On the implication of structural zeros as independent variables in regression analysis: applications to alcohol research

In alcohol studies, drinking outcomes such as number of days of any alcohol drinking (DAD) over a period of time do not precisely capture the differences among subjects in a study population of interest. For example, the value of 0 on DAD could mean that the subject was continually abstinent from drinking such as lifetime abstainers or the subject was alcoholic, but happened not to use any alcohol during the period of interest. In statistics, zeros of the first kind are called structural zeros, to distinguish them from the sampling zeros of the second type. As the example indicates, the structural and sampling zeros represent two groups of subjects with quite different psychosocial outcomes. In the literature on alcohol use, although many recent studies have begun to explicitly account for the differences between the two types of zeros in modeling drinking variables as a response, none has acknowledged the implications of the different types of zeros when such modeling drinking variables are used as a predictor. This paper serves as the first attempt to tackle the latter issue and illustrate the importance of disentangling the structural and sampling zeros by using simulated as well as real study data.

Causal inference for community-based multi-layered intervention study

Estimating causal treatment effect for randomized controlled trials under post-treatment confounding, that is, noncompliance and informative dropouts, is becoming an important problem in intervention/prevention studies when the treatment exposures are not completely controlled. When confounding is present in a study, the traditional intention-to-treat approach could underestimate the treatment effect because of insufficient exposure of treatment. In the recent two decades, many papers have been published to address such confounders to investigate the causal relationship between treatment and outcome of interest based on different modeling strategies. Most of the existing approaches, however, are suitable only for standard experiments. In this paper, we propose a new class of structural functional response model to address post-treatment confounding in complex multi-layered intervention studies within a longitudinal data setting. The new approach offers robust inference and is readily implemented. We illustrate and assess the performance of the proposed structural functional response model using both real and simulated data.

Parent and child psychopathology and suicide attempts among children of parents with alcohol use disorder

Parents with psychopathology such as alcohol use disorder (AUD) that confers risk for suicide attempt (SA) may have children who are more likely to develop such psychopathology and to attempt suicide, suggesting that risk may be "transmitted" from parents to children. We examined this phenomenon during the transition from childhood to adolescence, when risk for SA increases dramatically. A cohort of 418 children were examined at average age 9.4 (range 7-14) years at enrollment (Time 1, childhood) and approximately 5 years later, prior to reaching age 18 (Time 2, adolescence). One or both biological parents, oversampled for AUD, were also interviewed. Structural equation models (SEM) examined father-child, mother-child, and either/both parent-child associations. The primary outcome was SA over follow-up among offspring, assessed at Time 2. As hypothesized, parental antisocial personality disorder predicted conduct disorder symptoms in offspring both during childhood and adolescence (parent-child model, father-child model) and maternal AUD predicted conduct disorder symptoms during childhood (mother-child model). However, we did not find evidence to support transmission of depression from parents to offspring either during childhood or adolescence, and parent psychopathology did not show statistically significant associations with SA during adolescence. In conclusion, we conducted a rare study of parent-to-child "transmission" of risk for SA that used a prospective research design, included diagnostic interviews with both parents and offspring, and examined the transition from childhood to adolescence, and the first such study in children of parents with AUD. Results provided mixed support for hypothesized parent-child associations.

On assessing model fit for distribution-free longitudinal models under missing data

The generalized estimating equation (GEE), a distribution-free, or semi-parametric, approach for modeling longitudinal data, is used in a wide range of behavioral, psychotherapy, pharmaceutical drug safety, and healthcare-related research studies. Most popular methods for assessing model fit are based on the likelihood function for parametric models, rendering them inappropriate for distribution-free GEE. One rare exception is a score statistic initially proposed by Tsiatis for logistic regression (1980) and later extended by Barnhart and Willamson to GEE (1998). Because GEE only provides valid inference under the missing completely at random assumption and missing values arising in most longitudinal studies do not follow such a restricted mechanism, this GEE-based score test has very limited applications in practice. We propose extensions of this goodness-of-fit test to address missing data under the missing at random assumption, a more realistic model that applies to most studies in practice. We examine the performance of the proposed tests using simulated data and demonstrate the utilities of such tests with data from a real study on geriatric depression and associated medical comorbidities.

Distribution-free models for longitudinal count responses with overdispersion and structural zeros

Overdispersion and structural zeros are two major manifestations of departure from the Poisson assumption when modeling count responses using Poisson log-linear regression. As noted in a large body of literature, ignoring such departures could yield bias and lead to wrong conclusions. Different approaches have been developed to tackle these two major problems. In this paper, we review available methods for dealing with overdispersion and structural zeros within a longitudinal data setting and propose a distribution-free modeling approach to address the limitations of these methods by utilizing a new class of functional response models. We illustrate our approach with both simulated and real study data.

Mood, anxiety, and substance-use disorders and suicide risk in a military population cohort

There are meager prospective data from nonclinical samples on the link between anxiety disorders and suicide or the extent to which the association varies over time. We examined these issues in a cohort of 309,861 U.S. Air Force service members, with 227 suicides over follow-up. Mental disorder diagnoses including anxiety, mood, and substance-use disorders (SUD) were based on treatment encounters. Risk for suicide associated with anxiety disorders were lower compared with mood disorders and similar to SUD. Moreover, the associations between mood and anxiety disorders with suicide were greatest within a year of treatment presentation.

The hypertension scale of the system of Quality of Life Instruments for Chronic Diseases, QLICD-HY: a development and validation study

AIMS

This paper is aimed to develop and validate the hypertension scale of the system of Quality of Life Instruments for Chronic Diseases, QLICD-HY.

METHODS

The QLICD-HY instrument was developed based on programmed decision procedures with multiple nominal and focus group discussions and pilot testing. A total of 157 inpatients with hypertension were used to provide the data measuring QOL three times before and after treatment. The psychometric properties of the scale were evaluated with respect to validity, reliability and responsiveness employing correlation and factor analyses, and t-tests.

RESULTS

Correlation and factor analyses confirmed good construct validity and criterion-related validity when using Short Form (36) Health Survey (SF-36) as a criterion. Test-retest reliability coefficients (Pearson r and intra-class correlation (ICC)) for the overall instrument score and all domains except for the hypertension-specific domain (SPD) (0.75) were higher than 0.80 with a range of 0.75-0.91; the internal consistency α for all domains except for the hypertension-specific domain (0.66) was higher than 0.70. The overall score and scores for most facets within each domain except for the social domain (SOD) had statistically significant changes (t-tests) after treatment with moderate effect sizes.

CONCLUSION

QLICD-HY has good validity, reliability, responsiveness and can be used as the quality-of-life instrument for patients with hypertension.

On fitting generalized linear mixed-effects models for binary responses using different statistical packages

The generalized linear mixed-effects model (GLMM) is a popular paradigm to extend models for cross-sectional data to a longitudinal setting. When applied to modeling binary responses, different software packages and even different procedures within a package may give quite different results. In this report, we describe the statistical approaches that underlie these different procedures and discuss their strengths and weaknesses when applied to fit correlated binary responses. We then illustrate these considerations by applying these procedures implemented in some popular software packages to simulated and real study data. Our simulation results indicate a lack of reliability for most of the procedures considered, which carries significant implications for applying such popular software packages in practice.

Detection of depression in older adults by family and friends: distinguishing mood disorder signals from the noise of personality and everyday life

BACKGROUND

The capacity of friends and family member informants to make judgments about the presence of a mood disorder history in an older primary care patient has theoretical, clinical, and public health significance. This study examined the accuracy of informant-reported mood disorder diagnoses in a sample of primary care patients aged 65 years or older. We hypothesized that the accuracy (sensitivity and specificity) of informant reports would vary with the patient's personality.

METHODS

Hypotheses were tested in 191 dyads consisting of patients and their friends or relatives (informants) recruited from primary care settings. Gold-standard mood disorder diagnoses were established at consensus conferences based on a review of medical charts and data collected in a structured interview with the patient. Patients completed an assessment battery that included the NEO-Five Factor Inventory.

RESULTS

Sensitivity and specificity of informant-derived mood disorder diagnoses were related to patient personality. Sensitivity of informant-derived lifetime mood disorder diagnoses was compromised by higher Extraversion and higher Agreeableness. Specificity of informant-derived lifetime mood disorder diagnoses was compromised by lower Agreeableness and higher Conscientiousness.

CONCLUSION

Patient personality has implications for the accuracy of mood disorder histories provided by friends and family members. Given that false negatives can have grave consequences, we recommend that practitioners be particularly vigilant when interpreting collateral information about their extraverted, agreeable patients.

Test of a clinical model of drinking and suicidal risk

BACKGROUND

There are few data on the role of drinking patterns in suicidal thoughts or behavior among alcohol-dependent individuals (ADIs) and meager data on variables that may influence the role of drinking in suicidal thoughts and behavior. This study tested a heuristic model that predicts that drinking promotes suicidal thoughts and behavior, the association is mediated (accounted for) by depressive symptoms, and that anger moderates (increases) the risk associated with intense drinking.

METHODS

Data from Project MATCH, a multisite alcohol use disorders treatment trial, were analyzed using structural equation modeling. There were 1,726 participants including 24% women and a mean age of 40.2 ± 11.0 years. Subjects were assessed at baseline and at 3-, 9-, and 15-month follow-up. Two categorical measures (presence/absence) of suicidal ideation (SI) were used that were analyzed in separate models. Predictors of interest were continuous assessments of average drinking intensity (i.e., drinks per drinking day or DDD), drinking frequency (i.e., percent days abstinent or PDA), depression, and anger.

RESULTS

Both DDD and PDA were associated with SI at a statistically significant level, with PDA showing an inverse association. Depression scores served as a partial mediator or a full mediator of the drinking-SI relationship depending on the measure of SI used in the analysis. The models testing anger scores as a moderator fit the data poorly and did not support that anger serves as a moderator of the drinking-SI association.

CONCLUSIONS

Greater drinking intensity and drinking frequency predict SI among ADIs and depression serves as a mediator of these associations, but anger does not appear to serve as a moderator. Further research is required to clarify whether depression serves as a partial or full mediator and to see whether the results herein extend to suicidal behavior (i.e., suicide attempt, suicide).

A U-statistics-based approach for modeling Cronbach coefficient alpha within a longitudinal data setting

Cronbach coefficient alpha (CCA) is a classic measure of item internal consistency of an instrument and is used in a wide range of behavioral, biomedical, psychosocial, and health-care-related research. Methods are available for making inference about one CCA or multiple CCAs from correlated outcomes. However, none of the existing approaches effectively address missing data. As longitudinal study designs become increasingly popular and complex in modern-day clinical studies, missing data have become a serious issue, and the lack of methods to systematically address this problem has hampered the progress of research in the aforementioned fields. In this paper, we develop a novel approach to tackle the complexities involved in addressing missing data (at the instrument level due to subject dropout) within a longitudinal data setting. The approach is illustrated with both clinical and simulated data.

Effects of pretreatment and posttreatment depressive symptoms on alcohol consumption following treatment in Project MATCH

OBJECTIVE

This study examined the associations of pretreatment and posttreatment depressive symptoms with drinking outcomes in the year following treatment in Project MATCH (Matching Alcoholism Treatments to Client Heterogeneity), a multisite clinical trial of behavioral treatments for alcohol-use disorders.

METHOD

Data from 1,726 participants were modeled using generalized estimating equations to examine drinking frequency and intensity, as reflected by percentage days abstinent (PDA) and average drinks per drinking day (DDD). We predicted that patients who reported more pretreatment and posttreatment depressive symptoms would report greater drinking frequency (PDA) and more intense drinking (DDD) across the 12-month follow-up period.

RESULTS

Pretreatment Beck Depression Inventory (BDI) scores predicted more frequent and intense drinking in the year following treatment, although not after accounting for posttreatment BDI scores, which were associated with the drinking outcomes as hypothesized. Patients who reported more depressive symptoms in the year following treatment reported less abstinence (PDA) and more intense drinking (DDD) than patients with fewer posttreatment depressive symptoms.

CONCLUSIONS

Our findings underscore the importance of obtaining repeated assessments of depression during the course of substance use treatment. Moreover, the data suggest the potential utility of augmenting standard chemical dependency care with depression-focused interventions for alcohol-dependent patients whose depressive symptoms do not subside during treatment.

A New Analytic Framework for Moderation Analysis --- Moving Beyond Analytic Interactions

Conceptually, a moderator is a variable that modifies the effect of a predictor on a response. Analytically, a common approach as used in most moderation analyses is to add analytic interactions involving the predictor and moderator in the form of cross-variable products and test the significance of such terms. The narrow scope of such a procedure is inconsistent with the broader conceptual definition of moderation, leading to confusion in interpretation of study findings. In this paper, we develop a new approach to the analytic procedure that is consistent with the concept of moderation. The proposed framework defines moderation as a process that modifies an existing relationship between the predictor and the outcome, rather than simply a test of a predictor by moderator interaction. The approach is illustrated with data from a real study.

Adapting Interpersonal Psychotherapy for Older Adults at Risk for Suicide: Preliminary Findings

We report preliminary findings of the first ever study testing a 16-week course of Interpersonal Psychotherapy (IPT) modified for older outpatients at elevated risk for suicide. Participants were referred from inpatient and outpatient medicine and mental health services. Psychotherapy sessions took place in a therapist's office in a teaching hospital. Twelve adults 60 years or older (M=70.5, SD=6.1) with current thoughts of suicide (suicide ideation) or a wish to die (death ideation) or with recent self-injurious behavior were recruited into weekly sessions of IPT; one was subsequently excluded due to severe cognitive impairment. Participants completed measures of suicide ideation, death ideation, and depressive symptom severity at pre-treatment, mid-treatment, post-treatment, and at 3-month follow-up periods, and measures of therapeutic process variables. Preliminary findings of this uncontrolled pre-post-treatment study support the feasibility of recruiting and retaining older adults at-risk for suicide into psychotherapy research and suggest that adapted IPT is tolerable and safe. Findings indicate a substantial reduction in participant suicide ideation, death ideation, and depressive symptoms; controlled trials are needed to further evaluate these findings. We discuss implications for clinical care with at-risk older adults.

Hot flashes, fatigue, treatment exposures and work productivity in breast cancer survivors

INTRODUCTION

While fatigue has been associated with work limitations the combined influence of specific diagnosis and treatment exposures based on medical records on work limitations in breast cancer survivors is currently unknown. Since symptom burden and perceived health can interfere with work, the present study investigated the relationship among these variables and work outcomes.

METHODS

Medical chart abstraction, demographic measures, SF 36, the Work Limitations Questionnaire (WLQ) and measures of symptom burden, including hot flushes were obtained in 83 breast cancer survivors a mean of three years post treatment. OLS and poisson regression were used to determine the relationship of these factors to work productivity and work absences.

RESULTS

Breast cancer survivors reported a mean reduction in productivity of 3.1% below the healthy worker norm. This amounts to a loss of 2.48 hours of work over two weeks of full time employment. Stages 1 and 2 were related to work limitations. After controlling for stage, fatigue and hot flashes were each associated with work performance losses of 1.6% (p = 0.05) and 2.2% (p < 0.001), respectively. Protective factors included marriage and greater personal earned income.

CONCLUSIONS

Fatigue and hot flashes are important factors related to work productivity in breast cancer survivors even at three years post treatment. IMPLICATIONS FOR SURVIVORS: Therapy for hot flashes should be given serious consideration in breast cancer survivors who are experiencing work limitations.

Comparing Multiple Sensitivities and Specificities with Different Diagnostic Criteria: Applications to Sexual Abuse and Sexual Health Research

When comparing sensitivities and specificities from multiple diagnostic tests, particularly in biomedical research, the different test kits under study are applied to groups of subjects with the same disease status for a disease or medical condition under consideration. Although this process gives rise to clustered or correlated test outcomes, the associated inference issues are well recognized and have been widely discussed in the literature. In mental health and psychosocial research, sensitivity and specificity have also been widely used to study the reliability of instrument for diagnosing mental health and psychiatric conditions and assessing certain behavioral patterns. However, unlike biomedical applications, outcomes are often obtained under varying reference standards or different diagnostic criteria, precluding the application of existing methods for comparing multiple diagnostic tests to such a research setting. In this paper, we develop a new approach to address these problems (including that of missing data) by extending recent work on inference using inverse probability weighted estimates. The approach is illustrated with data from two studies in sexual abuse and health research as well as a limited simulation study, with the latter used to study the performance of the proposed procedure.

Predicting sexual risk behaviors among adolescent and young women using a prospective diary method

We describe the sexual risk behaviors, psychological distress, and substance use of 102 late adolescent girls and identify predictors of protected and unprotected vaginal sex. Participants completed questionnaires assessing hypothesized predictors and then daily behavioral diaries for 12 weeks. Protected intercourse was predicted by baseline sexual behavior, greater knowledge, positive condom attitudes, lower perceived condom-use difficulty, greater condom-use intentions, more drinking days, less binge drinking, less Ecstasy use, and lower psychological distress. Unprotected intercourse was predicted by baseline sexual behavior, binge drinking, Ecstasy and opiate use, fewer drinking days, and fewer daily drinks. These findings suggest that psychological distress, substance use, and sexual risk behavior are interconnected and should be considered collectively in interventions for adolescent females.

Correlation analysis for longitudinal data: applications to HIV and psychosocial research

Correlation analysis is widely used in biomedical and psychosocial research for assessing rater reliability, precision of diagnosis and accuracy of proxy outcomes. The popularity of longitudinal study designs has propelled the proliferation in recent years of new methods for longitudinal and other multi-level clustered data designs, such as the mixed-effect models and generalized estimating equations. Despite these advances, research and methodological development on addressing missing data for correlation analysis is woefully lacking. In this paper, we consider non-parametric inference for the product-moment correlation within a longitudinal data setting and address missing data under both the missing completely at random and missing at random assumptions. We illustrate the approach with real study data in mental health and HIV prevention research.

Religious involvement and depressive symptoms in primary care elders

BACKGROUND

Multiple lines of evidence indicate relationships between religious involvement and depression, although the specific nature of the relationships is yet to be clarified. Moreover, there appear to be no well controlled longitudinal studies to date examining this issue in primary care elders.

METHOD

The authors assessed the linear and non-linear relationships between three commonly identified types of religious involvement and observer-rated depressive symptoms in 709 primary care elders assessed at baseline and 1-year follow-up.

RESULTS

Cross-sectional analyses revealed a curvilinear, U-shaped association between depressive symptoms and organizational religious activity, an inverse linear relationship of depressive symptoms with private religious involvement, and a positive relationship of depressive symptoms with intrinsic religiosity. Longitudinal analyses revealed a U-shaped association between depressive symptoms and private religious involvement, such that those reporting moderate levels of private religiosity at baseline evidenced lower levels of depressive symptoms at 1-year follow-up than those reporting either high or low levels of private religious activity.

CONCLUSIONS

The relationships between religious involvement and depression in primary care elders are complex and dependent on the type of religiosity measured. The authors found the strongest evidence for an association of non-organizational, private religious involvement and the severity of depressive symptoms, although further study is warranted using carefully controlled longitudinal designs that test for both linear and curvilinear relationships.