care clinics (32%), and during hospitalization (47%), compared to the emergency department (11%), obstetrics/gynecology (9%), and other specialty clinics (10%) (P <.0001). More recent clinical visits (1997-2001) were more likely to have HIV addressed than earlier visits (P <.0001). Women were offered testing less often than men (P =.07). CONCLUSIONS: Missed opportunities for addressing HIV testing remain unacceptably high when patients seek medical care in the period before their HIV diagnosis. Despite improvement in recent years, variation by site of care remained important. In particular, the emergency department merits consideration for increased resource commitment to facilitate HIV testing. In order to detect HIV infection prior to advanced immunosuppression, clinicians must become more aware of clinical triggers that suggest a patient's increased risk for this infection and lower the threshold at which HIV testing is recommended.
  39. Krupitsky E, Zvartau E, Karandashova G, Horton NJ, Schoolwerth K, Bryant K and Samet JH. The onset of HIV infection in the Leningrad region of Russia: A focus on drug and alcohol dependence. HIV Medicine; 2004; 5:30-33.

    OBJECTIVES : Within 5 years, 5 million Russians may be infected with HIV. Currently, injection drug use is the major risk factor for HIV. As Russia's alcohol consumption per capita is among the highest in the world, alcohol-associated behaviours may be an important contributor to the HIV epidemic. Our objective was to examine the prevalence of HIV infection among drug- and alcohol-dependent patients in a regional narcology hospital and in the general population in Leningrad. METHODS : All patients in the Narcology Hospital, Leningrad Regional Center of Addictions (LRCA), were tested for HIV antibody between 1997 and 2001. We reviewed these clinical records (i.e. serostatus, gender, age, and addiction) and data from the HIV/AIDS Center in the Leningrad Region (1997-2001). RESULTS : From 1997 to 2001, HIV prevalence at the LRCA increased from 0 to 12.7% overall, 33.4% among drug-dependent patients and 1.2% among alcohol-dependent patients. During the same 5-year period (1997-2001), 2826 persons were registered at the HIV/AIDS Center: 6, 6, 51, 780, and 1983 persons in 1997, 1998, 1999, 2000 and 2001, respectively. CONCLUSIONS : HIV infection is exploding in the Leningrad Region, currently in injection drug users (IDUs) but potentially more broadly. The known high per capita alcohol intake in Russia heightens concern regarding the sexual transmission of HIV. Interventions to prevent such a development should include use, and assessment of the effectiveness, of known HIV prevention measures for at-risk and infected individuals.
  40. Ehrenstein V, Horton NJ, and Samet JH. Inconsistent condom use among HIV-infected patients with alcohol problems. Drug and Alcohol Dependence; 2004; 73(2):159-166.

    BACKGROUND: Unsafe sexual behavior is common among persons with negative or unknown HIV status and it is augmented by alcohol use in some populations. We examined the association between alcohol consumption level (abstinent, moderate, at-risk) and inconsistent condom use in a cohort of HIV-infected individuals with a history of alcohol problems. METHODS: Subjects (n=345) had up to seven structured interviews over 36 months. Identical questions on alcohol consumption and inconsistent condom use were asked at each interview. We used generalized estimating equations (GEE) multivariate logistic regression for repeated measurements analysis. We adjusted for potential confounding factors and explored possible interactions. RESULTS: At baseline, 132 (38%) participants reported inconsistent condom use. We detected a significant (P=0.0002) interaction between alcohol consumption and injection drug use (IDU) variables. Among active injection drug users, at-risk drinking was associated with inconsistent condom use, adjusted odds ratio (OR; 95% confidence interval) 4.3 (1.5, 12.2). Among those who did not inject drugs, at-risk drinking and inconsistent condom use were not associated, 0.7 (0.4, 1.3). Inconsistent condom use was more common among women, those believing condoms to be 'a hassle', and persons living with a partner. CONCLUSION: In HIV-infected drug-injecting individuals, excessive use of alcohol is associated with unsafe sexual practices.
  41. Saitz R, Larson MJ, Horton NJ, Winter M and Samet JH. Linkage with primary medical care in a prospective cohort of adults with addictions in inpatient detoxification: room for improvement. Health Services Research; 2004; 39(3):587-606.

    Objective. To identify patient characteristics and health care experiences associated with primary care linkage after alcohol or drug detoxification. Data Sources/Study Setting. Primary data collected over two years. Subjects were adults without primary medical care, in an urban residential detoxification program. Study Design. A prospective cohort study in the context of a randomized trial of a linkage intervention, and an expansion of Medicaid benefits. Data Collection/Extraction Methods. Data were collected by interview assessment of predisposing, enabling, and illness variables. Linkage was defined as self-report of at least one visit with a primary care clinician during follow-up. Principal Findings. Of 400 subjects, 63 percent linked with primary medical care. In a multivariable model adjusting for randomization assignment, predisposing, enabling, and illness variables, women, those with no recent incarceration, those with support for abstinence by family or friends, and those who had visited a medical clinic or physician recently were significantly more likely to link with primary care. Those with health insurance during follow-up were also more likely to link. Recent mental health or addictions treatment utilization and health status were not associated with linkage. Conclusions. A substantial proportion of adults with addictions do not link with primary medical care. These data suggest that efforts could be focused on those least likely to link, that contacts with mental health and addictions treatment providers are underutilized opportunities for these efforts, and that health policy changes such as expanding health insurance benefits may improve entry of substance-dependent patients into primary medical care.
  42. Palepu A, Horton NJ, Tibbetts N, Meli S, and Samet JH. Uptake and adherence to highly active antiretroviral therapy among HIV-infected persons with alcohol and other substance use problems: the impact of substance abuse treatment. Addiction; 2004; 99(3):361-8.

    AIM: We examined the association of substance abuse treatment with uptake, adherence and virological response to highly active antiretroviral therapy (HAART) among HIV-infected people with a history of alcohol problems. DESIGN: Prospective cohort study. METHODS: A standardized questionnaire was administered to 349 HIV-infected participants with a history of alcohol problems regarding demographics, substance use, use of substance abuse treatment and uptake of and adherence to HAART. These subjects were followed every 6 months for up to seven occasions. We defined substance abuse treatment services as any of the following in the past 6 months: 12 weeks in a half-way house or residential facility; 12 visits to a substance abuse counselor or mental health professional; or participation in any methadone maintenance program. Our outcome variables were uptake of antiretroviral therapy, 30-day self-reported adherence and HIV viral load suppression. FINDINGS: At baseline, 59% (205/349) of subjects were receiving HAART. Engagement in substance abuse treatment was independently associated with receiving antiretroviral therapy (adjusted OR; 95% CI: 1.70; 1.03-2.83). Substance abuse treatment was not associated with 30-day adherence or HIV viral load suppression. More depressive symptoms (0.48; 0.32-0.78) and use of drugs or alcohol in the previous 30 days (0.17; 0.11-0.28) were associated with worse 30-day adherence. HIV viral load suppression was positively associated with higher doses of antiretroviral medication (1.29; 1.15-1.45) and older age (1.04; 1.00-1.07) and negatively associated with use of drugs or alcohol in the previous 30 days (0.51; 0.33-0.78). CONCLUSION: Substance abuse treatment was associated with receipt of HAART; however, it was not associated with adherence or HIV viral load suppression. Substance abuse treatment programs may provide an opportunity for HIV-infected people with alcohol or drug problems to openly address issues of HIV care including enhancing adherence to HAART.
  43. Murphy JM, Horton NJ, Laird NM, Monson RR, Sobol AM and Leighton AH. Anxiety and depression: a 40-year perspective on relationships regarding prevalence, distribution, and comorbidity. Acta Psychiatrica Scandinavia; 2004; 109(5):355-75.

    OBJECTIVE: Building on a report about the prevalence of depression over time, this paper examines historical trends regarding anxiety in terms of its prevalence, its distribution by age and gender, and its comorbidity with depression. Methods for conducting such time trend analysis are reviewed. METHOD: Representative samples of adults were selected and interviewed in 1952, 1970, and 1992. Logistic regressions were used for statistical analysis. RESULTS: Although twice as common as depression, the prevalence of anxiety was equally stable. Anxiety was consistently and significantly more characteristic of women than men. A re-distribution of rates in 1992 indicated that depression but not anxiety had significantly increased among younger women (P = 0.03). Throughout the study, approximately half of the cases of anxiety also suffered depression. CONCLUSION: The relationships between anxiety and depression remained similar over time with the exception that depression came to resemble anxiety as a disorder to which women were significantly more vulnerable than men. Social and historical factors should be investigated to assess their relevance to this change.
  44. Samet JH, Horton NJ, Meli S, Freedberg KA and Palepu A. Alcohol consumption and antiretroviral adherence among HIV-infected persons with alcohol problems. Alcoholism: Clinical and Experimental Research; 2004; 28(4):572-577.

    BACKGROUND: Alcohol abuse has been associated with poor adherence to highly active antiretroviral therapy (HAART). We examined the relative importance of varying levels of alcohol consumption on adherence in HIV-infected patients with a history of alcohol problems. METHODS: We surveyed 349 HIV-infected persons with a history of alcohol problems at 6-month intervals. Of these subjects, 267 were taking HAART at one or more time periods during the 30-month follow-up period. Interviews assessed recent adherence to HAART and past month alcohol consumption, defined as "none", "moderate", and "at risk". We investigated the relationship between adherence to HAART and alcohol consumption at baseline and at each subsequent 6-month follow-up interval using multivariable longitudinal regression models, while controlling for potential confounders. RESULTS: Among the 267 HIV-infected persons with a history of alcohol problems who were receiving HAART, alcohol consumption was the most significant predictor of adherence (p < 0.0001), with better adherence being associated with recent abstinence from alcohol, compared with at-risk level usage (odds ratio = 3.6, 95% confidence interval = 2.1-6.2) or compared with moderate usage (odds ratio = 3.0, 95% confidence interval = 2.0-4.5). CONCLUSIONS: Any alcohol use among HIV-infected persons with a history of alcohol problems is associated with worse HAART adherence. Addressing alcohol use in HIV-infected persons may improve antiretroviral adherence and ultimately clinical outcomes.
  45. Horton NJ and Fitzmaurice GM. Regression analysis of multiple source data from complex survey samples. Statistics in Medicine; 2004; 23(18):2911-33.

    In this tutorial, we describe regression-based methods for analysing multiple source data arising from complex sample survey designs. We use the term 'multiple-source' data to encompass all cases where data are simultaneously obtained from multiple informants, or raters (e.g. self-reports, family members, health care providers, administrators) or via different/parallel instruments, indicators or methods (e.g. symptom rating scales, standardized diagnostic interviews, or clinical diagnoses). We review regression models for analysing multiple source risk factors or multiple source outcomes and show that they can be considered special cases of generalized linear models, albeit with correlated outcomes. We show how these methods can be extended to handle the common survey features of stratification, clustering, and sampling weights. We describe how to fit regression models with multiple source reports derived from complex sample surveys using general purpose statistical software. Finally, the methods are illustrated using data from two studies: the Stirling County Study and the Eastern Connecticut Child Survey.
  46. Wines JD, Saitz R, Horton NJ, Lloyd-Travaglini C and Samet JH. Suicidal behavior, drug use and depressive symptoms after detoxification: a 2-year prospective study. Drug and Alcohol Dependence, 2004; 76:S21-S29.

    INTRODUCTION: Individuals with substance-related disorders are at increased risk for suicidal behavior. Identifying those at higher risk for suicide among this population is difficult and informed mainly on the basis of cross-sectional data. METHODS: We examined factors associated with drug-related suicidal behavior using multivariable regression analyses in a 2-year prospective study of 470 inpatients enrolled from an unlocked, detoxification unit. Suicidal behavior included suicidal ideation (SI) and suicide attempt (SA). RESULTS: Lifetime prevalence for SI was 28.5%, and for SA, 21.9%. During the 2-year follow-up, 19.9% of the sample endorsed suicidal ideation and 6.9% reported a suicide attempt. Correlates of lifetime suicidal behavior included younger age, female, Hispanic, greater depressive symptoms, past sexual abuse, and problem sedative or alcohol use. Factors associated with suicidal behavior at follow-up included past suicidal behavior, more depressive symptoms, and more frequent benzodiazepine and alcohol use. Cocaine and heroin use did not reach statistical significance. CONCLUSIONS: Suicidal behavior is common among individuals with substance-related disorders. Differences in "suicide potential" may exist between drug categories with CNS depressants increasing the risk. These findings highlight the importance of addressing the recurrent 'suicide risk' of patients with substance-related disorders and regular monitoring for changes in depressive symptoms and drug use. Based on the prevalence and severity of this problem, the role of universal suicide screening of individuals with substance-related disorders merits greater attention.
  47. Horton NJ, Brown ER, and Qian L (SC '05 undergraduate co-author). Use of R as a toolbox for mathematical statistics exploration. The American Statistician, 2004; 58(4):343-357.

    The R language, a freely available environment for statistical computing and graphics is widely used in many fields. This expert-friendly system has a powerful command language and programming environment, combined with an active user community. We discuss how R is ideal as a platform to support experimentation in mathematical statistics, both at the undergraduate and graduate levels. Using a series of case studies and activities, we describe how R can be utilized in a mathematical statistics course as a toolbox for experimentation. Examples include the calculation of a running mean, maximization of a non-linear function, resampling of a statistic, simple Bayesian modeling, sampling from multivariate normal and estimation of power. These activities, often requiring only a few dozen lines of code, offer the student the opportunity to explore statistical concepts and experiment. In addition, they provide an introduction to the framework and idioms available in this rich environment.
  48. Olsen Y, Alford DP, Horton NJ, and Saitz R. Addressing smoking cessation in methadone programs. Journal of Addictive Diseases; 2005; 24(2):33-48.

    Little is known about how well methadone programs address smoking cessation. We describe the frequency of smoking cessation counseling, and factors affecting its provision in methadone programs. We conducted a cross-sectional survey of methadone patients and their counselors. Of 575 patients, 76% were eligible smokers. Although only 48% of patients reported receiving smoking cessation counseling within the previous six months, 97% of counselors reported providing it (p < 0.0001). Time with one counselor was significantly associated with patient report of receiving smoking cessation counseling (OR 1.19 [95% CI 1.04-1.36]). Although addiction counseling is required in methadone programs, nicotine addiction is addressed less than half the time. Methadone programs should prioritize the provision of effective smoking cessation and facilitate continuity of patient-counselor relationships.
  49. Krupitsky EM, Horton NJ, Wiliams EC, Lioznov D, Kuznetsova M, Zvartau E and Samet JH. Alcohol use and HIV risk behaviors among HIV-infected hospitalized patients in St. Petersburg, Russia. Drug and Alcohol Dependence, 2005: 79:251-256.

    PURPOSE: Russia has high per capita alcohol consumption and an injection-drug-use-driven HIV epidemic. However, the role of alcohol in the spread of HIV infection in Russia is largely unexplored. Thus, we assessed recent alcohol use and associated HIV risk behaviors among HIV-infected persons in St. Petersburg, Russia. METHODS: We recruited HIV-infected hospitalized patients from the Botkin Infectious Disease Hospital between June 2001 and March 2002. Interviewers assessed alcohol and drug use with the addiction severity index (ASI) and sex- and drug-risk behaviors with the risk assessment battery (RAB). Lifetime abuse or dependence diagnoses for alcohol and drugs were established by a physician with addiction medicine training. RESULTS: Among 201 subjects, diagnoses of abuse or dependence (AB/DEP) were common: 9% (19/201) had only alcohol AB/DEP; 39% (78/201) had alcohol and drug AB/DEP; 47% (95/201) had only drug AB/DEP; and 4% (9/201) had no diagnosis of alcohol or drug AB/DEP. Sex- and drug-risk behaviors varied significantly by substance use diagnosis. Subjects with any alcohol AB/DEP had higher sex-risk RAB scores than those with drug only AB/DEP (6.1 versus 3.9, p<.0001). Among subjects with any diagnosis of drug AB/DEP, having in addition an alcohol diagnosis was associated with unclean needle use in the last six months (33% (26/78) versus 21% (20/95), p=0.08). CONCLUSIONS: Lifetime alcohol diagnoses of abuse or dependence were present in nearly one-half of hospitalized HIV-infected patients in St. Petersburg, Russia and were associated with significantly higher sex-risk behaviors and borderline significantly higher drug-risk behaviors. As HIV infection spreads rapidly in Russia and Eastern Europe, these data support the need for HIV risk-reduction interventions in alcohol abusing populations and raise the potential of benefit by addressing alcohol use in HIV-infected populations.
  50. Kraemer KL, Roberts MS, Horton NJ, Palfai T, Samet JH, Freedner N, Tibbetts N and Saitz R. Health utility ratings for a spectrum of alcohol-related health states. Medical Care, 2005; 43(6):541-550.

    BACKGROUND: Preference-based utility ratings for health conditions are important components of cost-utility analyses and population burden of disease estimates. However, utility ratings for alcohol problems have not been determined. OBJECTIVES:: The objectives of this study were to directly measure utility ratings for a spectrum of alcohol-related health states and to compare different methods of utility measurement. DESIGN, SETTING, AND SUBJECTS:: The authors conducted a cross-sectional interview of 200 adults from a clinic and community sample. METHODS:: Subjects completed computerized visual analog scale (VAS), time tradeoff (TTO), and standard gamble (SG) utility measurement exercises for their current health, a blindness scenario, and for 6 alcohol-related health state scenarios presented in random order. The main outcome measures were the utility ratings, scaled from 0 to 1, and anchored by death (0) and perfect health (1). RESULTS:: The 200 subjects were middle-aged (mean, 41 +/- 14 years), 61% women, and racially diverse (48% black, 43% white). Utility ratings decreased as the severity of the alcohol-related health state increased, but differed significantly among the VAS, TTO, and SG methods within each health state. Adjusted mean (95% confidence interval) utility ratings for alcohol dependence (VAS, 0.38 [0.34-0.41]; TTO, 0.54 [0.48-0.60]; SG, 0.68 [0.63-0.73]) and alcohol abuse (VAS, 0.53 [0.49-0.56]; TTO, 0.71 [0.65-77]; SG, 0.76 [0.71-0.81]) were significantly lower than utility ratings for nondrinking, moderate drinking, at-risk drinking, current health, and blindness. CONCLUSIONS:: Utility ratings for alcohol-related health states decrease as the severity of alcohol use increases. The low utility ratings for alcohol abuse and alcohol dependence are similar to those reported for other severe chronic medical conditions.
  51. Kertesz SG, Larson MJ, Horton NJ, Winter M, Saitz R and Samet JH. Homeless chronicity and health-related quality of life trajectories among adults with addictions. Medical Care, 2005; 43(6):574-585.

    BACKGROUND:: New federal initiatives target funds toward chronically homeless as distinct from other homeless persons. Few data exist, however, to substantiate the implications of chronic homelessness for major health outcomes. OBJECTIVES:: Using data from a 2-year cohort of addicted persons, we tested whether changes in mental and physical health-related quality of life (HRQOL) differed according to homeless chronicity. METHODS:: Using self-reported homelessness, we classified subjects as chronically homeless (CH; n = 60), transitionally homeless (TRANS; n = 108), or as housed comparison subjects (HSD; n = 106). The Short Form-36 Health Survey, administered at baseline and 2 follow-ups over a period of 2 years, provided a Mental Component Summary (MCS) and a Physical Component Summary (PCS) for HRQOL. Mixed model linear regression was used to test the association between housing status, MCS, and PCS. Additional models assessed whether medical, psychiatric, addiction, and social support measures could account for HRQOL differences. RESULTS:: All subjects had low MCS scores at study entry (mean, 31.2; SD, 12.6). However, there was a significant housing status-by-time interaction (P = 0.01). At final follow-up, CH and TRANS subjects had lower adjusted MCS scores than HSD subjects (33.4, 38.8, and 43.7 for the 3 groups, respectively; all P < = 0.01). By contrast, housing status and PCS were not significantly associated (P = 0.19). Medical, psychiatric, addiction, and social support variables had significant associations with MCS, and their inclusion in the regression reduced the apparent effect of housing status on MCS. CONCLUSIONS:: Chronic homelessness was associated with especially poor mental but not physical HRQOL over time. These findings reinforce a new typology of homelessness.
  52. Samet JH, Horton NJ, Meli S, Dukes K, Tripp T, Sullivan L and Freedberg KA. A randomized controlled trial to enhance antiretroviral therapy adherence in patients with a history of alcohol problems. Antiviral Therapy, 2005; 10:83-93.

    OBJECTIVE: To assess the effectiveness of an individualized multicomponent intervention to promote adherence to antiretroviral therapy (ART) in a cohort of HIV-infected individuals with a history of alcohol problems. DESIGN: We conducted a randomized controlled trial to compare the usual medical follow-up with an adherence intervention. SETTING: The principal enrolment site was Boston Medical Center, a private, not-for-profit, academic medical institution. SUBJECTS: HIV-infected patients with a history of alcohol problems on ART. A total of 151 were enrolled and 141 (93%) were assessed at follow-up. Intervention: A nurse, trained in motivational interviewing, completed the following over 3 months in four encounters: addressed alcohol problems; provided a watch with a programmable timer to facilitate pill taking; enhanced perception of treatment efficacy; and delivered individually tailored assistance to facilitate medication use. MAIN OUTCOME MEASURES: Prior 30-day adherence > or =95%, prior 3-day adherence of 100%, CD4 cell count, HIV RNA and alcohol consumption, each at both short- and long-term follow-up. RESULTS: At follow-up, no significant differences in medication adherence, CD4 cell count, HIV RNA or alcohol consumption were found (all P values >0.25). CONCLUSIONS: A multicomponent intervention to enhance adherence among HIV-infected individuals with a history of alcohol problems was not associated with changes in medication adherence, alcohol consumption or markers of HIV disease progression. The failure to change adherence in a group at high risk for poor adherence, despite utilizing an intensive individual-focused patient intervention, supports the idea of addressing medication adherence with supervised medication delivery or markedly simplified dosing regimens.
  53. Webb KE, Horton NJ, and Katz DL. Parental IQ and cognitive development of malnourished Indonesian children. European Journal of Clinical Nutrition, 2005; 59(4):618-620.

    A cross-sectional study of children in West Kalimantan, Indonesia, was conducted to examine the relationship between malnutrition history, child IQ, school attendance, socioeconomic status, parental education and parental IQ. In unadjusted analyses, severely stunted children had significantly lower IQ scores than mild-moderately stunted children. This effect was significant when stunting, school attendance and parental education were included in multivariable models but was attenuated when parental IQ was included. Our research underscores the importance of accounting for parental IQ as a critical covariate when modeling the association between childhood stunting and IQ.
  54. Liebschutz JM, Geier JL, Horton NJ, Chuang CH, and Samet JH. Physical and sexual violence and health care utilization in HIV-infected persons with alcohol problems. AIDS Care, 2005; 17(5):566-578.

    We examined interpersonal violence and its association with health care utilization and substance use severity among a cohort of 349 HIV-infected men and women with histories of alcohol problems assessed biannually up to 36 months. Data included demographics, lifetime interpersonal violence histories, age at first violence exposure, recent violence (prior six months), substance use severity and health care utilization (ambulatory visits, Emergency Department (ED) visits, hospitalizations) and adherence to HIV medication. Kaplan-Meier survival curves estimated the proportion of subjects experiencing recent violence. Generalized estimating equation regression models evaluated the relationship between recent violence, utilization and substance use severity over time, controlling for demographics, CD4 counts and depressive symptoms. Subject characteristics included: 79% male; mean age 41 years; 44% black, 33% white and 23% other. Eighty percent of subjects reported lifetime interpersonal violence: 40% physical violence alone, and 40% sexual violence with or without physical violence. First violence occurred prior to age 13 in 46%. Twenty-four (41%) of subjects reported recent violence by 24 and 36 months, respectively. In multivariate analyses, recent violence was associated with more ambulatory visits, ED visits and hospitalizations and worse substance use severity, but not medication adherence. Due to the high incidence and associated increased health care services utilization, violence prevention interventions should be considered for HIV-infected patients with a history of alcohol problems.
  55. Saitz R, Horton NJ, Larson MJ, Winter M and Samet JH. Primary medical care and reductions in addiction severity: a prospective cohort study. Addiction, 2005; 100(1):70-8.

    AIMS: To assess whether receipt of primary medical care can lead to improved outcomes for adults with addictions. DESIGN: We studied a prospective cohort of adults enrolled in a randomized trial to improve linkage with primary medical care. METHODS: Subjects at a residential detoxification unit with alcohol, heroin or cocaine as a substance of choice, and no primary medical care were enrolled. Receipt of primary medical care was assessed over 2 years. Outcomes included (1) alcohol severity, (2) drug severity and (3) any substance use. FINDINGS: For the 391 subjects, receipt of primary care (> or = 2 visits) was associated with a lower odds of drug use or alcohol intoxication (adjusted odds ratio (AOR) 0.45, 95% confidence interval (CI) 0.29-0.69, 2 d.f. chi(2)P = 0.002). For 248 subjects with alcohol as a substance of choice, alcohol severity was lower in those who received primary care [predicted mean Addiction Severity Index (ASI) alcohol scores for those reporting > or = 2, 1 and 0 visits, respectively, 0.30, 0.26 and 0.34, P = 0.04]. For 300 subjects with heroin or cocaine as a substance of choice, drug severity was lower in those who received primary care (predicted mean ASI drug scores for those reporting > or = 2, 1 and 0 visits, respectively, 0.13, 0.15 and 0.16, P = 0.01). CONCLUSIONS: Receipt of primary medical care is associated with improved addiction severity. These results support efforts to link patients with addictions to primary medical care services.
  56. Palepu A, Raj A, Horton NJ, Tibbetts N, Meli S, and Samet JH. Substance abuse treatment and risk behaviors among HIV-infected persons with alcohol problems. Journal of Substance Abuse Treatment, 2005; 28(1):3-9.

    We examined the association of substance abuse treatment with sexual and drug use risk behaviors among 349 HIV-infected persons with a history of alcohol problems using a standardized questionnaire regarding sexual and drug use risk behaviors, demographics, substance use, and use of substance abuse treatment. We defined substance abuse treatment services as any of the following in the past 6 months: 12 weeks in a half-way house or residential facility; 12 visits to a substance abuse counselor or mental health professional; day treatment for at least 30 days; or participation in any methadone maintenance program. Our three outcome variables of high-risk behavior were the Risk Assessment Battery sex-risk and drug-risk scores and high-risk sex behavior which included any of the following: inconsistent condom use; having more than one sexual partner; and exchanging sex for money or drugs. Although sexual risk was high (51%) in our HIV-infected cohort, engagement in substance abuse treatment was not independently associated with lower frequency of any of our measures of high- risk behaviors. Although the opportunity exists to address HIV risk behaviors in the setting of substance abuse treatment, effective institutionalization of this challenging behavior change effort has not yet been realized.
  57. Horton NJ and Shapiro EC (SC '04 undergraduate co-author). Statistical sleuthing during epidemics: maternal influenza and schizophrenia. Chance, 2005; 18(1):11-18.

  58. Shanahan CW, Lincoln A, Horton NJ, Saitz R, Larson MJ and Samet JH. The relationship of depressive symptoms and mental health to detoxification unit readmission. Journal of Substance Abuse Treatment, 2005;29:117-123.

    To better understand residential detoxification use, we assessed the roles of depressive symptoms (DS) and mental health functioning (MHF) on repeat detoxification. A prospective cohort of residential detoxification patients (N=400) without primary medical care was followed over 2 years at 6-month intervals. Subsequent detoxification admissions were examined using a statewide administrative database and DS (Center for Epidemiologic Studies Depression Scale) and MHF (SF-36 mental component summary subscale) measurements at follow-up. Incidence rate ratios of return to detoxification were estimated using multivariable longitudinal Poisson regression. In separate analyses, greater DS and worse MHF predicted higher detoxification use rates. Clinically significant worsening (10 points) of DS and MHF on objective scales predicted a 20% increased rate of detoxification readmission. Male sex, heroin as a problem substance, and race/ethnicity each predicted detoxification use. These data suggest that identifying individuals with DS or worse MHF after detoxification may provide opportunities for clinical intervention to reduce recurrent residential detoxification.
  59. Palepu A, Horton NJ, Tibbetts N, Meli S and Samet JH. Substance abuse treatment and hospitalization among a cohort of HIV-infected persons with alcohol problems. Alcohol Clinical and Experimental Research, 2005; 29(3):389-394.

    BACKGROUND: We examined the association of substance abuse treatment services on hospitalization among participants in the HIV-Alcohol Longitudinal Cohort (HIV-ALC) study of HIV-infected individuals with a history of alcohol problems. METHODS: A standardized questionnaire that inquired about demographics, substance use, use of substance abuse treatment services, and hospitalization was administered to 349 HIV-ALC participants. We defined substance abuse treatment services as any of the following in the past 6 months: 12 weeks in a half-way house or residential facility, 12 visits to a substance abuse counselor or mental health professional, or participation in any methadone maintenance program. RESULTS: Almost one third of this cohort were hospitalized in the past 6 months. Substance abuse treatment was not significantly associated with hospitalization [adjusted odds ratio (AOR) 1.0; 95% confidence interval (CI) 0.7-1.5), whereas homelessness (AOR 2.3; 95% CI 1.5-3.6), injection drug use (AOR 1.7; 95% CI 1.0-2.7), severity of alcohol dependence (AOR 1.02; 95% CI 1.00-1.05), CD4 cell count (AOR 0.999; 95% CI 0.998-1.00), and HIV RNA (AOR 1.1; 95% CI 1.0-1.2) were independently associated with increased odds of hospitalization over time. CONCLUSIONS: Engagement in substance abuse treatment was not associated with a decrease in hospital use by HIV-infected individuals with a history of alcohol problems. The period of substance abuse treatment may present an opportunity to address health care utilization patterns of HIV-infected individuals.
  60. Rosenberg L, Wise LA, Palmer JR, Horton NJ, Adams-Campbell LL. A multilevel study of socioeconomic predictors of regular mammography use among African-American women. Cancer Epidemiology, Biomarkers & Prevention, 2005; 14(11):2628-2633.

    BACKGROUND: Little is known about the predictors of adherence to mammography guidelines by African-American women. We assessed individual-level and group-level socioeconomic predictors of regular mammography use in a large cohort of African-American women. METHODS: We included 14,706 participants in the Black Women's Health Study who were 40 to 69 years of age at baseline in 1995. Data were obtained through mail questionnaires at entry and biennially during three 2-year follow-up periods until 2001. We linked the women's addresses to census block (neighborhood) socioeconomic data. With multilevel multivariable regression analysis, we assessed the relation of individual-level and neighborhood socioeconomic factors to "regular" mammography use (use in all three follow-up periods) relative to nonregular use, with control for other predictors. RESULTS: Most participants had health insurance and almost half had graduated college. Having health insurance was the socioeconomic variable most strongly associated with regular mammography use (odds ratio, 3.59; 95% confidence interval 3.02-4.28); the association was present at all levels of educational attainment, household income, and neighborhood socioeconomic status. Regular mammography use increased with individual household income: odds ratio, 2.00 (95% confidence interval, 1.58-2.53) for household income > USD $100,000 relative to < USD $15,001. Regular mammography use did not vary across level of education. Higher neighborhood socioeconomic status was significantly associated with regular mammography use before, but not after, control for household income. CONCLUSIONS: Access to health insurance is strongly associated with regular mammography use among African-American women, even at higher levels of educational attainment and socioeconomic status. Neighborhood socioeconomic characteristics do not materially influence regular use.
  61. Chuang CH, Liebschutz JM, Horton NJ and Samet JH. Association of victimization with unsafe condom use in HIV-infected persons. AIDS and Behavior, 2006;10(2):201-207.

    The association of violence victimization with current condom use in HIV-infected persons was examined in this cross-sectional study. The HIV-Alcohol Longitudinal Cohort (HIV-ALC) recruited HIV-infected participants with a history of alcohol problems. Interviews assessed violence histories and current sexual behaviors. Of the 349 participants (79% men), 38% reported inconsistent condom use and 80% reported a violence history. Lifetime sexual violence was reported by 40% and lifetime physical violence (without sexual violence) by 40%. Participants reporting lifetime sexual violence had greater odds of inconsistent condom use than participants without any history of violence. A history of childhood sexual violence was also associated with greater odds of inconsistent condom use than participants without a history of childhood sexual violence. A history of sexual violence may in part explain HIV-infected persons' greater risk for transmitting HIV through high-risk sexual behaviors.
  62. Horton NJ and Switzer SS (SC '06 undergraduate co-author). Statistical methods in the Journal (research letter). New England Journal of Medicine, 2005;353(13):1977-1979.

  63. Saitz R, Freedner N, Palfai T, Horton NJ and Samet JH. The severity of unhealthy alcohol use in hospitalized medical patients. The spectrum is narrow. Journal of General Internal Medicine, 2006;21(4):381-385.

    BACKGROUND: Professional organizations recommend screening and brief intervention for unhealthy alcohol use; however, brief intervention has established efficacy only for people without alcohol dependence. Whether many medical inpatients with unhealthy alcohol use have nondependent use, and thus might benefit from brief intervention, is unknown. OBJECTIVE: To determine the prevalence and spectrum of unhealthy alcohol use in medical inpatients. DESIGN: Interviews of medical inpatients (March 2001 to June 2003). SUBJECTS: Adult medical inpatients (5,813) in an urban teaching hospital. MEASUREMENTS: Proportion drinking risky amounts in the past month (defined by national standards); proportion drinking risky amounts with a current alcohol diagnosis (determined by diagnostic interview). RESULTS: Seventeen percent (986) were drinking risky amounts; 97% exceeded per occasion limits. Most scored > or =8 on the Alcohol Use Disorders Identification Test, strongly correlating with alcohol diagnoses. Most of a subsample of subjects who drank risky amounts and received further evaluation had dependence (77%). CONCLUSIONS: Drinking risky amounts was common in medical inpatients. Most drinkers of risky amounts had dependence, not the broad spectrum of unhealthy alcohol use anticipated. Screening on a medicine service largely identifies patients with dependence--a group for whom the efficacy of brief intervention (a recommended practice) is not well established.
  64. Voss SE, Horton NJ, Tabucchi TH (SC '05 undergraduate co-author, Folowosele FO (SC '05 undergraduate co-author) and Shera CA. Posture-induced changes in distortion-product otoacoustic emissions and the potential for noninvasive monitoring of changes in intracranial pressure. Neurocritical Care, 2006;4(3):251-257.

    INTRODUCTION: Intracranial pressure (ICP) monitoring is currently an invasive procedure that requires access to the intracranial space through an opening in the skull. Noninvasive monitoring of ICP via the auditory system is theoretically possible because changes in ICP transfer to the inner ear through connections between the cerebral spinal fluid and the cochlear fluids. In particular, low-frequency distortion-product otoacoustic emissions (DPOAEs), measured noninvasively in the external ear canal, have magnitudes that depend on ICP. Postural changes in healthy humans cause systematic changes in ICP. Here, we quantify the effects of postural changes, and presumably ICP changes, on DPOAE magnitudes. METHODS: DPOAE magnitudes were measured on seven normal-hearing, healthy subjects at four postural positions on a tilting table (angles 90 degrees , 0 degrees , - 30 degrees , and - 45 degrees to the horizontal). At these positions, it is expected that ICP varied from about 0 (90 degrees ) to 22 mm Hg ( - 45 degrees ). DPOAE magnitudes were measured for a set of frequencies 750 < f2 < 4000, with f2/f1 = 1.2. RESULTS: For the low-frequency range of 750 < f2 < 1500, the differences in DPOAE magnitude between upright and - 45 degrees were highly significant (all p < 0.01), and above 1500 Hz there were minimal differences between magnitudes at 90 degrees versus - 45 degrees. There were no significant differences in the DPOAE magnitudes with subjects at 90 degrees and 0 degrees postures. CONCLUSIONS: Changes in ICP can be detected using the auditory-based measurement of DPOAEs. In particular, changes are largest at low frequencies. Although this approach does not allow for absolute measurement of ICP, it appears that measurement of DPOAEs may be a useful means of noninvasively monitoring ICP.
  65. Smith KL, Horton NJ, Saitz R and Samet JH. The use of the mini-mental state examination in recruitment for substance abuse research studies. Drug and Alcohol Dependence, 2006;82(3):231-237.

    BACKGROUND: Substance abuse is associated with cognitive impairment. Participation in clinical addiction research can be cognitively demanding. Screening tools can identify cognitively impaired subjects. We examined the use of the mini-mental state examination (MMSE) as an entry criterion in three randomized controlled substance abuse clinical trials. METHODS: In each of the three studies, we calculated the proportion of subjects excluded due to MMSE scores (<21) suggestive of cognitive impairment. We estimated the potential impact on enrollment based on the number of excluded subjects. Separately, for two of the studies, we assessed the impact of cognitive function on participation in follow-up using multivariable logistic regression. RESULTS: Of all persons screened for enrollment, 1.6% (171/10,791) were ineligible based solely on a MMSE score of <21. We estimate that 119 of these 171 ineligible persons would have consented and enrolled. These 119 persons would have represented 9.3% of all enrolled subjects across these studies. For subjects in a study in an inpatient detoxification unit, a higher MMSE score was associated with higher odds (adjusted odds ratio 1.15, 95% CI 1.03-1.30) of completing at least one follow-up assessment. A similar impact on subject follow-up was not observed in a study of medical inpatients with unhealthy alcohol use (adjusted odds ratio 1.01, 95% CI 0.86-1.20). CONCLUSION: Screening for cognitive impairment using the MMSE excludes a small, but substantial, number of persons from addiction research studies. Cognitive ability, as captured by the MMSE may impact follow-up. These data support cognitive screening of substance abuse research subjects.
  66. Zamboanga BL, Horton NJ, Leitkowski LK (SC '05 undergraduate co-author) and Wang SC (SC '05 undergraduate co-author). Do good things come to those who drink? A longitudinal investigation of drinking expectancies and hazardous alcohol use in female college athletes. Journal of Adolescent Health, 2006;39(2):229-236.

    PURPOSE: To prospectively examine the link between positive and negative drinking expectancies and hazardous alcohol use, and to explore the reciprocal associations between expectancies and hazardous use among female college athletes. METHOD: Participants were part of a larger (n = 189), ongoing longitudinal study on female athletes' socialization experiences and health behaviors from an all-women's college in the Northeastern United States. The sample for the present study were 85 college-aged (17-22 years) female students (Mean age = 19.6) who indicated alcohol use in the past year and provided data at both time points (baseline and one year later). Respondents completed self-report questionnaires on their age, hazardous alcohol use, and expectations about the effects of drinking. RESULTS: Logistic regression analyses showed that positive drinking expectancies predicted increased odds of hazardous alcohol use at baseline and one year later (even after controlling for baseline hazardous use). Positive expectancies accounted for a higher proportion of variance in predicting increased odds of hazardous alcohol use at both time points than did negative expectancies. Bivariate correlational analyses examining the bi-directional associations between expectancies and hazardous alcohol use also revealed some indication that positive expectancies were associated with hazardous alcohol use rather than vice-versa. No such relations emerged for negative expectancies. CONCLUSIONS: Findings shed light on the utility of positive expectancies in predicting concurrent and subsequent hazardous alcohol use among female student-athletes. Future research directions and potential implications for prevention efforts are discussed.
  67. Cozier Y, Palmer JR, Horton NJ, Fredman L, Wise LA and Rosenberg L. Racial Discrimination and the Incidence of Hypertension in US Black Women. Annals of Epidemiology, 16(9):681-7.

    PURPOSE: Unique experiences associated with "race," such as racism, may adversely affect health. Our goal is to assess whether racism is associated with the occurrence of hypertension in African-American women. METHODS: In the first prospective examination of perceived experiences of racism in relation to the incidence of hypertension, we used data from the Black Women's Health Study, a follow-up study of US black women that began in 1995. The 1997 follow-up questionnaire contained eight questions designed to measure personally mediated racism and institutionalized racism. Cox proportional hazard models were used to estimate incidence rate ratios (IRRs), with control for age, body mass index, and questionnaire period. RESULTS: There were 2316 incident cases of hypertension reported during 104,574 person-years of observation from 1997 to 2001. Most women reported experiences of racism. In the total sample, IRRs for the association of racism with incident hypertension were close to the null. However, some positive associations were observed for personally mediated racism in women born outside the United States. CONCLUSIONS: There may be an increase in hypertension associated with experiences of racism in certain subgroups of African-American women.
  68. Larson MJ, Saitz R, Horton NJ, Lloyd-Travaglini C and Samet JH. Emergency Department and Hospital Utilization Among Alcohol and Drug-Dependent Detoxification Patients without Primary Medical Care. American Journal of Drug and Alcohol Dependence, 2006;32(3):435-452.

    Utilization of emergency department (ED) services and hospitalization among a cohort of substance abusers are described based on structured research interviews with 470 adults without primary care admitted to an urban residential detoxification program. Cross-sectional analysis of baseline data of subjects found nearly 19% of subjects went to an ED on 2 or more occasions in the 6 months prior to detoxification and 14% were admitted for an overnight hospitalization. Upon further analysis of past 6-month ED utilization, the following factors were independently associated with increased odds of ED use: White race; at least one month homeless in the past 5 years chronic health condition; injury in past 6 months; and subject perception that their substance abuse interfered with seeking care from a regular doctor. Subjects with cocaine as a primary problem had lower odds of ED utilization than a reference group with alcohol as a primary problem.
  69. O'Brien LM, Fitzmaurice GM and Horton NH. Maximum likelihood estimation of marginal pairwise associations with multiple source predictors. Biometrical Journal, 2006; 48(5):860-875.

    Researchers interested in the association of a predictor with an outcome will often collect information about that predictor from more than one source. Standard multiple regression methods allow estimation of the effect of each predictor on the outcome while controlling for the remaining predictors. The resulting regression coefficient for each predictor has an interpretation that is conditional on all other predictors. In settings in which interest is in comparison of the marginal pairwise relationships between each predictor and the outcome separately (e.g., studies in psychiatry with multiple informants or comparison of the predictive values of diagnostic tests), standard regression methods are not appropriate. Instead, the generalized estimating equations (GEE) approach can be used to simultaneously estimate, and make comparisons among, the separate pairwise marginal associations. In this paper, we consider maximum likelihood (ML) estimation of these marginal relationships when the outcome is binary. ML enjoys benefits over GEE methods in that it is asymptotically efficient, can accommodate missing data that are ignorable, and allows likelihood-based inferences about the pairwise marginal relationships. We also explore the asymptotic relative efficiency of ML and GEE methods in this setting.
  70. Kim TW, Kertesz SG, Horton NJ, Tibbetts N and Samet JH. Episodic homelessness and health care utilization in a prospective cohort of HIV-infected persons with alcohol problems. BMC Health Services Research, 2006; 6:19.

    BACKGROUND: Because individuals with HIV/AIDS often have complex medical and social needs, the impact of housing status on medical service utilization is difficult to isolate from the impact of conditions that may worsen during periods of homelessness such as depression and substance abuse. We examine whether episodes of homelessness are independently associated with suboptimal medical utilization even when accounting for concurrent addiction severity and depression. METHODS: We used data from a 30-month cohort of patients with HIV/AIDS and alcohol problems. Housing status, utilization (ambulatory visits, emergency department (ED) visits, and hospitalizations) and other features were assessed with standardized research interviews at 6-month intervals. Multivariable longitudinal regression models calculated incidence rate ratios (IRR) comparing utilization rates during 6-month intervals (homeless versus housed). Additional models assessed whether addiction severity and depressive symptoms could account for utilization differences. RESULTS: Of the 349 subjects, 139 (39%) reported homelessness at least once during the study period; among these subjects, the median number of nights homeless per 6-month interview period was 30. Homelessness was associated with higher ED utilization (IRR = 2.17; 95% CI = 1.72-2.74) and hospitalizations (IRR = 2.30; 1.70-3.12), despite no difference in ambulatory care utilization (IRR = 1.09; 0.89-1.33). These associations were attenuated but remained significant when adjusting for addiction severity and depressive symptoms. CONCLUSION: In patients with HIV/AIDS and alcohol problems, efforts to improve housing stability may help to mitigate intensive medical utilization patterns.
  71. Zamboanga BL, Raffaelli M and Horton NJ. Acculturation status and heavy alcohol use among Mexican American college students: Investigating the moderating role of gender. Addictive Behaviors, 31(12):2188-98.

    We examined whether gender moderates the association between acculturation and heavy alcohol use. The sample consisted of 126 Mexican American college students (Mean age=24.7 years; 57% female) who completed self-report measures of heavy alcohol use, acculturation status (global acculturation and ethnic identity), and relevant control variables (age, peer alcohol use). Multivariable regression revealed that higher levels of ethnic identity were associated with greater frequency of heavy alcohol among men. Conversely, neither measure of acculturation was associated with heavy alcohol use among women. These findings suggest that interventions for Latino/a students should consider the role of culturally relevant variables in heavy alcohol use, particularly for men. They also have implications regarding how acculturation is operationalized in alcohol studies, and suggest directions for future research.
  72. Cozier Y, Palmer JR, Horton NJ, Fredman L, Wise LA and Rosenberg L. Relationship between neighborhood median housing value and hypertension risk among Black women in the United States. American Journal of Public Health, 2007; 97(4):718-724.

    OBJECTIVES: We examined the relation between median housing value and hypertension risk among US Black women. METHODS: We gathered data from the Black Women's Health Study, a prospective follow-up of 59000 Black women aged 21 to 69 years in 1995. Median housing value from US census data was used to measure neighborhood socioeconomic status. Cases of hypertension were identified through postal questionnaires mailed in 1997, 1999, and 2001. Clustered survival regression models were used to estimate incidence rate ratios. RESULTS: During 180294 person-years of observation, 3780 cases of hypertension were reported. A significant inverse, graded association was found between median housing value and hypertension. The incidence rate ratio for women living in low median housing value neighborhoods relative to high was 1.29 (95% confidence interval=1.14, 1.45) after adjustment for individual risk factors. The association was evident even at higher individual levels of income and education. CONCLUSIONS: Median housing value is inversely associated with hypertension in Black women, independent of individual risk factors. Lowering hypertension risk in Black women will require a greater understanding of the underlying social inequalities that adversely affect health.
  73. Switzer SS (SC '06 undergraduate first-author) and Horton NJ. What your doctor should know about statistics (but perhaps doesn't), Chance, 2007; 20(1):17-21.

  74. Litman HJ, Horton NJ, Hernandez B and Laird NM. Incorporating missingness for estimation of marginal regression models with multiple source predictors, Statistics in Medicine, 2007; 26(5):1055-68.

    Multiple informant data refers to information obtained from different individuals or sources used to measure the same construct; for example, researchers might collect information regarding child psychopathology from the child's teacher and the child's parent. Frequently, studies with multiple informants have incomplete observations; in some cases the missingness of informants is substantial. We introduce a Maximum Likelihood (ML) technique to fit models with multiple informants as predictors that permits missingness in the predictors as well as the response. We provide closed form solutions when possible and analytically compare the ML technique to the existing Generalized Estimating Equations (GEE) approach. We demonstrate that the ML approach can be used to compare the effect of the informants on response without standardizing the data. Simulations incorporating missingness show that ML is more efficient than the existing GEE method. In the presence of MCAR missing data, we find through a simulation study that the ML approach is robust to a relatively extreme departure from the normality assumption. We implement both methods in a study investigating the association between physical activity and obesity with activity measured using multiple informants (children and their mothers).
  75. Litman HJ, Horton NJ, Murphy JM and Laird NM. Marginal regression models with a time to event outcome and discrete multiple source predictors. Lifetime Data Analysis, 2006; 12(3):249-65.

    Information from multiple informants is frequently used to assess psychopathology. We consider marginal regression models with multiple informants as discrete predictors and a time to event outcome. We fit these models to data from the Stirling County Study; specifically, the models predict mortality from self report of psychiatric disorders and also predict mortality from physician report of psychiatric disorders. Previously, Horton et al. found little relationship between self and physician reports of psychopathology, but that the relationship of self report of psychopathology with mortality was similar to that of physician report of psychopathology with mortality. Generalized estimating equations (GEE) have been used to fit marginal models with multiple informant covariates; here we develop a maximum likelihood (ML) approach and show how it relates to the GEE approach. In a simple setting using a saturated model, the ML approach can be constructed to provide estimates that match those found using GEE. We extend the ML technique to consider multiple informant predictors with missingness and compare the method to using inverse probability weighted (IPW) GEE. Our simulation study illustrates that IPW GEE loses little efficiency compared with ML in the presence of monotone missingness. Our example data has non-monotone missingness; in this case, ML offers a modest decrease in variance compared with IPW GEE, particularly for estimating covariates in the marginal models. In more general settings, e.g., categorical predictors and piecewise exponential models, the likelihood parameters from the ML technique do not have the same interpretation as the GEE. Thus, the GEE is recommended to fit marginal models for its flexibility, ease of interpretation and comparable efficiency to ML in the presence of missing data.
  76. Bruno A (SC '05 graduate student first-author), Marcotrigiano M and Horton NJ. A three-stage micropropagation system for a novel color morph of the wildflower Porteranthus trifoliatus (L.) Britt., Plant Cell, Tissue & Organ Culture, 2007; 90(1):9-13.

    A three-stage micropropagation system was devised for Porteranthus trifoliatus Pink Profusion, a cultivar distinguished from the wild type by its pink flowers, purple stems and darker reddish leaves. Nodes of young shoots were used as explants, disinfested, and placed on several different Murashige and Skoogs (MS) media formulations containing two levels of indole-3-butyric acid (IBA) and four levels of benzylaminopurine (BAP) in a factorial combination. An optimal number of commercially-usable shoots was achieved with BAP concentrations between 10 mu M and 30 mu M, while the addition of IBA made no significant difference in the number of shoots produced. Proliferated shoots could be rooted ex vitro without auxin treatment.
  77. Zamboanga BL, Horton NJ, Tyler KM-B (SC '09 undergraduate co-author), O'Riordan SS (SC '07 undergraduate co-author), Calvert BD (SC '07 undergraduate co-author) and McCollum EC (SC '08 undergraduate co-author). The utility of the AUDIT in screening for drinking game involvement among female college students, Journal of Adolescent Health, 2007; 40(4):359-61.

    Drinking games (DG) facilitate heavy alcohol consumption in a short period and are associated with negative experiences. We examined the utility of Alcohol Use Disorders Identification Test (AUDIT) cut-off scores to identify DG involvement. Findings indicated an AUDIT score of at least 5 is needed to identify gamers among students at a women's college. Prevention implications are discussed.
  78. Williams EC, Horton NJ, Samet JH and Saitz R. Do brief measures of readiness to change predict alcohol consumption and consequences in primary care patients with unhealthy alcohol use?, Alcoholism: Clinical and Experimental Research, 2007; 31(3):428-35.

    BACKGROUND: Assessing readiness to change is recommended as part of brief interventions for patients with unhealthy alcohol use. However, the utility and predictive validity of readiness measures have not been well established. METHODS: In a prospective cohort study, we assessed primary care patients with unhealthy alcohol use (past-month drinking of risky amounts, or any amount and an affirmative response to CAGE alcohol screening questionnaire) and reassessed them 6 months later. At study entry, we assessed readiness to change using 1 multi-item measure of stage of change, and 5 single-item measures (readiness per se, importance of changing, confidence in ability to change, intention to cut down, intention to abstain). Outcomes included alcohol consumption and alcohol-related consequences. Multivariable regression models were fit for each measure of readiness and each outcome. RESULTS: Of 312 patients with unhealthy alcohol use, 228 (73%) were assessed at study entry and 6 months later and had complete data. Among readiness measures, only confidence and intention to abstain (1 point changes on single-item measures) were associated with consumption 6 months later: less heavy episodic drinking [adjusted odds ratio (AOR) 0.88, 95% CI 0.80-0.98 and AOR 0.79, 0.64-0.98, respectively], and less drinking of risky amounts (AOR 0.89, 0.79-1.00 and AOR 0.78, 0.62-0.98, respectively). Intention to abstain was also associated with more abstinence (AOR 1.43, 1.09-1.88). Single-item measures of readiness, importance, and intention to cut down were significantly associated with higher odds of alcohol consequences. Greater confidence (single item) was associated with a lower odds of any consequences (AOR 0.88, 0.79-0.98). CONCLUSIONS: Greater readiness, as measured by several brief assessments, was associated with more consequences and was not predictive of consumption. However, assessing confidence in the ability to change one's alcohol use may have a role in predicting subsequent decreases in both consumption and consequences in primary care patients.
  79. Zamboanga BL, Rodriguez L and Horton NJ. Athletic involvement and its relevance to hazardous alcohol use and drinking games participation in female college athletes: A preliminary investigation, Journal of American College Health, 2008; 56(6)651-656. OBJECTIVE AND PARTICIPANTS: The authors designed this cross-sectional study to examine sports team differences in hazardous alcohol use and drinking game participation, as well as the social correlates of these behaviors among female college athletes (N = 176; M age = 19.9 years, SD = 1.24, range = 18-22). METHODS: Respondents completed self-report surveys in small groups. They reported drinking behaviors, frequency of team social events involving alcohol use, and team cohesion. RESULTS: The authors found significant differences across sports teams with regard to hazardous alcohol use and participation in drinking games with teammates. Findings also revealed that a high frequency of team social events involving alcohol use was associated with elevated use and an increased likelihood of drinking game participation. CONCLUSIONS: The authors discuss future research directions and implications.

  80. Saitz R, Palfai T, Cheng DM, Horton NJ, Freedner N, Dukes K, Kraemer KL, Roberts MS and Samet JH. Brief intervention for medical inpatients with unhealthy alcohol use: a randomized, controlled trial, Annals of Internal Medicine, 2007; 146(3):167-176 (Summary for patients in: Annals of Internal Medicine, 2007; 146(3):I22).

    BACKGROUND: The efficacy of brief intervention in reducing alcohol consumption is well established for selected outpatients but not for medical inpatients. OBJECTIVE: To determine whether brief intervention improves alcohol outcomes in medical inpatients who were identified by screening as having unhealthy alcohol use. DESIGN: Randomized, controlled trial. SETTING: Medical service of an urban hospital. PATIENTS: 341 medical inpatients who were drinking risky amounts of alcohol (defined for eligibility as >14 drinks/wk or > or =5 drinks/occasion for men and >11 drinks/wk or > or =4 drinks/occasion for women and persons > or =66 y); 77% had alcohol dependence as determined by the Composite International Diagnostic Interview Alcohol Module. INTERVENTION: A 30-minute session of motivational counseling given by trained counselors during a patient's hospitalization (n = 172) versus usual care (n = 169). MEASUREMENTS: Self-reported primary outcomes were receipt of alcohol assistance (for example, alcohol disorders specialty treatment) by 3 months in dependent drinkers and change in the mean number of drinks per day from enrollment to 12 months in all patients. RESULTS: The intervention was not significantly associated with receipt of alcohol assistance by 3 months among alcohol-dependent patients (adjusted proportions receiving assistance, 49% for the intervention group and 44% for the control group; intervention-control difference, 5% [95% CI, -8% to 19%]) or with drinks per day at 12 months among all patients (adjusted mean decreases, 1.5 for patients who received the intervention and 3.1 for patients who received usual care; adjusted mean group difference, -1.5 [CI, -3.7 to 0.6]). There was no significant interaction between the intervention and alcohol dependence in statistical models predicting drinks per day (P = 0.24). LIMITATIONS: Baseline imbalances existed between randomized groups. Patients who received usual care were assessed and advised that they could discuss their drinking with their physicians. CONCLUSIONS: Brief intervention is insufficient for linking medical inpatients with treatment for alcohol dependence and for changing alcohol consumption. Medical inpatients with unhealthy alcohol use require more extensive, tailored alcohol interventions.
  81. Horton NJ, Kim E (SC '07 undergraduate co-author) and Saitz R. A cautionary note regarding count models for alcohol consumption in randomized clinical trials, BMC Medical Research Methodology, 2007; 7:9.

    Background Alcohol consumption is commonly used as a primary outcome in randomized alcohol treatment studies. The distribution of alcohol consumption is highly skewed, particularly in subjects with alcohol dependence.
    Methods In this paper, we will consider the use of count models for outcomes in a randomized clinical trial setting. These include the Poisson, over-dispersed Poisson, negative binomial, zero-inflated Poisson and zero-inflated negative binomial. We compare the Type-I error rate of these methods in a series of simulation studies of a randomized clinical trial, and apply the methods to the ASAP (Addressing the Spectrum of Alcohol Problems) trial.
    Results Standard Poisson models provide a poor fit for alcohol consumption data from our motivating example, and did not preserve Type-I error rates for the randomized group comparison when the true distribution was over-dispersed Poisson. For the ASAP trial, where the distribution of alcohol consumption featured extensive over-dispersion, there was little indication of significant randomization group differences, except when the standard Poisson model was fit.
    Conclusion As with any analysis, it is important to choose appropriate statistical models. In simulation studies and in the motivating example, the standard Poisson was not robust when fit to over-dispersed count data, and did not maintain the appropriate Type-I error rate. To appropriately model alcohol consumption, more flexible count models should be routinely employed.
  82. Wines JD, Saitz R, Horton NJ, Lloyd-Travaglini C and Samet JH. Overdose after detoxification: a prospective study, Drug and Alcohol Dependence, 2007; 89(2-3):161-169.

    OBJECTIVE: The aim of this study was to determine predictors of non-fatal overdose (OD) among a cohort of 470 adults after detoxification from heroin, cocaine or alcohol. METHODS: We examined factors associated with time to OD during 2 years after discharge from an urban detoxification unit in Boston, MA, USA using multivariable regression analyses. Separate analyses were performed for both the total sample and a subgroup with problem opioid use. RESULTS: Lifetime prevalence for any OD was 30.9% (145/470) in the total sample and 42.3% (85/201) in patients with opioid problems. During the 2-year follow-up, OD was estimated to occur in 16.9% of the total sample and 26.7% of the opioid problem subgroup, with new-onset (incidence) OD estimated at 5.7% and 11.0%, respectively. Factors associated with an increased hazard of OD in both samples included white race, more depressive symptoms, and prior OD regardless of intent. Prior suicidal ideation or attempt was not associated with future OD. CONCLUSIONS: Findings underscore both the high prevalence of non-fatal OD among detoxification patients especially opioid users, and the potency of prior OD as a risk factor for future OD. Depressive symptoms, a modifiable risk factor, may represent a potential intervention target to prevent OD, including some "unintentional" ODs.
  83. Horton NJ and Kleinman KP. Much ado about nothing: A comparison of missing data methods and software to fit incomplete data regression models (plus appendix), The American Statistician, 2007; 61(1):79-90.

    Missing data are a recurring problem that can cause bias or lead to inefficient analyses. Development of statistical methods to address missingness have been actively pursued in recent years, including imputation, likelihood and weighting approaches. Each approach is more complicated when there are many patterns of missing values, or when both categorical and continuous random variables are involved. Implementations of routines to incorporate observations with incomplete variables in regression models are now widely available. We review these routines in the context of a motivating example from a large health services research dataset. While there are still limitations to the current implementations, and additional efforts are required of the analyst, it is feasible to incorporate partially observed values, and these methods should be utilized in practice.
  84. Henle J, Jakus S (SC '05 undergraduate co-author) and Horton NJ. Modelling inequality with a single parameter, Modeling Income Distributions and Lorenz Curves (Series of Economic studies in inequality, social exclusion and well-being), 2008, Springer ISBN 978-0-387-72756-1.

  85. Finucane MM (SC '05 undergraduate first-author), Samet JH and Horton NJ. Translational methods in biostatistics: Longitudinal regression models of alcohol consumption and HIV disease progression, Epidemiologic Perspectives & Innovations, 2007; 4:8.

    Longitudinal studies are helpful in understanding how subtle associations between factors of interest change over time. Our goal is to apply statistical methods which are appropriate for analyzing longitudinal data to a repeated measures epidemiological study as a tutorial in the appropriate use and interpretation of random effects models. To motivate their use, we study the association of alcohol consumption on markers of HIV disease progression in an observational cohort. To make valid inferences, the association among measurements correlated within a subject must be taken into account. We describe a linear mixed effects regression framework that accounts for the clustering of longitudinal data and that can be fit using standard statistical software. We apply the linear mixed effects model to a previously published dataset of HIV infected individuals with a history of alcohol problems who are receiving HAART (n=197). The researchers were interested in determining the effect of alcohol use on HIV disease progression over time. Fitting a linear mixed effects multiple regression model with a random intercept and random slope for each subject accounts for the association of observations within subjects and yields parameters interpretable as in ordinary multiple regression. A significant interaction between alcohol use and adherence to HAART is found: subjects who use alcohol and are not fully adherent to their HIV medications had higher log RNA (ribonucleic acid) viral load levels than fully adherent non-drinkers, fully adherent alcohol users, and non-drinkers who were not fully adherent. Longitudinal studies are increasingly common in epidemiological research. Software routines that account for correlation between repeated measures using linear mixed effects methods are now generally available and straightforward to utilize. These models allow the relaxation of assumptions needed for approaches such as repeated measures ANOVA, and should be routinely incorporated into the analysis of cohort studies.
  86. Litman HJ, Horton NJ, Hernandez B and Laird NM. Estimation of marginal regression models with multiple source predictors. In Handbook of Statistics (volume 27): Epidemiology and Medical Statistics, 2007; chapter 25, 730-746.

  87. Horton NJ. Clinician attitudes towards biostatistics (research letter). Mayo Clinic Proceedings, 2007; 82(12):1578.

  88. Voss SE, Horton NJ, Woodbury RR (SC '06 undergraduate co-author) and Sheffield KN (SC '07 undergraduate co-author). Sources of variability in reflectance measurements on normal cadaver ears. Ear and Hearing, 2008; 29(4):651-656.

    OBJECTIVES: The development of acoustic reflectance measurements may lead to noninvasive tests that provide information currently unavailable from standard audiometric testing. One factor limiting the development of these tests is that normal-hearing human ears show substantial intersubject variations. This work examines intersubject variability that results from measurement location within the ear canal, estimates of ear-canal area, and variations in middle-ear cavity volume. DESIGN: Energy reflectance (ER) measurements were made on nine human-cadaver ears to study three variables. (1) ER was measured at multiple ear-canal locations. (2) The ear-canal area at each measurement location was measured and the ER was calculated with the measured area, a constant area, and an acoustically estimated area. (3) The ER was measured with the middle-ear cavity in three conditions: (1) normal, (2) the mastoid widely opened (large air space), and (3) the mastoid closed off at the aditus ad antrum (small air space). RESULTS: Measurement-location effects are generally largest at frequencies below about 2000 Hz, where in some ears reflectance magnitudes tend to decrease systematically as the measurement location moves away from the tympanic membrane but in other ears the effects seem minimal. Intrasubject variations in reflectance due to changes in either measurement location within the ear canal or differences in the estimate of the ear-canal area are smaller than variations produced by large variations in middle-ear cavity air volume or intersubject differences. At frequencies below 2000 Hz, large increases in cavity volume systematically reduce the ER, with more variable changes above 2000 Hz. CONCLUSIONS: ER measurements depend on all variables studied: measurement location, ear-canal cross-sectional area, and middle-ear cavity volume. Variations within an individual ear in either measurement location or ear-canal cross-sectional area result in relatively small effects on the ER, supporting the notion that diagnostic tests (1) need not control for measurement location and (2) can assume a constant ear-canal area across most subjects. Variations in cavity volume produce much larger effects in ER than measurement location or ear-canal area, possibly explaining some of the intersubject variation in ER reported among normal ears.
  89. Saitz R, Horton NJ, Cheng DM and Samet JH. Alcohol counseling reflects higher quality of primary care. Journal of General Internal Medicine, 2008; 23(9):1482-1486.

    BACKGROUND: Some primary care physicians do not conduct alcohol screening because they assume their patients do not want to discuss alcohol use. OBJECTIVES: To assess whether (1) alcohol counseling can improve patient-perceived quality of primary care, and (2) higher quality of primary care is associated with subsequent decreased alcohol consumption. DESIGN: A prospective cohort study. SUBJECTS: Two hundred eighty-eight patients in an academic primary care practice who had unhealthy alcohol use. MEASUREMENTS: The primary outcome was quality of care received [measured with the communication, whole-person knowledge, and trust scales of the Primary Care Assessment Survey (PCAS)]. The secondary outcome was drinking risky amounts in the past 30 days (measured with the Timeline Followback method). RESULTS: Alcohol counseling was significantly associated with higher quality of primary care in the areas of communication (adjusted mean PCAS scale scores: 85 vs. 76) and whole-person knowledge (67 vs. 59). The quality of primary care was not associated with drinking risky amounts 6 months later. CONCLUSIONS: Although quality of primary care may not necessarily affect drinking, brief counseling for unhealthy alcohol use may enhance the quality of primary care.
  90. Murphy JM, Burke JD, Monson RR, Horton NJ, Laird NM, Lesage A, Sobol AM and Leighton AH. Mortality associated with depression: A forty-year perspective from the Stirling County Study. Social Psychiatry and Psychiatric Epidemiology, 2008; 43(8):594-601.

    BACKGROUND: This report concerns long-term mortality risks associated with depression, and the potentially confounding factors of alcoholism and cigarette smoking, as experienced by a general population assessed at a baseline in 1952, followed for re-assessment of survivors in 1968, and for death by 1992. METHODS: Self-report and physician-report information was gathered in 1952 and again in 1968 about a sample of 1,079 adults. At the end of follow-up in 1992, the vital status of all subjects was known. Comorbidity among depression, alcoholism, and smoking was investigated. Cox regression models were employed to estimate hazard ratios (HRs) as indicators of mortality risk. Models including age, gender, and depression were fit for the complete sample at baseline as well as for re-assessed survivors. Models simultaneously controlling for the mortality risks associated with depression, alcoholism, and heavy smoking were fit for men. RESULTS: At the baseline in 1952, depression was somewhat more common among women than men (4% compared to 6%) but was found to carry a significant mortality risk only among men (HR 2.7, 95% CI 1.6-4.7). Based on re-assessments made in 1968, depression was associated with mortality risk among both men (HR 2.2, 95% CI 1.0-4.5) and women (HR 2.1, 95% CI 1.2-3.8). In 1952, more than 20% of men smoked cigarettes excessively and 8% abused alcohol, but very few of these groups of men were also depressed. In the original sample and also among the survivors, depression, alcoholism, and heavy smoking were separately associated with mortality among men. Depression and alcoholism carried a more immediate mortality risk while heavy smoking a more delayed one. CONCLUSIONS: At the baseline of the Stirling County Study, the mortality risk associated with depression among men was not enhanced or explained by abuse of alcohol or nicotine, mainly because comorbidity was rare at that time. The longitudinal research of the study has pointed to a number of psychiatrically-relevant time-trends such as the fact that an association between depression and cigarette smoking did not appear until the 1990s. It is hypothesized that a similar trend may emerge over time regarding the comorbidity of depression and alcoholism. A trend reported here was that, while depressed women in the original sample did not carry a significant mortality risk, the surviving women who were depressed at the time of re-assessment exhibited a mortality risk that was as significant as that for men. Such information may provide a useful back-drop for future investigations.
  91. Banh MK, Saxe G, Mangione T and Horton NJ. Physician reported practice managing childhood posttraumatic stress in pediatric primary care. General Hospital Psychiatry, 2008; 30(6):536-545.

    OBJECTIVE: This study investigated pediatrician-reported practices in identifying, assessing, and treating traumatic exposure and posttraumatic stress disorder (PTSD) in children. METHOD: Focus groups guided the development of a survey that was mailed to primary care pediatricians in Massachusetts in 2005. Descriptive statistics and multivariate analyses were used to describe clinical practices and perceived barriers to care. RESULTS: A 60% (N=597) survey response-rate was obtained. On average, pediatricians reported that less than 8% of patients had psychological problems that may be related to traumatic exposure. Only 18% of pediatricians agreed that they had adequate knowledge of childhood PTSD. About 15% of pediatricians reported frequently learning about traumatic event(s) from direct inquiry in the past year. Only 10% of pediatricians reported frequent assessment and treatment of posttraumatic stress symptoms. Most pediatricians (72%) agreed that greater collaborations with mental health providers would improve pediatric assessment of PTSD. Finally, having received PTSD-specific training and believing that pediatricians should identify and manage PTSD were each significantly associated with learning about a traumatic event from direct inquiry. CONCLUSION(S): Providing PTSD-specific training and changing pediatricians' attitudes about childhood PTSD may be useful first steps in improving care for children.
  92. Horton NJ, Roberts K, Ryan L, Suglia S, and Wright RJ. A maximum-likelihood latent variable model for multiple informants. Statistics in Medicine, 2008; 27:4992-5004.

    Studies pertaining to childhood psychopathology often incorporate information from multiple sources (or informants). For example, measurement of some factor of particular interest might be collected from parents, teachers as well as the children being studied. We propose a latent variable modeling framework to incorporate multiple informant predictor data. Several related models are presented, and likelihood ratio tests are introduced to formally compare fit. The incorporation of partially observed subjects is addressed under a variety of missing data mechanisms. The methods are motivated by and applied to a study of the association of chronic exposure to violence on asthma in children.
  93. Caria MP, Bellocco R, Zambon A, Horton NJ and Galanti MR. Overweight and perception of overweight as predictors of smoking and smokeless tobacco use in a cohort of Swedish adolescents. Addiction, 2009; 104(4):661-668.

    AIMS: To study the association between measured or perceived overweight in adolescence and subsequent uptake of cigarette smoking and of the Swedish smokeless tobacco 'snus' (oral moist snuff). DESIGN: Population-based prospective cohort study with 7 years' follow-up. SETTING: Self-administered questionnaires and school nurses' visits. PARTICIPANTS: A total of 2922 children of both sexes and mean age 11.6 years at recruitment, resident in the Stockholm region, Sweden. MEASUREMENTS: Tobacco use was self-reported at baseline and on six subsequent surveys. Subjects' height and weight were measured by school nurses during the first 4 years, self-reported thereafter. Overweight perception was self-reported at the age of 15 years. FINDINGS: Overweight and perception of overweight were not associated with subsequent uptake of either smoking or snus among males. Among females, overweight at baseline was associated with uptake of smoking [adjusted hazard ratio (HR) 1.34, 95% confidence interval (CI) 1.09-1.63], but not of snus. A similar pattern was found with overweight status during follow-up. Among girls with low-educated parents, overweight at baseline predicted the uptake of both snus and smoking. Among 15-year-old females who never used tobacco perceived overweight was associated with subsequent uptake of smoking (adjusted HR 1.71, 95% CI 1.20-2.46), but not of snus. CONCLUSIONS: In Sweden, adolescent girls with actual or perceived overweight are at increased risk to start smoking, while indications that this increased risk applies to smokeless tobacco (snus) are limited to girls of low socio-economic status.
  94. Bertholet N, Horton NJ and Saitz R. Improvements in readiness to change and drinking in primary care patients with unhealthy alcohol use: a prospective study. BMC Public Health, 2009; 9:101.

    BACKGROUND: The course of alcohol consumption and cognitive dimensions of behavior change (readiness to change, importance of changing and confidence in ability to change) in primary care patients are not well described. The objective of the study was to determine changes in readiness, importance and confidence after a primary care visit, and 6-month improvements in both drinking and cognitive dimensions of behavior change, in patients with unhealthy alcohol use. METHODS: Prospective cohort study of patients with unhealthy alcohol use visiting primary care physicians, with repeated assessments of readiness, importance, and confidence (visual analogue scale (VAS), score range 1-10 points). Improvements 6 months later were defined as no unhealthy alcohol use or any increase in readiness, importance, or confidence. Regression models accounted for clustering by physician and adjusted for demographics, alcohol consumption and related problems, and discussion with the physician about alcohol. RESULTS: From before to immediately after the primary care physician visit, patients (n = 173) had increases in readiness (mean +1.0 point), importance (+0.2), and confidence (+0.5) (all p < 0.002). In adjusted models, discussion with the physician about alcohol was associated with increased readiness (+0.8, p = 0.04). At 6 months, many participants had improvements in drinking or readiness (62%), drinking or importance (58%), or drinking or confidence (56%). CONCLUSION: Readiness, importance and confidence improve in many patients with unhealthy alcohol use immediately after a primary care visit. Six months after a visit, most patients have improvements in either drinking or these cognitive dimensions of behavior change.
  95. Saitz R, Palfai T, Cheng DM, Horton NJ, Dukes K, Kraemer KL, Roberts MS, Guerriero RT and Samet JH. Some medical inpatients with unhealthy alcohol use may benefit from brief intervention. Journal of Studies on Alcohol and Drugs, 2009; 70(3):426-435.

    OBJECTIVE: Studies of alcohol brief intervention for medical inpatients have mixed results. We explored potential moderators of the effectiveness of brief intervention for unhealthy alcohol use among medical inpatients. METHOD: This is a secondary analysis of a randomized controlled trial of brief motivational counseling among 341 urban-hospital medical inpatients (99 women) with unhealthy alcohol use. Self-reported main outcomes were receipt of alcohol treatment by 3 months in subjects with dependence and change in the mean number of drinks per day 3 and 12 months after enrollment in all subjects. RESULTS: Among subjects with dependence, the effect of brief intervention on receipt of alcohol treatment differed significantly by gender and age (p = .02 for each interaction). In stratified analyses, brief intervention was associated with receipt of alcohol treatment in women (adjusted odds ratio [AOR] = 3.9, 95% confidence interval [CI]: 1.2-12.7), and younger (<44 years) subjects (AOR = 3.6, 95% CI: 1.3-10.1). Among subjects with nondependent, unhealthy alcohol use, brief intervention was significantly associated with fewer drinks per day and better physical health-related quality of life at 3 months. However, among those with dependence, intervention was associated with worse physical health-related quality of life and more hospital use, and no changes in drinking. In adjusted analyses among those with and without dependence, brief intervention was not associated with mental health-related quality of life, alcohol problems, or readiness to change. Effects of brief intervention on consumption outcomes were not consistently moderated by demographic characteristics, comorbidity/health, or readiness to change. CONCLUSIONS: Some medical inpatients with unhealthy alcohol use, particularly women, younger adults, and patients without dependence may benefit from brief intervention. Few factors that were expected to moderate brief intervention effects did so. Additional research should assess which medical inpatients, if any, can benefit from brief intervention.
  96. Bertholet N, Dukes KM, Horton NJ, Palfai T, Pedley A and Saitz R. Factor structure of the SOCRATES questionnaire in hospitalized medical patients. Addictive Behaviors, 2009; 34(6-7):568-572.

    The Stages of Change Readiness and Treatment Eagerness Scale (SOCRATES), a 19-item instrument developed to assess readiness to change alcohol use among individuals presenting for specialized alcohol treatment, has been used in various populations and settings. Its factor structure and concurrent validity has been described for specialized alcohol treatment settings and primary care. The purpose of this study was to determine the factor structure and concurrent validity of the SOCRATES among medical inpatients with unhealthy alcohol use not seeking help for specialized alcohol treatment. The subjects were 337 medical inpatients with unhealthy alcohol use, identified during their hospital stay. Most of them had alcohol dependence (76%). We performed an Alpha Factor Analysis (AFA) and Principal Component Analysis (PCA) of the 19 SOCRATES items, and forced 3 factors and 2 components, in order to replicate findings from Miller and Tonigan (Miller, W. R., & Tonigan, J. S., (1996). Assessing drinkers' motivations for change: The Stages of Change Readiness and Treatment Eagerness Scale (SOCRATES). Psychology of Addictive Behavior, 10, 81-89.) and Maisto et al. (Maisto, S. A., Conigliaro, J., McNeil, M., Kraemer, K., O'Connor, M., & Kelley, M. E., (1999). Factor structure of the SOCRATES in a sample of primary care patients. Addictive Behavior, 24(6), 879-892.). Our analysis supported the view that the 2 component solution proposed by Maisto et al. (Maisto, S.A., Conigliaro, J., McNeil, M., Kraemer, K., O'Connor, M., & Kelley, M.E., (1999). Factor structure of the SOCRATES in a sample of primary care patients. Addictive Behavior, 24(6), 879-892.) is more appropriate for our data than the 3 factor solution proposed by Miller and Tonigan (Miller, W. R., & Tonigan, J. S., (1996). Assessing drinkers' motivations for change: The Stages of Change Readiness and Treatment Eagerness Scale (SOCRATES). Psychology of Addictive Behavior, 10, 81-89.). The first component measured Perception of Problems and was more strongly correlated with severity of alcohol-related consequences, presence of alcohol dependence, and alcohol consumption levels (average number of drinks per day and total number of binge drinking days over the past 30 days) compared to the second component measuring Taking Action. Our findings support the view that the SOCRATES is comprised of two important readiness constructs in general medical patients identified by screening.
  97. Murphy JM, Horton NJ, Burke JD Jr, Monson RR, Laird NM, Lesage A and Sobol AM. Obesity and weight gain in relation to depression: findings from the Stirling County Study. International Journal of Obesity, 2009; 33(3):335-341.

    OBJECTIVE: This study concerns the question of whether obese subjects in a community sample experience depression in a different way from the nonobese, especially whether they overeat to the point of gaining weight during periods of depression. DESIGN: A representative sample of adults was interviewed regarding depression and obesity. SUBJECTS: The sample consisted of 1396 subjects whose interviews were studied regarding relationships between obesity and depression and among whom 114 had experienced a major depressive episode at some point in their lives and provided information about the symptoms experienced during the worst or only episode of major depression. MEASUREMENTS: The Diagnostic Interview Schedule (DIS) was used to identify major depressive episodes. Information was also derived from the section on Depression and Anxiety (DPAX) of the Stirling Study Schedule. Obesity was calculated as a body mass index >30. Logistic regressions were employed to assess relationships, controlling for age and gender, by means of odds ratios and 95% confidence intervals. RESULTS: In the sample as a whole, obesity was not related to depression although it was associated with the symptom of hopelessness. Among those who had ever experienced a major depressive episode, obese persons were 5 times more likely than the nonobese to overeat leading to weight gain during a period of depression (P<0.002). These obese subjects, compared to the nonobese, also experienced longer episodes of depression, a larger number of episodes, and were more preoccupied with death during such episodes. CONCLUSIONS: Depression among obese subjects in a community sample tends to be more severe than among the nonobese. Gaining weight while depressed is an important marker of that severity. Further research is needed to understand and possibly prevent the associations, sequences and outcomes among depression, obesity, weight gain and other adversities.
  98. Zamboanga BL, Olthuis JV (SC '08 undergraduate co-author), Horton NJ, McCollum EC (SC '08 undergraduate co-author), Lee JJ, Shaw R. Where's the house party? Hazardous drinking behaviors and related risk factors, Journal of Psychology, 2009: 143(3):228-244.

    The authors examined differences in drinking behaviors and related risk factors across campus housing at a women's liberal arts college. Participants (N = 362) living in residence-style housing or house-style residences completed self-report questionnaires. Results showed that students in residence hall-style houses reported higher levels of hazardous alcohol use and perceived that their college, housemates or roommates, and close peers are more permissive of alcohol use than did students living in house-style residences. Findings highlight the role of the environmental structure of a college campus on students' perceptions of alcohol use and their drinking behaviors. The authors discuss implications for college housing and programming.
  99. Kypri K, Hallett J, Howat P, McManus A, Maycock B, Bowe S and Horton NJ. Randomized controlled trial of proactive web-based alcohol screening and brief intervention for university students., Archives of Internal Medicine, 2009; 169(16):1508-1514.

    BACKGROUND: University students drink more heavily than their nonstudent peers and are often unaware that their drinking is risky and exceeds normative levels. We tested the efficacy of a proactive Web-based alcohol screening and brief intervention program. METHODS: A randomized controlled trial was conducted at an Australian university in 2007. Invitations were sent to 13 000 undergraduates (age range, 17-24 years) to complete a Web-based Alcohol Use Disorders Identification Test. Of 7237 students who responded, 2435 scored in the hazardous/harmful range (> or =8) and were randomized, and 2050 (84%) completed at least 1 follow-up assessment. Intervention was 10 minutes of Web-based motivational assessment and personalized feedback. Controls received only screening. Follow-up assessments were conducted at 1 and 6 months with observers and participants blinded to allocation. Outcome measures were drinking frequency, typical occasion quantity, overall volume, number of personal problems, an academic problems score, prevalence of binge drinking, and prevalence of heavy drinking. RESULTS: Mean (SD) baseline Alcohol Use Disorders Identification Test scores for control and intervention groups were 14.3 (5.1) and 14.2 (5.1), respectively. After 1 month, participants receiving intervention drank less often (rate ratio [RR], 0.89; 95% confidence interval [CI], 0.83-0.94), smaller quantities per occasion (RR, 0.93; 95% CI, 0.88-0.98), and less alcohol overall (RR, 0.83; 95% CI, 0.78-0.90) than did controls. Differences in alcohol-related harms were nonsignificant. At 6 months, intervention effects persisted for drinking frequency (RR, 0.91; 95% CI, 0.85-0.97) and overall volume (RR, 0.89; 95% CI, 0.82-0.96) but not for other variables. CONCLUSION: Proactive Web-based screening and intervention reduces drinking in undergraduates, and such a program could be implemented widely.
  100. White IR, Horton NJ, Carpenter J. Complete-cases analysis is appropriate for randomised trials with pre-test-post-test designs (research letter), Psychiatry Research, 2009; 168(3):268.

  101. Kleinman KP and Horton NJ. SAS and R: Data management, statistical analysis and graphics, Chapman and Hall/CRC Press, 2010.

    This book shows how equivalent statistical methods can be applied in either SAS or R, enabling users of each software to learn how to apply the methods in the other. It covers data management, simple statistical procedures, modeling and regression, and graphics. Each section begins with a brief introduction to the procedures and then presents the code for each software side-by-side. The book provides detailed worked examples together with output from the software to illustrate how the methods are applied in practice. It also includes an index for both SAS and R to help facilitate use by a wide range of users.
  102. Horton NJ, White IR and Carpenter J. The performance of multiple imputation for missing covariates relative to complete case analysis (research letter), Statistics in Medicine, 2010; 29(12):1357.

  103. Voss SE, Adegoke MF (SC ’10 undergraduate co-author), Horton NJ, Sheth KN, Rosand J and Shera CA. Posture systematically alters ear-canal reflectance and DPOAE properties, Hearing Research, 2010; 263:43-51.

    Several studies have demonstrated that the auditory system is sensitive to changes in posture, presumably through changes in intracranial pressure (ICP) that in turn alter the intracochlear pressure, which affects the stiffness of the middle-ear system. This observation has led to efforts to develop an ear-canal based noninvasive diagnostic measure for monitoring ICP, which is currently monitored invasively via access through the skull or spine. Here, we demonstrate the effects of postural changes, and presumably ICP changes, on distortion product otoacoustic emissions (DPOAE) magnitude, DPOAE angle, and power reflectance. Measurements were made on 12 normal-hearing subjects in two postural positions: upright at 90 degrees and tilted at 45 degrees to the horizontal. Measurements on each subject were repeated five times across five separate measurement sessions. All three measures showed significant changes p < 0.001 between upright and tilted for frequencies between 500 and 2000 Hz, and DPOAE angle changes were significant at all measured frequencies (500–4000 Hz). Intra-subject variability, assessed via standard deviations for each subject's multiple measurements, were generally smaller in the upright position relative to the tilted position.
  104. Pfannkuch M, Regan M, Wild C, and Horton NJ. Telling data stories: essential dialogues for comparative reasoning, Journal of Statistics Education, 2010; 18(1).

    Language and the telling of data stories have fundamental roles in advancing the GAISE agenda of shifting the emphasis in statistics education from the operation of sets of procedures towards conceptual understanding and communication. In this paper we discuss some of the major issues surrounding story telling in statistics, challenge current practices, open debates about what constitutes good verbalization of structure in graphical and numerical summaries, and attempt to clarify what underlying concepts should be brought to students' attention, and how. Narrowing in on the particular problem of comparing groups, we propose that instead of simply reading and interpreting coded information from graphs, students should engage in understanding and verbalizing the rich conceptual repertoire underpinning comparisons using plots. These essential data-dialogues include paying attention to language, invoking descriptive and inferential thoughts, and determining informally whether claims can be made about the underlying populations from the sample data. A detailed teacher guide on comparative reasoning is presented and discussed.
  105. Wild CJ, Pfannkuch M, Regan M and Horton NJ. Towards more accessible conceptions of statistical inference, Journal of the Royal Statistical Society: Series A (Statistics in Society), 2011; 174(part 2):247-295 (read before the Royal Statistical Society on October 20, 2010).

    There is a compelling case, based on research in statistics education, for first courses in statistical inference to be underpinned by a staged development path. Preferably over a number of years, students should begin working with precursor forms of statistical inference, much earlier than they now do. A side benefit is giving younger students more straightforward and more satisfying ways of answering interesting real-world questions. We discuss the issues involved in formulating precursor versions of inference and then present some specific and highly visual proposals. These build upon novel ways of experiencing sampling variation and have intuitive connections to the standard formal methods of making inferences in first university courses in statistics. Our proposal uses visual comparisons to enable the inferential step to be made without taking the eyes off relevant graphs of the data. This allows the time and conceptual distances between questions, data and conclusions to be minimised, so that the most critical linkages can be made. Our approach was devised for use in high schools but is also relevant to adult education and some introductory tertiary courses.
  106. Murphy, JM, Gilman SE, Lesage A, Horton NJ, Rasic D, Trinh N-H, Alamiri B, Sobol AM, Fava M and Smoller J. Time trends in mortality associated with depression: findings from the Stirling County Study, Canadian Journal of Psychiatry (Revue Canadienne de Psychiatrie), 2010; 55(12):776-783.

  107. Merchant GR (SC '09 undergraduate first author), Voss SE and Horton NJ. Normative reflectance and transmittance measurements on healthy newborn and one-month old infants, Ear and Hearing, 2010; 31(6):746-54.

    OBJECTIVE: Ear-canal-based wideband reflectance (WBR) measurements may provide objective measures to assess and monitor middle-ear status in young babies. This work presents WBR measurements of power reflectance and transmittance on populations of healthy newborn babies (3 to 5 days) and healthy 1-mo-old babies (28 to 34 days). Thus, this work determines how power reflectance and transmittance vary between newborn and 1-mo-old babies and characterizes the range of these measures in normal populations. DESIGN: Power reflectance and transmittance were calculated from pressure measurements made in the ear canals of seven newborn (12 ears) and eleven 1-mo-old (19 ears) babies. Permutation tests, t tests, and regression (random effects) models were used to test the effects of age (newborn versus 1 mo), gender, and ear side (right versus left). RESULTS: The power reflectance and transmittance did not differ significantly for the age comparison (newborn versus 1 mo), although the results suggest a possible difference between newborn and 1-mo-old ears near 2000 Hz. There were no differences between the male and female ears. There are small but significant differences between left and right ears in three frequency bands encompassing 500 to 4000 Hz, where the predicted power reflectance mean for the left ear differs from the right ear by 0.02 to -0.07 depending on the frequency band. CONCLUSIONS: At most frequencies, power reflectance and transmittance are indistinguishable for newborn and 1-mo-old healthy babies, with limited or no differences between the two age groups and the males and females. There were small differences in some frequency bands for left and right ears. The measurements made here are similar to other published results in some frequency ranges but differ in other frequency ranges; differences among other studies from neonatal intensive care unit babies, healthy newborn babies, and healthy 1-mo-old babies are discussed.
  108. Horton NJ and Kleinman KP. Using R for data management, statistical analysis and graphics, Chapman and Hall/CRC Press, 2011.

    This book shows how statistical methods can be applied in R. It covers data management, simple statistical procedures, modeling and regression, and graphics. Each section begins with a brief introduction to the procedures and then presents the code for R. The book provides detailed worked examples together with output from the software to illustrate how the methods are applied in practice. It also includes an index for both syntax and concepts to help facilitate use by a wide range of users.
  109. Kleinman KP and Horton NJ. Using SAS for data management, statistical analysis and graphics, Chapman and Hall/CRC Press, 2011.

    This book shows how statistical methods can be applied in SAS. It covers data management, simple statistical procedures, modeling and regression, and graphics. Each section begins with a brief introduction to the procedures and then presents the code for SAS. The book provides detailed worked examples together with output from the software to illustrate how the methods are applied in practice. It also includes an index for both syntax and concepts to help facilitate use by a wide range of users.
  110. White IR, Horton NJ, Carpenter J and Pocock SJ. An intention-to-treat analysis strategy for randomized trials with missing outcome data, British Medical Journal, 2011, Feb 7, 342:d40.

    Loss to follow-up is often hard to avoid in randomised trials. This article suggests a framework for intention to treat analysis that depends on making plausible assumptions about the missing data and including all participants in sensitivity analyses.
  111. Field AE, Corliss HL, Skinner HH and Horton NJ. Loss of control eating as a predictor of weight gain and the development of overweight, depressive symptoms, binge drinking, and substance use. In Striegel-Moore R, Wonderlich SA, Walsh BT, & Mitchell JE (eds). Toward an Evidence-Based Classification of Eating Disorders. Arlington, VA: American Psychiatric Association; 2011:77-88.

  112. Caria MP, Bellocco R, Galanti MR and Horton NJ. The impact of different sources of body mass index assessment on smoking onset: an application of multiple informant models, Stata Journal, 2011; 11(3):386-402.

    Multiple-source data are often collected to provide better information of some underlying construct that is difficult to measure or likely to be missing. In this article, we describe regression-based methods for analyzing multiple-source data in Stata. We use data from the BROMS Cohort Study, a cohort of Swedish adolescents who collected data on body mass index that was self-reported and that was measured by nurses. We draw together into a single frame of reference both source reports and relate these to smoking onset. This unified method has two advantages over traditional approaches: 1) the relative predictiveness of each source can be assessed and 2) all subjects contribute to the analysis. The methods are applicable to other areas of epidemiology where multiple-source reports are used.
  113. Yoon FB, Fitzmaurice GM, Lipsitz SR, Horton NJ and Normand S-LT. Alternative methods for testing treatment effects on the basis of multiple outcomes: simulation and case study, Statistics in Medicine, 2011 Jul 20; 30(16):1917-32.

    In clinical trials multiple outcomes are often used to assess treatment interventions. This paper presents an evaluation of likelihood-based methods for jointly testing treatment effects in clinical trials with multiple continuous outcomes. Specifically, we compare the power of joint tests of treatment effects obtained from joint models for the multiple outcomes with univariate tests based on modeling the outcomes separately. We also consider the power and bias of tests when data are missing, a common feature of many trials, especially in psychiatry. Our results suggest that joint tests capitalize on the correlation of multiple outcomes and are more powerful than standard univariate methods, especially when outcomes are missing completely at random. When outcomes are missing at random, test procedures based on correctly specified joint models are unbiased, while standard univariate procedures are not. Results of a simulation study are reported, and the methods are illustrated in an example from the Clinical Antipsychotic Trials of Intervention Effectiveness for schizophrenia.
  114. Tyler KM-B (SC '09 undergraduate first author), Normand SL and Horton NJ. The use and abuse of multiple outcomes in randomized controlled trials of depression, Contemporary Clinical Trials, 2011; 32:299-304.

    OBJECTIVE: Multiple outcomes are commonly analyzed in randomized trials. Interpretation of the results of trials with many outcomes is not always straightforward. We characterize the prevalence and factors associated with multiple outcomes in reports of clinical trials of depression, methods used to account for these outcomes, and concordance between published analyses and original protocol specifications. METHODS: A PubMed search for randomized controlled depression trials that included multiple outcomes published between January 2007 and October 2008 in 6 medical journals. Original study protocols were reviewed where available. Parallel data collection by 2 abstractors was used to determine trial registration information, the number of outcomes, and analytical method. RESULTS: Of the 55 included trials, nearly half of the papers reported more than 1 primary outcome, while almost all (90.9%, n = 50) reported more than 2 combined primary or secondary outcomes. Relatively few of the studies (5.8%, n = 3) adjusted for multiple outcomes. While most studies had published protocols in clinical trial registries (76.4%, n = 42), many did not specify outcomes in the protocol (n = 11) and a number had discrepancies with the published report. CONCLUSIONS: Multiple outcomes are prevalent in randomized controlled depression trials and appropriate statistical analyses to account for these methods are rarely used. Not all studies filed protocols, and there were discrepancies between these protocols and published reports. These issues complicate interpretability of trial results, and in some cases may lead to spurious conclusions. Promulgation of guidelines to improve analysis and reporting of multiple outcomes is warranted.
  115. Horton NJ. Estimation of error covariances in a linear model, Stata Journal, 2011; 11(1):145-148.

  116. Konold K, Madden S, Pollatsek A, Pfannkuch M, Wild C, Ziedins I, Finzer W, Horton NJ and Kazak S. Conceptual challenges in coordinating theoretical and data-centered estimates of probability, Mathematical Thinking and Learning, 2011; 13:68-86.

    A core component of informal statistical inference is the recognition that judgments based on sample data are inherently uncertain. This implies that instruction aimed at developing informal inference needs to foster basic probabilistic reasoning. In this article, we analyze and critique the now-common practice of introducing students to both “theoretical” and “experimental” probability, typically with the hope that students will come to see the latter as converging on the former as the number of observations grows. On the surface of it, this approach would seem to fit well with objectives in teaching informal inference. However, our in-depth analysis of one eighth-grader's reasoning about experimental and theoretical probabilities points to various pitfalls in this approach. We offer tentative recommendations about how some of these issues might be addressed.
  117. White IR, Horton NJ, Carpenter J and Pocock SJ. Strategy for intention-to-treat analysis in randomized trials with missing outcome data, British Medical Journal, 2011, Feb 7; 342:d40.

    The intention to treat principle requires all participants in a clinical trial to be included in the analysis in the groups to which they were randomised, regardless of any departures from randomised treatment. This principle is a key defence against bias, since participants who depart from randomised treatment are usually a non-random subset whose exclusion can lead to serious selection bias. However, it is unclear how to apply the intention to treat principle when investigators are unable to follow up all randomised participants. Filling in (imputing) the missing values is often seen as the only alternative to omitting participants from the analysis. In particular, imputing by “last observation carried forward” is widely used, but this approach has serious drawbacks.3 For example, last observation carried forward was applied in a recent trial of a novel drug treatment in Alzheimer’s disease. The analysis was criticised because it effectively assumed that loss to follow-up halts disease progression, but the authors argued that their analysis was in fact conservative. Increasingly, trialists are expected to justify their handling of missing data and not simply rely on techniques that have been used in other clinical contexts. To guide investigators dealing with these tricky issues, we propose a four point framework for dealing with incomplete observations (box). Our aim is not to describe specific methods for analysing missing data, since these are described elsewhere, but to provide the framework within which methods can be chosen and implemented. We argue that all observed data should be included in the analysis, but undue focus on including all randomised participants can be unhelpful because participants with no post-randomisation data can contribute to the results only through untestable assumptions. The key issue is therefore not how to include all participants but what assumptions about the missing data are most plausibly correct, and how to perform appropriate analyses based on these assumptions.
  118. White IR, Carpenter J and Horton NJ. . Including all individuals is not enough: lessons for intention-to-treat analysis, Clinical Trials, 2012; 9(4):396-407.

    Abstract BACKGROUND: Intention-to-treat (ITT) analysis requires all randomised individuals to be included in the analysis in the groups to which they were randomised. However, there is confusion about how ITT analysis should be performed in the presence of missing outcome data. PURPOSES: To explain, justify, and illustrate an ITT analysis strategy for randomised trials with incomplete outcome data. METHODS: We consider several methods of analysis and compare their underlying assumptions, plausibility, and numbers of individuals included. We illustrate the ITT analysis strategy using data from the UK700 trial in the management of severe mental illness. RESULTS: Depending on the assumptions made about the missing data, some methods of analysis that include all randomised individuals may be less valid than methods that do not include all randomised individuals. Furthermore, some methods of analysis that include all randomised individuals are essentially equivalent to methods that do not include all randomised individuals. LIMITATIONS: This work assumes that the aim of analysis is to obtain an accurate estimate of the difference in outcome between randomised groups and not to obtain a conservative estimate with bias against the experimental intervention. CONCLUSIONS: Clinical trials should employ an ITT analysis strategy, comprising a design that attempts to follow up all randomised individuals, a main analysis that is valid under a stated plausible assumption about the missing data, and sensitivity analyses that include all randomised individuals in order to explore the impact of departures from the assumption underlying the main analysis. Following this strategy recognises the extra uncertainty arising from missing outcomes and increases the incentive for researchers to minimise the extent of missing data.
  119. Sonneville KR, Calzo JP, Horton NJ, Austin SB, Haines J and Field AE. Body satisfaction, weight gain and binge eating among overweight adolescent girls., International Journal of Obesity, , 2012; 36(7):944-9.

    OBJECTIVE: To examine if body satisfaction is associated with body mass index (BMI) change and whether it protects against the development of frequent binge eating among overweight and obese adolescent girls. METHODS: We used prospective data from nine waves of an ongoing cohort study of adolescents, the Growing Up Today Study. At enrollment in 1996, participants were 9-14 years old. Questionnaires were mailed to participants annually until 2001, then biennially through 2007. Girls who were overweight or obese in 1996 were included in the analysis (n = 1559). Our outcomes were annual change in BMI and incident frequent binge eating, defined as binge eating at least weekly and no use of compensatory behaviors. RESULTS: At baseline, 57.2% of the overweight and obese girls were at least somewhat satisfied with their bodies. During 11 years of follow-up, 9.5% (95% confidence interval (CI) (7.8, 10.8)) of the girls started to binge eat frequently. Controlling for BMI and other confounders, overweight and obese girls who reported being at least somewhat satisfied with their bodies made smaller BMI gains (β = -0.10 kg m(-2), 95% CI (-0.19, -0.02)) and had 61% lower odds of starting to binge eat frequently (odds ratio (OR) = 0.39, 95% CI (0.24, 0.64)) than their less satisfied peers. Compared with girls who were the least satisfied with their bodies, girls who were the most satisfied had 85% lower odds of starting to binge eat frequently (OR = 0.15, 95% CI (0.06, 0.37)). The association between body satisfaction and starting to binge eat frequently was stronger for younger adolescents than older adolescents. CONCLUSIONS: Whereas body dissatisfaction is common among overweight and obese girls, body satisfaction may protect against excessive weight gain and binge eating. Prevention of body dissatisfaction must begin early and should be considered as a component of both obesity and eating disorder prevention programs.
  120. Voss SE, Merchant GR (SC '09 undergraduate co-author) and Horton NJ. Effects of middle-ear disorders on power reflectance measured in cadaveric ear canals, Ear and Hearing, 2012; 33(2):195-208.

    OBJECTIVE: Reflectance measured in the ear canal offers a noninvasive method to monitor the acoustic properties of the middle ear, and few systematic measurements exist on the effects of various middle-ear disorders on the reflectance. This work uses a human cadaver-ear preparation and a mathematical middle-ear model to both measure and predict how power reflectance R is affected by the middle-ear disorders of static middle-ear pressures, middle-ear fluid, fixed stapes, disarticulated incudostapedial joint, and tympanic-membrane perforations. DESIGN: R was calculated from ear-canal pressure measurements made on human-cadaver ears in the normal condition and five states: (1) positive and negative pressure in the middle-ear cavity, (2) fluid-filled middle ear, (3) stapes fixed with dental cement, (4) incudostapedial joint disarticulated, and (5) tympanic-membrane perforations. The middle-ear model of Kringlebotn (1988) was modified to represent the middle-ear disorders. Model predictions are compared with measurements. RESULTS: For a given disorder, the general trends of the measurements and model were similar. The changes from normal in R, induced by the simulated disorder, generally depend on frequency and the extent of the disorder (except for the disarticulation). Systematic changes in middle-ear static pressure (up to ±300 daPa) resulted in systematic increases in R. These affects were most pronounced for frequencies up to 1000 to 2000 Hz. Above about 2000 Hz there were some asymmetries in behavior between negative and positive pressures. Results with fluid in the middle-ear air space were highly dependent on the percentage of the air space that was filled. Changes in R were minimal when a smaller fraction of the air space was filled with fluid, and as the air space was filled with more saline, R increased at most frequencies. Fixation of the stapes generally resulted in a relatively small low-frequency increase in R. Disarticulation of the incus with the stapes led to a consistent low-frequency decrease in R with a distinctive minimum below 1000 Hz. Perforations of the tympanic membrane resulted in a decrease in R for frequencies up to about 2000 Hz; at these lower frequencies, smaller perforations led to larger changes from normal when compared with larger perforations. CONCLUSIONS: These preliminary measurements help assess the utility of power reflectance as a diagnostic tool for middle-ear disorders. In particular, the measurements document (1) the frequency ranges for which the changes are largest and (2) the extent of the changes from normal for a spectrum of middle-ear disorders.
  121. Field AE, Sonneville KR, Micali N, Crosby RD, Swanson SA, Laird NM, Treasure J, Solmi F, Horton NJ. Prospective association of common eating disorders and adverse outcomes, Pediactrics, 2012; 130(2):e289-95.

    Abstract OBJECTIVE: Anorexia nervosa and bulimia nervosa (BN) are rare, but eating disorders not otherwise specified (EDNOS) are relatively common among female participants. Our objective was to evaluate whether BN and subtypes of EDNOS are predictive of developing adverse outcomes. METHODS: This study comprised a prospective analysis of 8594 female participants from the ongoing Growing Up Today Study. Questionnaires were sent annually from 1996 through 2001, then biennially through 2007 and 2008. Participants who were 9 to 15 years of age in 1996 and completed at least 2 consecutive questionnaires between 1996 and 2008 were included in the analyses. Participants were classified as having BN (≥ weekly binge eating and purging), binge eating disorder (BED; ≥ weekly binge eating, infrequent purging), purging disorder (PD; ≥ weekly purging, infrequent binge eating), other EDNOS (binge eating and/or purging monthly), or nondisordered. RESULTS: BN affected ∼1% of adolescent girls; 2% to 3% had PD and another 2% to 3% had BED. Girls with BED were almost twice as likely as their nondisordered peers to become overweight or obese (odds ratio [OR]: 1.9 [95% confidence interval: 1.0-3.5]) or develop high depressive symptoms (OR: 2.3 [95% confidence interval: 1.0-5.0]). Female participants with PD had a significantly increased risk of starting to use drugs (OR: 1.7) and starting to binge drink frequently (OR: 1.8). CONCLUSIONS: PD and BED are common and predict a range of adverse outcomes. Primary care clinicians should be made aware of these disorders, which may be underrepresented in eating disorder clinic samples. Efforts to prevent eating disorders should focus on cases of subthreshold severity.
  122. Sonneville KR, Horton NJ, Micali N, Crosby RD, Swanson SA, Solmi F and Field AE. Longitudinal associations between binge eating and overeating and adverse outcomes among adolescents and young adults: does loss of control matter?, Archives of Pediatric and Adolescent Medicine , 2013; 167(2):149-55.

    OBJECTIVE: To investigate the association between overeating (without loss of control) and binge eating (overeating with loss of control) and adverse outcomes. DESIGN: Prospective cohort study. SETTING: Adolescents and young adults living throughout the United States. PARTICIPANTS: Sixteen thousand eight hundred eighty-two males and females participating in the Growing Up Today Study who were 9 to 15 years old at enrollment in 1996. MAIN EXPOSURE: Overeating and binge eating assessed via questionnaire every 12 to 24 months between 1996 and 2005. MAIN OUTCOME MEASURES: Risk of becoming overweight or obese, starting to binge drink frequently, starting to use marijuana, starting to use other drugs, and developing high levels of depressive symptoms. Generalized estimating equations were used to estimate associations. All models controlled for age and sex; additional covariates varied by outcome. RESULTS: Among this large cohort of adolescents and young adults, binge eating was more common among females than males. In fully adjusted models, binge eating, but not overeating, was associated with incident overweight/obesity (odds ratio, 1.73; 95% CI, 1.11-2.69) and the onset of high depressive symptoms (odds ratio, 2.19; 95% CI, 1.40-3.45). Neither overeating nor binge eating was associated with starting to binge drink frequently, while both overeating and binge eating predicted starting to use marijuana and other drugs. CONCLUSIONS: Although any overeating, with or without loss of control, predicted the onset of marijuana and other drug use, we found that binge eating is uniquely predictive of incident overweight/obesity and the onset of high depressive symptoms. These findings suggest that loss of control is an important indicator of severity of overeating episodes.
  123. Kypros K, McCambridge J, Vater T, Bowe SJ, Saunders JB, Cunningham JA and Horton NJ. Web-based alcohol intervention for Māori university students: double-blind, multi-site randomized controlled trial., Addiction, 2013; 108(2):331-8.

    AIMS: Like many indigenous peoples, New Zealand Māori bear a heavy burden of alcohol-related harm relative to their non-indigenous compatriots, and disparities are greatest among young adults. We tested the effectiveness of web-based alcohol screening and brief intervention (e-SBI) for reducing hazardous drinking among Māori university students. DESIGN: Parallel, double-blind, multi-site, randomized controlled trial. SETTING: Seven of New Zealand's eight universities. PARTICIPANTS: In April 2010, we sent e-mail invitations to all 6697 17-24-year-old Māori students to complete a brief web questionnaire including the Alcohol Use Disorders Identification Test (AUDIT)-C, a screening tool for hazardous and harmful drinking. Those screening positive were computer randomized to: <10 minutes of web-based alcohol assessment and personalized feedback (intervention) or screening alone (control). MEASUREMENTS: We conducted a fully automated 5-month follow-up assessment with observers and participants blinded to study hypotheses, design and intervention delivery. Pre-determined primary outcomes were: (i) frequency of drinking, (ii) amount consumed per typical drinking occasion, (iii) overall volume of alcohol consumed and (iv) academic problems. FINDINGS: Of the participants, 1789 were hazardous or harmful drinkers (AUDIT-C ≥ 4) and were randomized: 850 to control, 939 to intervention. Follow-up assessments were completed by 682 controls (80%) and 733 intervention group members (78%). Relative to controls, participants receiving intervention drank less often [RR = 0.89; 95% confidence interval (CI): 0.82-0.97], less per drinking occasion (RR = 0.92; 95% CI: 0.84-1.00), less overall (RR = 0.78; 95% CI: 0.69-0.89) and had fewer academic problems (RR = 0.81; 95% CI: 0.69-0.95). CONCLUSIONS: Web-based screening and brief intervention reduced hazardous and harmful drinking among non-help-seeking Māori students in a large-scale pragmatic trial. The study has wider implications for behavioural intervention in the important but neglected area of indigenous health.
  124. Bell ML, Kenward MG, Fairclough DL, Horton NJ. Differential dropout and bias in randomised controlled trials: when it matters and when it may not., British Medical Journal, 2013; 346:e8668.

    Dropout in randomised controlled trials is common and threatens the validity of results, as completers may differ from people who drop out. Differing dropout rates between treatment arms is sometimes called differential dropout or attrition. Although differential dropout can bias results, it does not always do so. Similarly, equal dropout may or may not lead to biased results. Depending on the type of missingness and the analysis used, one can get a biased estimate of the treatment effect with equal dropout rates and an unbiased estimate with unequal dropout rates. We reinforce this point with data from a randomised controlled trial in patients with renal cancer and a simulation study.
  125. Borsari B, Zamboanga BL, Correia CJ, Olthuis VJ, Van Tyne K, Zadworny Z (SC ’13 undergraduate co-author), Grossbard JR and Horton NJ. Characterizing high school students who play drinking games using latent class analysis, Addictive Behaviors, 2013; 38(10):2532-2540.

    Heavy alcohol use and its associated negative consequences continue to be an important health issue among adolescents. Of particular concern are risky drinking practices such as playing drinking games. Although retrospective accounts indicate that drinking game participation is common among high school students, it has yet to be assessed in current high school students. Utilizing data from high school students who reported current drinking game participation (n=178), we used latent class analysis to investigate the negative consequences resulting from gaming and examined underlying demographic and alcohol-related behavioral characteristics of students as a function of the resultant classes. Three classes of "gamers" emerged: (1) a "lower-risk" group who had a lower probability of endorsing negative consequences compared to the other groups, (2) a "higher-risk" group who reported that they experienced hangovers and difficulties limiting their drinking, got physically sick, and became rude, obnoxious, or insulting, and (3) a "sexual regret" group who reported that they experienced poor recall and unplanned sexual activity that they later regretted. Although the frequency of participating in drinking games did not differ between these three groups, results indicated that the "lower-risk" group consumed fewer drinks in a typical gaming session compared to the other two groups. The present findings suggest that drinking games are common among high school students, but that mere participation and frequency of play are not necessarily the best indicators of risk. Instead, examination of other constructs such as game-related alcohol consumption, consequences, or psychosocial variables such as impulsivity may be more useful.
  126. Elwy AR, Horton NJ, and Saitz R. Physicians' attitudes toward unhealthy alcohol use and self-efficacy for screening and counseling as predictors of their counseling and primary care patients' drinking outcomes., Substance Abuse Treatment, Prevention and Policy, 2013; 8(1):17.

    OBJECTIVE: Patients' unhealthy alcohol use is often undetected in primary care. Our objective was to examine whether physicians' attitudes and their perceived self-efficacy for screening and counseling patients is associated with physicians' counseling of patients with unhealthy alcohol use, and patients' subsequent drinking. METHODS: This study is a prospective cohort study (nested within a randomized trial) involving 41 primary care physicians and 301 of their patients, all of whom had unhealthy alcohol use. Independent variables were physicians' attitudes toward unhealthy substance use and self-efficacy for screening and counseling. Outcomes were patients' reports of physicians' counseling about unhealthy alcohol use immediately after a physician visit, and patients' drinking six months later. RESULTS: Neither physicians' attitudes nor self-efficacy had any impact on physicians' counseling, but greater perceived self-efficacy in screening, assessing and intervening with patients was associated with more drinking by patients six months later. CONCLUSIONS: Future research needs to further explore the relationship between physicians' attitudes towards unhealthy alcohol use, their self-efficacy for screening and counseling and patients' drinking outcomes, given our unexpected findings.
  127. Abur D (SC '13 undergraduate co-author), Horton NJ, and Voss SE. Intra-subject variability in power reflectance, Journal of the American Academy of Audiology, in press.

    Background: Power reflectance measurements are an active area of research related to the development of noninvasive middle-ear assessment methods. There are limited data related to test-retest measures of power reflectance. Purpose: The study investigates test-retest features of power reflectance, including comparisons of intra- versus inter-subject variability and how ear-canal measurement location a↵ects measurements. Research Design: Repeated measurements of power reflectance were made at approximately weekly intervals. The subjects returned for four to eight sessions. Measurements were made at three ear-canal locations: a deep insertion depth (foam plug being flush at the entrance to the ear canal), and both 3 and 6 mm more lateral to this deep insertion. Study Sample: Repeated measurements on seven subjects are reported. All subjects were females between 19 and 22 years old. They were selected because the study was done at an undergraduate women’s college and these were the demographics of the available subjects. Data Collection and Analysis: Measurements on both the right and left ears were made at three ear-canal locations during each of four to eight measurement sessions. Random effects regression models were used for the analysis to account for repeated measures within subjects. The mean power reflectance for each position over all sessions was calculated for each subject. Results: The comparison of power reflectance from the left and right ears of an individual subject varied greatly over the seven subjects; the difference between the power reflectance measured on the left to that measured on the right was compared at 248 frequencies, and depending on the subject, the percentage of tested frequencies for which the left and right differed significantly ranged from 10 to 93 percent. While the individual subjects showed left-right differences, the overall population generally did not show significant differences between left and right ears. The mean power reflectance for each measurement position over all sessions depended on the location of the probe in the ear for frequencies below 1000 Hz. The standard deviation between 47 subjects mean power reflectance, after controlling for ear (left or right), was found to be greater than the standard deviation within individual subject’s mean power reflectance. The intra-subject standard deviation in power reflectance was smallest at the deepest insertion depths. Conclusions: All subjects had differences in power reflectance between their left and right ears at some frequencies; the percentage of frequencies where differences occurred varied greatly across subjects. The intra-subject standard deviations were smallest for the deepest probe insertion depths, suggesting clinical measurements should be made with as deep an insertion as practically possible in order to minimize variability. This deep insertion will reduce both acoustic leaks and the effect of low-frequency ear-canal losses. The within subject standard deviations were about half the size of the overall standard deviations, quantifying the extent of intra- versus inter-subject variability.
  128. Reviews

    Horton NJ. Harvard School of Public Health tracks research projects with CrossGraphs, DM Review Magazine, 1999.

    Horton, NJ. review of Brown and Prescott, Applied Mixed Models in Medicine, Statistical Methods in Medical Research, 2003; 12(1):88-89.

    Horton, NJ. review of Rabe-Hesketh and Skrondal, Multilevel and longitudinal modeling using Stata, The American Statistician, 2006; 60(3):293.

    Horton, NJ. review of Rabe-Hesketh and Skrondal, Multilevel and longitudinal modeling using Stata (2nd edition), Stata Journal, 2008; 8(1):579-582.