Random-effects regression analysis of correlated grouped-time survival data

Does government food demonstration intervention influence household dietary diversity in the Upper West Region of Ghana?

Dietary diversity is crucial in ensuring food and nutrition security. In low-middle-income countries, people frequently prioritize the quantity of food they consume over its quality due to a lack of availability and financial limitations. As a result, achieving dietary diversity is often overlooked in favor of ensuring adequate caloric intake. Through a social cognitive theory perspective, our study examines the relationship between food demonstrations and household dietary diversity in Ghana's Upper West Region utilizing cross-sectional survey data from 517 smallholder farmer households. The results from ordered logistic regression presented in odds ratio (OR) show that participating in food demonstrations (OR: 2.585, p<0.01), engaging in home gardening (OR: 1.932, p<0.001), having access to credit (OR: 1.609, p<0.01), self-rated good nutritional status (OR: 1.747, p<0.01), and Waala ethnicity (OR: 3.686, p<0.001) were all positively associated with high household dietary diversity. Conversely, living in the Wa West district was associated with lower dietary diversity (OR: 0.326, p<0.001). Our research findings suggest that policymakers may want to consider implementing community-based educational programs, such as home and school visits for food demonstrations and sensitizations, promoting mother-to-mother support groups for dietary diversity education, nutrition counseling services, and using role-play and local media. In addition, strengthening local agricultural policies through food banks, indigenous seed development, and mobile food markets and enhancing public-private partnerships like the Ghana Schools Feeding Programme and National Food Buffer Stock company could improve the supply chain and distribution networks for diverse food items. Implementing these interventions in the Upper West Region of Ghana could improve health, well-being, food security, and nutritional outcomes.

Alcohol misuse as a social determinant of food insecurity among smallholder farmers

The availability and affordability of alcohol in smallholder communities have surged the misuse of alcohol. Misusing alcohol has dire health and nutrition consequences in smallholder communities. Alcohol misuse can divert household resources from essential household needs such as food and also hinder local food production. In the context of multiple stressors on smallholder farmers' livelihoods, it is crucial to assess the relationship between alcohol consumption and smallholder farmers' experience of hunger. Therefore, we used data from a cross-sectional survey involving 1100 smallholder farmers in the Upper West region of Ghana to examine the association between alcohol consumption and household food insecurity. Results showed that daily (OR = 3.81; p ≤ 0.001) and weekly/frequent (OR = 2.32; p ≤ 0.001) consumption of alcohol was significantly associated with higher odds of household food insecurity compared to no consumption. The relationship between alcohol and food insecurity was bidirectional. The experience of food insecurity was also significantly associated with higher odds of occasional or frequent alcohol consumption. While alcohol misuse can transition smallholder households into food insecurity, the household heads of food insecure households may resort to alcohol to cope with underlying stressors such as climate change and food insecurity. This calls for policy interventions to mitigate alcohol misuse through regulations, surveillance, economic disincentives and improving the social mechanisms of resilience to climate change and food insecurity in smallholder communities. However, policy approaches must be cautious not to disrupt the livelihoods of vulnerable smallholder farmers.

Statistical analysis of the longitudinal fundamental movement skills data in the REACT project using the multilevel ordinal logistic model

OBJECTIVES

The REACT project was designed around two main aims: (1) to assess children's growth and motor development after the COVID-19 pandemic and (2) to follow their fundamental movement skills' developmental trajectories over 18 months using a novel technological device (Meu Educativo®) in their physical education classes. In this article, our goal is to describe statistical analysis of the longitudinal ordinal motor development data that was obtained from these children using the multilevel ordinal logistic model.

METHODS

Longitudinal ordinal data are often collected in studies on motor development. For example, children or adolescents might be rated as having poor, good, or excellent performance levels in fundamental movement skills, and such ratings may be obtained yearly over time to assess changes in fundamental movement skills levels of performance. However, such longitudinal ordinal data are often analyzed using either methods for continuous outcomes, or by dichotomizing the ordinal outcome and using methods for binary data. These approaches are not optimal, and so we describe in detail the use of the multilevel ordinal logistic model for analysis of such data from the REACT project. Our intent is to provide an accessible description and application of this model for analysis of ordinal motor development data.

DISCUSSION

Our analyses show both the between-subjects and within-subjects effects of age on motor development outcomes across three timepoints. The between-subjects effect of age indicate that children that are older have higher motor development ratings, relative to thoese that are younger, whereas the within-subject effect of age indicates higher motor development ratings as a child ages. It is the latter effect that is particularly of interest in longitudinal studies of motor development, and an important advantage of using the multilevel ordinal logistic model relative to more traditional methods.

Implications of analysing time-to-event outcomes as binary in meta-analysis: empirical evidence from the Cochrane Database of Systematic Reviews

Background

Systematic reviews and meta-analysis of time-to-event outcomes are frequently published within the Cochrane Database of Systematic Reviews (CDSR). However, these outcomes are handled differently across meta-analyses. They can be analysed on the hazard ratio (HR) scale or can be dichotomized and analysed as binary outcomes using effect measures such as odds ratios (OR) or risk ratios (RR). We investigated the impact of reanalysing meta-analyses from the CDSR that used these different effect measures.

Methods

We extracted two types of meta-analysis data from the CDSR: either recorded in a binary form only (“binary”), or in binary form together with observed minus expected and variance statistics (“OEV”). We explored how results for time-to-event outcomes originally analysed as “binary” change when analysed using the complementary log–log (clog-log) link on a HR scale. For the data originally analysed as HRs (“OEV”), we compared these results to analysing them as binary on a HR scale using the clog-log link or using a logit link on an OR scale.

Results

The pooled HR estimates were closer to 1 than the OR estimates in the majority of meta-analyses. Important differences in between-study heterogeneity between the HR and OR analyses were also observed. These changes led to discrepant conclusions between the OR and HR scales in some meta-analyses. Situations under which the clog-log link performed better than logit link and vice versa were apparent, indicating that the correct choice of the method does matter. Differences between scales arise mainly when event probability is high and may occur via differences in between-study heterogeneity or via increased within-study standard error in the OR relative to the HR analyses.

Conclusions

We identified that dichotomising time-to-event outcomes may be adequate for low event probabilities but not for high event probabilities. In meta-analyses where only binary data are available, the complementary log–log link may be a useful alternative when analysing time-to-event outcomes as binary, however the exact conditions need further exploration. These findings provide guidance on the appropriate methodology that should be used when conducting such meta-analyses.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12874-022-01541-9.

Design and study protocol for a cluster randomized trial of a multi-faceted implementation strategy to increase the uptake of the USPSTF hypertension screening recommendations: the EMBRACE study

Background

The US Preventive Services Task Force (USPSTF) recommends out-of-office blood pressure (BP) testing to exclude white coat hypertension prior to hypertension diagnosis. Despite improved availability and coverage of home and 24-h ambulatory BP monitoring (HBPM, ABPM), both are infrequently used to confirm diagnoses. We used the Behavior Change Wheel (BCW) framework, a multi-step process for mapping barriers to theory-informed behavior change techniques, to develop a multi-component implementation strategy for increasing out-of-office BP testing for hypertension diagnosis. Informed by geographically diverse provider focus groups (n = 63) exploring barriers to out-of-office testing and key informant interviews (n = 12), a multi-disciplinary team (medicine, psychology, nursing) used rigorous mixed methods to develop, refine, locally adapt, and finalize intervention components.

The purpose of this report is to describe the protocol of the Effects of a Multi-faceted intervention on Blood pRessure Actions in the primary Care Environment (EMBRACE) trial, a cluster randomized control trial evaluating whether a theory-informed multi-component strategy increased out-of-office testing for hypertension diagnosis.

Methods/design

The EMBRACE Trial patient sample will include all adults ≥ 18 years of age with a newly elevated office BP (≥ 140/90 mmHg) at a scheduled visit with a primary care provider from a study clinic. All providers with scheduled visits with adult primary care patients at enrolled ACN primary care clinics were included. We determined that the most feasible, effective implementation strategy would include delivering education about out-of-office testing, demonstration/instruction on how to perform out-of-office HBPM and ABPM testing, feedback on completion rates of out-of-office testing, environmental prompts/cues via computerized clinical decision support (CDS) tool, and a culturally tailored, locally accessible ABPM testing service. We are currently comparing the effect of this locally adapted multi-component strategy with usual care on the change in the proportion of eligible patients who complete out-of-office BP testing in a 1:1 cluster randomized trial across 8 socioeconomically diverse clinics.

Conclusions

The EMBRACE trial is the first trial to test an implementation strategy for improving out-of-office testing for hypertension diagnosis. It will elucidate the degree to which targeting provider behavior via education, reminders, and decision support in addition to providing an ABPM testing service will improve referral to and completion of ABPM and HBPMs.

Trial registration

Clinicaltrials.gov, NCT03480217, Registered on 29 March 2018

Hepatitis B knowledge among women and men in the upper west region of Ghana: What sources of health information matter?

Despite the vital role of accurate health information in reducing the spread of Hepatitis B virus (HBV) in endemic contexts such as Ghana, little is known about how health information sources may influence disparities in the knowledge of HBV transmission among women and men. This study examines the association between sources of health information and knowledge of HBV transmission in the Upper West Region (UWR) of Ghana. Data from a cross-sectional survey (n = 1061) was analyzed using gender-specific multivariate ordered logistic regression models. The results show that, women who obtained health information from religious-based programs (OR = 4.04, P < .05), higher-level facilities (OR = 2.37, P < .05), and primary health facilities (OR = 1.83, P < .1) were more likely to have good knowledge of HBV transmission compared to non-facility-based programs. Similarly, men who accessed health information from religious-based programs only, were more likely to have good knowledge of HBV transmission (OR = 2.14, P < .05) compared to non-facility-based programs. The results demonstrate the importance of health information sources on knowledge of disease transmission and prevention in a resource-poor context. Based on our findings, we suggest the scaling-up of information programs at health facilities in rural areas and the expansion of HBV services in the UWR in contribution towards the attainment of SDG #3.3.

Developmental considerations in survival models as applied to substance use research

Survival analysis is a class of models that are ideal for evaluating questions of timing of events, which makes them well-suited for modeling the development of a process such as initiation of substance use, development of addiction, or post-treatment recovery. The focus of this review paper is to demonstrate how survival models operate in a broader developmental framework and to offer guidance on selecting the appropriate model on the basis of the research question at hand. We provide a basic overview of survival models and then identify several key issues, explain how they pertain to research in the addiction field, and describe studies that utilize survival models to address questions about timing. We discuss the importance of carefully selecting the metric and origin of the time scale that corresponds to developmental process under investigation and we describe types of censoring/truncation. We describe the value of modeling covariates as time-invariant versus time-varying, and make the distinction between time-varying covariates and time-varying effects of covariates. We also explain how to test for substantive differences due to the timing of the assessment of the predictor. We finish the paper with a presentation of relatively novel extensions of survival models, including models that integrate standard statistical mediational analysis with discrete-time survival analysis, models that simultaneously consider order and timing of multiple events, and models that involve joint modeling of longitudinal and survival data. We also present our own substantive examples of various models in an Appendix containing annotated syntax and output.

A mixed-effects location scale model for time-to-event data: A smoking behavior application

In general, mixed-effects location scale models (MELS) allow assessment of within-person and between-person variability with time-to-event data for outcomes that follow a normal or ordinal distribution. In this article, we extend the mixed-effects location scale model to time-to-event data in relation to smoking data. Better understanding of the time-graded within-person variability of factors involved in nicotine dependence can be helpful to researchers in their efforts to fine-tune smoking cessation programs. We illustrate the MELS model with data on time to first cigarette measured every day for 7 days in smokers randomized to two groups: a) those asked to keep smoking, or b) those asked to stop. Our results show that some individuals remain very stable in their time to first cigarette over the week, while others show variable patterns. The stable individuals smoked every day, did not smoke immediately upon waking, and were all in the group asked to keep smoking. Conversely, the variable individuals had at least one day during which they did not smoke, other days during which they smoked within the first 5 min of waking, and they were almost all in the group asked to quit smoking. These findings suggested that MELS have the potential to provide insights on how people try to stop smoking. More importantly, this model can be applied to other clinically important outcomes such as time to relapse in a range of cessation programs.

Adapting the HIV Infant Tracking System to Support Prevention of Mother-to-Child Transmission of HIV in Kenya: Protocol for an Intervention Development Pilot Study in Two Hospitals

Background

Despite progress to expand access to HIV testing and treatment during pregnancy in Kenya, gaps still remain in prevention of mother-to-child transmission of HIV (PMTCT) services. This study addresses the need for effective and scalable interventions to support women throughout the continuum of care for PMTCT services in low-resource settings. Our research team has successfully implemented the HIV Infant Tracking System (HITSystem), a Web-based, system-level intervention to improve early infant diagnosis (EID) outcomes.

Objective

This study will expand the scope of the HITSystem to address PMTCT services to bridge the gap between maternal and pediatric HIV services and improve outcomes. This paper describes the intervention development protocol to adapt and pilot an HITSystem version 2.0 to assess acceptability, feasibility, and preliminary PMTCT outcomes in Kenya.

Methods

This is a 3-year intervention development study to adapt the current HITSystem intervention to support a range of PMTCT outcomes including appointment attendance, antiretroviral therapy (ART) adherence, hospital deliveries, and integration of maternal and pediatric HIV services in low-resource settings. The study will be conducted in 3 phases. Phase 1 will elicit feedback from intervention users (patients and providers) to guide development and refinement of the new PMTCT components and inform optimal implementation. In Phase 2, we will design and develop the HITSystem 2.0 features to support key PMTCT outcomes guided by clinical content experts and findings from Phase 1. Phase 3 will assess complete PMTCT retention (before, during, and after delivery) using a matched randomized pilot study design in 2 hospitals over 18 months. A total of N=108 HIV-positive pregnant women (n=54 per site) will be enrolled and followed from their first PMTCT appointment until infant HIV DNA Polymerase Chain Reaction testing at the target age of 6 weeks (<7 weeks) postnatal.

Results

Funding for this study was received in August 2015, enrollment in Phase 1 began in March 2016, and completion of data collection is expected by May 2019.

Conclusions

This protocol will extend, adapt, and pilot an HITSystem 2.0 version to improve attendance of PMTCT appointments, increase ART adherence and hospital-based deliveries, and prompt EID by 6 weeks postnatal. The HITSystem 2.0 aims to improve the integration of maternal and pediatric HIV services.

Trial Registration

ClinicalTrials.gov NCT02726607; https://clinicaltrials.gov/ct2/show/NCT02726607 (Archived by WebCite at http://www.webcitation.org/78VraLrOb)

International Registered Report Identifier (IRRID)

DERR1-10.2196/13268

A multivariate discrete failure time model for the analysis of infant motor development

We develop a multivariate discrete failure time model for the analysis of infant motor development. We use the model to jointly evaluate the time (in months) to achievement of three well-established motor milestones: sitting up, crawling, and walking. The model includes a subject-specific latent factor that reflects underlying heterogeneity in the population and accounts for within-subject dependence across the milestones. The factor loadings and covariate effects are allowed to vary flexibly across milestones, and the milestones are permitted to have unique at-risk intervals corresponding to different developmental windows. We adopt a Bayesian inferential approach and develop a convenient data-augmented Gibbs sampler for posterior computation. We conduct simulation studies to illustrate key features of the model and use the model to analyze data from the Nurture study, a birth cohort examining infant health and development during the first year of life.

Modeling clustered binary data with excess zero clusters

We establish a zero-inflated (random-effects) logistic-Gaussian model for clustered binary data in which members of clusters in one latent class have a zero response with probability one, and members of clusters in a second latent class yield correlated outcomes. Response probabilities in terms of random-effects models are formulated, and maximum marginal likelihood estimation procedures based on Gaussian quadrature are developed. Application to esophageal cancer data in Chinese families is presented.

ROC Estimation from Clustered Data with an Application to Liver Cancer Data

In this article, we propose a regression model to compare the performances of different diagnostic methods having clustered ordinal test outcomes. The proposed model treats ordinal test outcomes (an ordinal categorical variable) as grouped-survival time data and uses random effects to explain correlation among outcomes from the same cluster. To compare different diagnostic methods, we introduce a set of covariates indicating diagnostic methods and compare their coefficients. We find that the proposed model defines a Lehmann family and can also introduce a location-scale family of a receiver operating characteristic (ROC) curve. The proposed model can easily be estimated using standard statistical software such as SAS and SPSS. We illustrate its practical usefulness by applying it to testing different magnetic resonance imaging (MRI) methods to detect abnormal lesions in a liver.

Comparing vocational training courses through a discrete-time multilevel hazard model

In this article, the author provides an econometric approach for comparing vocational training courses, and aims to study how courses affect individual behaviour. Given the hierarchical nature of the data and the purposes of the analysis, he proposes the application of a discrete-time multilevel hazard model. The variable of interest is the duration (in months) of the first job-search after the end of the course. The author focusses on contextual and correlated (course) effects and comments on the results also in terms of school effectiveness. The approach may be easily applied to several other clustered structures.

Methods for Multilevel Ordinal Data in Prevention Research

This paper discusses statistical models for multilevel ordinal data that may be more appropriate for prevention outcomes than models that assume continuous measurement and normality. Prevention outcomes often have distributions that make them inappropriate for many popular statistical models that assume normality and are more appropriately considered ordinal outcomes. Despite this, the modeling of ordinal outcomes is often not well understood. This article discusses ways to analyze multilevel ordinal outcomes that are clustered or longitudinal, including the proportional odds regression model for ordinal outcomes, which assumes that the covariate effects are the same across the levels of the ordinal outcome. The article will cover how to test this assumption and what to do if it is violated. It will also discuss application of these models using computer software programs.

An introduction and integration of cross-classified, multiple membership, and dynamic group random-effects models

In longitudinal studies, time-varying group membership and group effects are important issues that need to be addressed. In this article we describe use of cross-classified and multiple membership random-effects models to address time-varying group membership, and dynamic group random-effects models to address time-varying group effects. We propose new models that integrate features of existing models, evaluate these models through simulation, provide guidance on how to fit these models, and apply the models in 2 real data examples. The discussion focuses on challenges in the application of these models.

Predictors of waterpipe smoking progression among youth in Irbid, Jordan: A longitudinal study (2008-2011)

BACKGROUND

The predictors of waterpipe smoking progression are yet to be examined using a longitudinal study that is guided by a theoretical model of behavioral change. This study identifies the gender-specific predictors of waterpipe smoking progression among adolescents in Irbid, Jordan.

METHODS

This study uses data from a school longitudinal study of smoking behavior in Irbid, Jordan. A random sample of 19 schools was selected by probability proportionate to size. A total of 1781 seventh graders were enrolled at baseline, and completed a questionnaire annually from 2008 through 2011. Students who reported ever smoking waterpipe (N=864) at any time point were assessed for progression (escalation in the frequency of waterpipe smoking) in the subsequent follow-up. Grouped-time survival analysis was used to identify the risk of progression.

RESULTS

During the three years of follow-up, 29.6% of students progressed in waterpipe smoking. Predictors of waterpipe smoking progression were higher mother's education, enrollment in public school, frequent physical activity, and low refusal self-efficacy among boys, having ever smoked cigarettes, and having friends and siblings who smoke waterpipe among girls. Awareness of harms of waterpipe was protective among boys and seeing warning labels on the tobacco packs was protective among girls.

CONCLUSIONS

Even at this early stage, about a third of waterpipe smokers progressed in their habit during the 3 year follow up. Factors predicting progression of use differed by gender, which calls for gender-specific approaches to waterpipe interventions among Jordanian youth.

Improving early infant HIV diagnosis in Kenya: study protocol of a cluster-randomized efficacy trial of the HITSystem

Background

Early infant diagnosis among human immunodeficiency virus (HIV)-exposed infants is a critical component of prevention of mother-to-child transmission programs. Barriers to early infant diagnosis include poor uptake, low retention at designated re-testing intervals, delayed test results, passive systems of communication, and poor linkage to treatment. This study will evaluate the HIV Infant Tracking System (HITSystem), an eHealth intervention that streamlines communication and accountability between the key early infant diagnosis stakeholders: HIV+ mothers and their HIV-exposed infants, healthcare providers, and central laboratory personnel. It is hypothesized that the HITSystem will significantly improve early infant diagnosis retention at 9 and 18 months postnatal and the timely provision of services.

Methods/design

Using a phased cluster-randomized controlled trial design, we will evaluate the impact of the HITSystem on eight primary benchmarks in the 18-month long cascade of care for early infant diagnosis. Study sites are six government hospitals in Kenya matched on geographic region, resource level, and patient volume. Early infant diagnosis outcomes of mother-infant dyads (n = 120 per site) at intervention hospitals (n = 3) where the HITSystem is deployed at baseline will be compared to the matched control sites providing standard care. After allowing for sufficient time for enrollment and 18-month follow-up of dyads, the HITSystem will be deployed at the control sites in the end of Year 3. Primary outcomes are retention among mother-infant dyads, initiation of antiretroviral therapy among HIV-infected infants, and the proportion of services delivered within the optimal time window indicated by national and study guidelines. Satisfaction interviews with participants and providers will inform intervention improvements. Cost-effectiveness analyses will be conducted to inform the sustainability of the HITSystem. Hypothesized outcomes include significantly higher retention throughout the 18-month early infant diagnosis process, significantly more services provided on-time at intervention sites, and a potential savings to the healthcare system.

Discussion

This study will evaluate the public health impact of the HITSystem to improve critical early infant diagnosis outcomes in low-resource settings. Cost-effectiveness analyses will inform the feasibility of scale-up in other settings.

Trial registration

ClinicalTrials.gov: NCT02072603

Predictors of Cigarette Smoking Progression Among a School-Based Sample of Adolescents in Irbid, Jordan: A Longitudinal Study (2008-2011)

INTRODUCTION

Little evidence regarding longitudinal predictors of cigarette smoking progression is available from developing countries. This study aimed to identify gender-specific individual and social predictors of cigarette smoking progression among a school-based sample of adolescents in Irbid, Jordan.

METHODS

A total of 1781 seventh graders (participation rate 95%) were enrolled and completed an annual self-administered questionnaire from 2008 through 2011. Students who reported "ever-smoking a cigarette" at baseline or in the subsequent follow-up but not being "heavy daily smokers" (>10 cigarettes per day) were eligible for this analysis (N = 669). Grouped-time survival analyses were used to identify predictors of cigarette smoking progression in boys and girls.

RESULTS

Among the study sample, 38.3% of students increased the frequency and /or amount of cigarette smoking during the 3 years of follow-up. Among individual factors, the urge to smoke in the morning predicted smoking progression for boys and girls. The independent predictors of cigarette smoking progression were friends' smoking and attending public schools in boys, and siblings' smoking in girls. Discussing the dangers of smoking with family members was protective for girls.

CONCLUSION

Boys and girls progressed similarly in cigarette smoking once they initiated the habit. Progression among girls was solely family-related, while it was peer-related for boys.

Waterpipe a gateway to cigarette smoking initiation among adolescents in Irbid, Jordan: a longitudinal study

SETTING

According to anecdotal evidence, waterpipe smoking may lead to the initiation of cigarette smoking among young people. This hypothesis is yet to be examined using an appropriate study design and a theoretical model for behavioral change.

OBJECTIVE

To compare the risk of cigarette smoking initiation among waterpipe-only smokers and never smokers in a school-based sample of adolescents from Irbid, Jordan.

METHODS

A total of 1454 cigarette-naïve participants were drawn from a longitudinal study on smoking behavior conducted in Irbid among 1781 seventh graders who were enrolled at baseline (2008) and completed the study questionnaire on smoking behavior annually until 2011. Grouped time-survival analysis was used to compare the risk of subsequent initiation of cigarette smoking between waterpipe smokers (n = 298) and never smokers (n = 1156) using adjusted hazard ratios (aHRs) and 95% confidence intervals (95%CI).

RESULTS

Risk of initiation of cigarette smoking among waterpipe smokers was significantly higher than among never smokers after adjusting for potential confounders (aHR 1.67, 95%CI 1.46-1.92). The association between waterpipe and cigarette smoking initiation was dose-dependent. The risk of initiating cigarette smoking increased with increase in the frequency of waterpipe smoking (P for linear trend < 0.001).

CONCLUSIONS

Waterpipe smoking led to the initiation of cigarette smoking among this cohort of Jordanian adolescents; the effect was dose-dependent.

Risk of suicidal behavior with antidepressants in bipolar and unipolar disorders

OBJECTIVE

To examine the risk of suicidal behavior (suicide attempts and deaths) associated with antidepressants in participants with bipolar I, bipolar II, and unipolar major depressive disorders.

DESIGN

A 27-year longitudinal (1981-2008) observational study of mood disorders (Research Diagnostic Criteria diagnoses based on Schedule for Affective Disorders and Schizophrenia and review of medical records) was used to evaluate antidepressants and risk for suicidal behavior. Mixed-effects logistic regression models examined propensity for antidepressant exposure. Mixed-effects survival models that were matched on the propensity score examined exposure status as a risk factor for time until suicidal behavior.

SETTING

Five US academic medical centers.

RESULTS

Analyses of 206 participants with bipolar I disorder revealed 2,010 exposure intervals (980 exposed to antidepressants; 1,030 unexposed); 139 participants with bipolar II disorder had 1,407 exposure intervals (694 exposed; 713 unexposed); and 361 participants with unipolar depressive disorder had 2,745 exposure intervals (1,328 exposed; 1,417 unexposed). Propensity score analyses confirmed that more severely ill participants were more likely to initiate antidepressant treatment. In mixed-effects survival analyses, those with bipolar I disorder had a significant reduction in risk of suicidal behavior by 54% (HR = 0.46; 95% CI, 0.31-0.69; t = -3.74; P < .001) during periods of antidepressant exposure compared to propensity-matched unexposed intervals. Similarly, the risk was reduced by 35% (HR = 0.65; 95% CI, 0.43-0.99; t = -2.01; P = .045) in bipolar II disorder. By contrast, there was no evidence of an increased or decreased risk with antidepressant exposure in unipolar disorder.

CONCLUSIONS

Based on observational data adjusted for propensity to receive antidepressants, antidepressants may protect patients with bipolar disorders but not unipolar depressive disorder from suicidal behavior.

Recovery from multiple episodes of bipolar I depression

OBJECTIVE

To describe the duration of bipolar I major and minor depressive episodes and factors associated with time to recovery.

METHOD

As part of the National Institute of Mental Health Collaborative Depression Study, 219 participants with bipolar I disorder based on Research Diagnostic Criteria analogs to DSM-IV-TR criteria were recruited at 5 academic medical centers from 1978 to 1981 and followed for up to 25 years with the Longitudinal Interval Follow-Up Evaluation. The probability of recovery over time from depressive episodes, the primary outcome measure, was examined with mixed-effects grouped-time survival models.

RESULTS

The median duration of major depressive episodes was 14 weeks, and over 70% of participants recovered within 12 months of episode onset. The median duration of minor depressive episodes was 8 weeks, and approximately 90% of participants recovered within 6 months of onset of the episode. Aggregated data demonstrated similar durations of the first 3 major depressive episodes. However, for each participant with multiple episodes of major depression or minor depression, the duration of each episode was not consistent (intraclass correlation coefficient = 0.07 and 0.25 for major and minor depression, respectively). The total number of years in episode over follow-up with major plus minor depression prior to onset of a major depressive episode was significantly associated with a decreased probability of recovery from that episode; with each additional year, the likelihood of recovery was reduced by 7% (hazard ratio = 0.93; 95% CI, 0.89-0.98; P = .002).

CONCLUSIONS

Bipolar I major depression generally lasts longer than minor depression, and the duration of multiple episodes within an individual varies. However, the probability of recovery over time from an episode of major depression appears to decline with each successive episode.

Partnerships for the design, conduct, and analysis of effectiveness, and implementation research: experiences of the prevention science and methodology group

What progress prevention research has made comes through strategic partnerships with communities and institutions that host this research, as well as professional and practice networks that facilitate the diffusion of knowledge about prevention. We discuss partnership issues related to the design, analysis, and implementation of prevention research and especially how rigorous designs, including random assignment, get resolved through a partnership between community stakeholders, institutions, and researchers. These partnerships shape not only study design, but they determine the data that can be collected and how results and new methods are disseminated. We also examine a second type of partnership to improve the implementation of effective prevention programs into practice. We draw on social networks to studying partnership formation and function. The experience of the Prevention Science and Methodology Group, which itself is a networked partnership between scientists and methodologists, is highlighted.

The effect of recurrent episodes of clinical mastitis caused by gram-positive and gram-negative bacteria and other organisms on mortality and culling in Holstein dairy cows

The objective of this study was to estimate the effects of recurrent episodes of different types of clinical mastitis (CM) caused by gram-positive (Streptococcus spp., Staphylococcus aureus, Staphylococcus spp.) and gram-negative (Escherichia coli, Klebsiella, Citrobacter, Enterobacter, Pseudomonas) bacteria, and other organisms (Arcanobacterium pyogenes, Mycoplasma, Corynebacterium bovis, yeast, miscellaneous) on the probability of mortality and culling in Holstein dairy cows. Data from 30,233 lactations in cows of 7 dairy farms in New York State were analyzed. Cows were followed for the first 10 mo in lactation, or until death or culling occurred, or until the end of our study period. Generalized linear mixed models with a Poisson error distribution were used to study the effects of recurrent cases of the different types of CM and several other factors (herd, parity, month of lactation, current year and season, profitability, net replacement cost, other diseases) on cows' probability of death (model 1) or being culled (model 2). Primiparous and multiparous cows were modeled separately because they had different risks of mortality and culling and potentially different CM effects on mortality and culling. Approximately 30% of multiparous cows had at least one case of CM in lactation compared with 16.6% of primiparous cows. Multipara also had higher lactational incidence risks of second (10.7%) and third (4.4%) cases than primipara (3.7% and 1.1%, respectively). For primipara, CM increased the probability of death, with each successive case occurring in a month being increasingly lethal. In multipara, gram-negative CM increased the probability of death, especially when the gram-negative case was the first or second CM case in lactation. Primiparous cows with CM were more likely to be culled after CM than if they did not have CM, particularly after a second or third case. In multipara, any type of CM increased the probability of being culled. Gram-negative CM cases were associated with the numerically highest risk of culling.

Antiepileptic drugs for bipolar disorder and the risk of suicidal behavior: a 30-year observational study

OBJECTIVE

In 2009 the U.S. Food and Drug Administration issued a warning regarding suicidality and antiepileptic drugs based on meta-analyses of 199 randomized trials (over 43,000 subjects with different illnesses) of 11 antiepileptics. The present study examines the hypothesis that the three antiepileptics approved for bipolar disorder (carbamazepine, lamotrigine, and valproate) are associated with an elevated risk of suicide attempts and suicides.

METHOD

A prospective observational study was conducted at five U.S. academic medical centers from 1978 to 2009. Analyses included 199 participants with bipolar disorder for whom 1,077 time intervals were classified as either exposed to an antiepileptic (carbamazepine, lamotrigine, or valproate) or not exposed to an antiepileptic, an antidepressant, or lithium during 30 years of follow-up.

RESULTS

Participants who had more severe manic symptoms were more likely to receive antiepileptic drugs. Mixed-effects grouped-time survival models revealed no elevation in risk of suicide attempt or suicide during periods when participants were receiving antiepileptics relative to periods when they were not (hazard ratio=0.93, 95% CI=0.45-1.92), controlling for demographic and clinical variables through propensity score matching.

CONCLUSIONS

In this longitudinal observational study, the risk of suicide attempts or suicides was not associated with the antiepileptics approved for bipolar disorder.

Evaluation of psychiatric interventions in an observational study: issues in design and analysis

Characteristics of randomized controlled clinical trials (RCTs) and observational studies of psychiatric intervention effectiveness are contrasted. Randomization drives treatment assignment in an RCT, whereas clinician and patient selection determine treatment in an observational study. Strengths and weaknesses of randomized and observational designs are considered. The propensity adjustment, a statistical approach that allows for intervention evaluation in a nonrandomized observational study, is described here. The plausibility of propensity adjustment assumptions must be carefully evaluated. This data analytic strategy is illustrated with the longitudinal observational data from the National Institute of Mental Health Collaborative Depression Study. Evaluations presented here examine acute and maintenance antidepressant effectiveness and demonstrate effectiveness of the higher categorical doses.

Antidepressants and risks of suicide and suicide attempts: a 27-year observational study

OBJECTIVE

The 2007 revision of the black box warning for suicidality with antidepressants states that patients of all ages who initiate antidepressants should be monitored for clinical worsening or suicidality. The objective of this study was to examine the association of antidepressants with suicide attempts and with suicide deaths.

METHOD

A longitudinal, observational study of mood disorders with prospective assessments for up to 27 years was conducted at 5 US academic medical centers. The study sample included 757 participants who enrolled from 1979 to 1981 during an episode of mania, depression, or schizoaffective disorder, each based on Research Diagnostic Criteria. Unlike randomized controlled clinical trials of antidepressants, the analyses included participants with psychiatric and other medical comorbidity and those receiving acute or maintenance therapy, polypharmacy, or no psychopharmacologic treatment at all. Over follow-up, these participants had 6,716 time periods that were classified as either exposed to an antidepressant or not exposed. Propensity score-adjusted mixed-effects survival analyses were used to examine risk of suicide attempt or suicide, the primary outcome.

RESULTS

The propensity model showed that antidepressant therapy was significantly more likely when participants' symptom severity was greater (odds ratio [OR] = 1.16; 95% CI, 1.12-1.21; z = 8.22; P < .001) or when it was worsening (OR = 1.69; 95% CI, 1.50-1.89; z = 9.02; P < .001). Quintile-stratified, propensity-adjusted safety analyses using mixed-effects grouped-time survival models indicate that the risk of suicide attempts or suicides was reduced by 20% among participants taking antidepressants (hazard ratio, 0.80; 95% CI, 0.68-0.95; z = -2.54; P = .011).

CONCLUSIONS

This longitudinal study of a broadly generalizable cohort found that, although those with more severe affective syndromes were more likely to initiate treatment, antidepressants were associated with a significant reduction in the risk of suicidal behavior. Nonetheless, we believe that clinicians must closely monitor patients when an antidepressant is initiated.

Longitudinal course of bipolar I disorder: duration of mood episodes

CONTEXT

The phenomenology of bipolar I disorder affects treatment and prognosis.

OBJECTIVE

To describe the duration of bipolar I mood episodes and factors associated with recovery from these episodes.

DESIGN

Subjects with Research Diagnostic Criteria bipolar I disorder were prospectively followed up for as long as 25 years. The probability of recovery over time from multiple successive mood episodes was examined with survival analytic techniques, including a mixed-effects grouped-time survival model.

SETTING

Five US academic medical centers.

PARTICIPANTS

Two hundred nineteen subjects with bipolar I disorder.

MAIN OUTCOME MEASURES

Level of psychopathology was assessed with the Longitudinal Interval Follow-up Evaluation every 6 months for the first 5 years of follow-up and annually thereafter.

RESULTS

The median duration of bipolar I mood episodes was 13 weeks. More than 75% of the subjects recovered from their mood episodes within 1 year of onset. The probability of recovery was significantly less for an episode with severe onset (psychosis or severe psychosocial impairment in week 1 of the episode) (hazard ratio [HR] = 0.746; 95% confidence interval [CI], 0.578-0.963; P = .02) and for subjects with greater cumulative morbidity (total number of years spent ill with any mood episode) (HR = 0.917; 95% CI, 0.886-0.948; P < .001). Compared with the probability of recovery from a major depressive episode, there was a significantly greater probability of recovery from an episode of mania (HR = 1.713; 95% CI, 1.373-2.137; P < .001), hypomania (HR = 4.502; 95% CI, 3.466-5.849; P < .001), or minor depression (HR = 2.027; 95% CI, 1.622-2.534; P < .001) and, conversely, a significantly reduced probability of recovery from a cycling episode (switching from one pole to the other without an intervening period of recovery) (HR = 0.438; 95% CI, 0.351-0.548; P < .001).

CONCLUSIONS

The median duration of bipolar I mood episodes was 13 weeks, and the probability of recovery was significantly decreased for cycling episodes, mood episodes with severe onset, and subjects with greater cumulative morbidity.

Design and analysis of the Community Youth Development Study longitudinal cohort sample

Communities That Care (CTC) is a prevention system designed to reduce adolescent substance use and delinquency through the selection of effective preventive interventions tailored to a community's specific profile of risk and protection. A community-randomized trial of CTC, the Community Youth Development Study, is currently being conducted in 24 communities across the United States. This article describes the rationale, multilevel analyses, and baseline comparability for the study's longitudinal cohort design. The cohort sample consists of 4,407 fifth- and sixth-grade students recruited in 2004 and 2005 and surveyed annually through ninth grade. Results of mixed-model ANOVAs indicated that students in CTC and control communities exhibited no significant differences (ps > .05) in baseline levels of student outcomes.

Discrete-Time Survival Factor Mixture Analysis for Low-Frequency Recurrent Event Histories

In this article, the latent class analysis framework for modeling single event discrete-time survival data is extended to low-frequency recurrent event histories. A partial gap time model, parameterized as a restricted factor mixture model, is presented and illustrated using juvenile offending data. This model accommodates event-specific baseline hazard probabilities and covariate effects; event recurrences within a single time period; and accounts for within- and between-subject correlations of event times. This approach expands the family of latent variable survival models in a way that allows researchers to explicitly address questions about unobserved heterogeneity in the timing of events across the lifespan.

Do risk factors for suicidal behavior differ by affective disorder polarity?

BACKGROUND

Suicide is a leading cause of death and has been strongly associated with affective disorders. The influence of affective disorder polarity on subsequent suicide attempts or completions and any differential effect of suicide risk factors by polarity were assessed in a prospective cohort.

METHOD

Participants with major affective disorders in the National Institute of Mental Health (NIMH) Collaborative Depression Study (CDS) were followed prospectively for up to 25 years. A total of 909 participants meeting prospective diagnostic criteria for major depressive and bipolar disorders were followed through 4204 mood cycles. Suicidal behavior was defined as suicide attempts or completions. Mixed-effects, grouped-time survival analysis assessed risk of suicidal behavior and differential effects of risk factors for suicidal behavior by polarity. In addition to polarity, the main effects of age, gender, hopelessness, married status, prior suicide attempts and active substance abuse were modeled, with mood cycle as the unit of analysis.

RESULTS

After controlling for age of onset, there were no differences in prior suicide attempts by polarity although bipolar participants had more prior severe attempts. During follow-up, 40 cycles ended in suicide and 384 cycles contained at least one suicide attempt. Age, hopelessness and active substance abuse but not polarity predicted suicidal behavior. The effects of risk factors did not differ by polarity.

CONCLUSIONS

Bipolarity does not independently influence risk of suicidal behavior or alter the influence of well-established suicide risk factors within affective disorders. Suicide risk assessment strategies may continue to appraise these common risk factors without regard to mood polarity.

Widowhood and mortality among the elderly: the modifying role of neighborhood concentration of widowed individuals

The effect of death of a spouse on the mortality of the survivor (the "widowhood effect") is well-established. We investigated how the effect of widowhood on mortality depends on the neighborhood concentration of widowed individuals in the United States. We developed a large, nationally representative, and longitudinal dataset from Medicare claims and other data sources characterizing 200,000 elderly couples, with nine years of follow-up (1993-2002), and estimated multilevel grouped discrete-time hazard models. In neighborhoods with a low concentration of widowed individuals, widowhood increased the odds of death for men by 22% and for women by 17%, compared to 17% for men, and 15% for women in neighborhoods with a high concentration of widowed individuals. Our findings suggest that neighborhood structural contexts - that provide opportunities for interacting with others and favoring new social engagements - could be potential modifiers of the widowhood effects and as such requires more systematic consideration in future research of widowhood effects on well-being and mortality.

Shared random effects analysis of multi-state Markov models: application to a longitudinal study of transitions to dementia

Multi-state models are appealing tools for analysing data about the progression of a disease over time. In this paper, we consider a multi-state Markov chain with two competing absorbing states: dementia and death and three transient non-demented states: cognitively normal, amnestic mild cognitive impairment (amnestic MCI), and non-amnestic mild cognitive impairment (non-amnestic MCI). The likelihood function for the data is derived and estimates for the effects of the covariates on transitions are determined when the process can be viewed as a polytomous logistic regression model with shared random effects. The presence of a shared random effect not only complicates the formulation of the likelihood but also its evaluation and maximization. Three approaches for maximizing the likelihood are compared using a simulation study; the first method is based on the Gauss-quadrature technique, the second method is based on importance sampling ideas, and the third method is based on an expansion by Taylor series. The best approach is illustrated using a longitudinal study on a cohort of cognitively normal subjects, followed annually for conversion to mild cognitive impairment (MCI) and/or dementia, conducted at the Sanders Brown Center on Aging at the University of Kentucky.

Bias reduction in effectiveness analyses of longitudinal ordinal doses with a mixed-effects propensity adjustment

A mixed-effects propensity adjustment is described that can reduce bias in longitudinal studies involving non-equivalent comparison groups. There are two stages in this data analytic strategy. First, a model of propensity for treatment intensity examines variables that distinguish among subjects who receive various ordered doses of treatment across time using mixed-effects ordinal logistic regression. Second, the effectiveness model examines multiple times until recurrence to compare the ordered doses using a mixed-effects grouped-time survival model. Effectiveness analyses are initially stratified by propensity quintile. Then the quintile-specific results are pooled, assuming that there is not a propensity x treatment interaction. A Monte Carlo simulation study compares bias reduction in fully specified propensity model relative to misspecified models. In addition, type I error rate and statistical power are examined. The approach is illustrated by applying it to a longitudinal, observational study of maintenance treatment of major depression.

A flexible model for multivariate interval-censored survival times with complex correlation structure

We focus on the analysis of multivariate survival times with highly structured interdependency and subject to interval censoring. Such data are common in developmental genetics and genetic epidemiology. We propose a flexible mixed probit model that deals naturally with complex but uninformative censoring. The recorded ages of onset are treated as possibly censored ordinal outcomes with the interval censoring mechanism seen as arising from a coarsened measurement of a continuous variable observed as falling between subject-specific thresholds. This bypasses the requirement for the failure times to be observed as falling into non-overlapping intervals. The assumption of a normal age-of-onset distribution of the standard probit model is relaxed by embedding within it a multivariate Box-Cox transformation whose parameters are jointly estimated with the other parameters of the model. Complex decompositions of the underlying multivariate normal covariance matrix of the transformed ages of onset become possible. The new methodology is here applied to a multivariate study of the ages of first use of tobacco and first consumption of alcohol without parental permission in twins. The proposed model allows estimation of the genetic and environmental effects that are shared by both of these risk behaviours as well as those that are specific.

A mixed-effects quintile-stratified propensity adjustment for effectiveness analyses of ordered categorical doses

Observational studies can be used to evaluate treatment effectiveness among patients with a broader range of illness severity than typically seen in randomized controlled clinical trials. However, there are several difficulties with observational evaluations including non-equivalent comparison groups, treatment doses and durations that vary widely, and, in longitudinal studies, multiple courses of treatment per subject. A mixed-effects approach to the propensity adjustment for non-equivalent comparison groups is described that can account for each of these perturbations. The strategy involves two stages. First, characteristics that distinguish among subjects who receive various levels of treatment are examined in a model of propensity for treatment intensity using mixed-effects ordinal logistic regression. Second, the propensity-stratified effectiveness of ordered categorical doses is compared in a mixed-effects grouped time survival model of time until recovery. The model is applied in a longitudinal, observational study of antidepressant effectiveness. Then a Monte Carlo simulation study indicates that the strategy has acceptable type I error rates and minimal bias in the estimates of treatment effectiveness. Statistical power exceeds 0.90 for an odds ratio of 1.5 with N = 250 and 500, and is acceptable for an odds ratio of 2.0 with N = 100. Nevertheless, with N = 100, the models that had high intraclass correlation coefficients had greater tendency towards non-convergence. This approach is a useful strategy for observational studies of treatment effectiveness. It is capable of adjusting for selection bias, incorporating multiple observations per subject, and comparing effectiveness of ordinal doses.

Context or composition: what explains variation in SCHIP disenrollment?

OBJECTIVE

To investigate (1) the relative contributions of family and contextual characteristics to observed variation in disenrollment rates from the State Children's Health Insurance Program (SCHIP), and (2) whether context explains observed family-level patterns.

DATA SOURCES

We use secondary data on 24,628 families enrolled in New Jersey's SCHIP program (NJ KidCare), and county-level data from the Area Resource File, the Census, and the NJ FamilyCare provider roster.

STUDY DESIGN

Information on family characteristics, SCHIP plan, and dates of enrollment and disenrollment are taken from NJ KidCare administrative records, which provided surveillance data from January 1998 through April 2000.

DATA COLLECTION/ANALYSIS

We estimate a multilevel discrete-time-hazards model of SCHIP disenrollment.

FINDINGS

Families enrolled in plans involving cost-sharing, blacks, and those with only one enrolled child have higher than average rates of disenrollment. Disenrollment rates for blacks are lower in counties with a high share of black physicians. These characteristics account for part of the intercounty variation in disenrollment rates; remaining intercounty variation is largely explained by physician density or population density.

POLICY IMPLICATIONS

It may be worthwhile to pay special attention to black families and counties with high disenrollment rates to address the reasons for their lower retention. Addressing cultural differences between physician and client and the geographic distribution of medical providers might reduce disenrollment.

Design and analysis of group-randomized trials: a review of recent methodological developments

We review recent developments in the design and analysis of group-randomized trials (GRTs). Regarding design, we summarize developments in estimates of intraclass correlation, power analysis, matched designs, designs involving one group per condition, and designs in which individuals are randomized to receive treatments in groups. Regarding analysis, we summarize developments in marginal and conditional models, the sandwich estimator, model-based estimators, binary data, survival analysis, randomization tests, survey methods, latent variable methods and nonlinear mixed models, time series methods, global tests for multiple endpoints, mediation effects, missing data, trial reporting, and software. We encourage investigators who conduct GRTs to become familiar with these developments and to collaborate with methodologists who can strengthen the design and analysis of their trials.

Assessing the effects of a school-based intervention on unscheduled school transfers during elementary school

Raising Healthy Children is a cluster-randomized study of a school-based intervention aimed at preventing problem behaviors among children recruited into the project in the first or second grade of elementary school. Multilevel analysis was used to compare students in intervention and control schools with respect to whether they transferred out of their original schools. Students in intervention schools were less likely to transfer within the first 5 years of the project. A multilevel discrete-time survival model that included both time-varying and contextual variables revealed that the difference in hazard of transfer was greatest in the earlier years of the project.

Multilevel models for censored and latent responses

Multilevel models were originally developed to allow linear regression or ANOVA models to be applied to observations that are not mutually independent. This lack of independence commonly arises due to clustering of the units of observations into 'higher level units' such as patients in hospitals. In linear mixed models, the within-cluster correlations are modelled by including random effects in a linear model. In this paper, we discuss generalizations of linear mixed models suitable for responses subject to systematic and random measurement error and interval censoring. The first example uses data from two cross-sectional surveys of schoolchildren to investigate risk factors for early first experimentation with cigarettes. Here the recalled times of the children's first cigarette are likely to be subject to both systematic and random measurement errors as well as being interval censored. We describe multilevel models for interval censored survival times as special cases of generalized linear mixed models and discuss methods of estimating systematic recall bias. The second example is a longitudinal study of mental health problems of patients nested in clinics. Here the outcome is measured by multiple questionnaires allowing the measurement errors to be modelled within a linear latent growth curve model. The resulting model is a multilevel structural equation model. We briefly discuss such models both as extensions of linear mixed models and as extensions of structural equation models. Several different model structures are examined. An important goal of the paper is to place a number of methods that readers may have considered as being distinct within a single overall modelling framework.