The Impact of Prenatal Alcohol Exposure on Longitudinal Growth, Nutritional Status, and Insulin-Like Growth Factor 1 in Early Childhood in Leyte, the Philippines.

OBJECTIVE: To assess the impact and potential mechanistic pathways of prenatal alcohol exposure (PAE) on longitudinal growth and nutritional status in early childhood. STUDY DESIGN: A cohort of 296 mother-infant dyads (32% with PAE vs 68% unexposed) were recruited in Leyte, the Philippines, and followed from early gestation through 24 months of age. PAE was assessed using serum phosphatidylethanol (PEth) captured twice prenatally and in cord blood and supplemented with self-reported alcohol consumption. Linear mixed models were used to examine longitudinal effects of PAE on growth from birth through 2 years including key potential mediating factors (placental histopathology, and infant serum leptin and Insulin-like Growth Factor 1 [IGF-1]). RESULTS: After adjusting for potential confounders, we found that PAE was significantly associated with a delayed blunting of linear growth trajectories (height-for-age z-score, body length) and weight (weight-for-age z-score, body weight) that manifested between 4 and 6 months and continued through 12-24 months. PAE was also associated with a decreased rate of mid-upper-arm circumference growth from birth to 12 months, and a lower mean IGF-1 levels at birth and 6 months. CONCLUSION: This study demonstrates a delayed impact of PAE on growth that manifested around 6 months of age, underscoring the importance of routine clinical monitoring in early childhood. Furthermore, the findings supported prior animal model findings that suggest a mechanistic role for IGF-1 in PAE-induced growth delay.

Gaussian process emulation to improve efficiency of computationally intensive multidisease models: a practical tutorial with adaptable R code.

BACKGROUND: The rapidly growing burden of non-communicable diseases (NCDs) among people living with HIV in sub-Saharan Africa (SSA) has expanded the number of multidisease models predicting future care needs and health system priorities. Usefulness of these models depends on their ability to replicate real-life data and be readily understood and applied by public health decision-makers; yet existing simulation models of HIV comorbidities are computationally expensive and require large numbers of parameters and long run times, which hinders their utility in resource-constrained settings. METHODS: We present a novel, user-friendly emulator that can efficiently approximate complex simulators of long-term HIV and NCD outcomes in Africa. We describe how to implement the emulator via a tutorial based on publicly available data from Kenya. Emulator parameters relating to incidence and prevalence of HIV, hypertension and depression were derived from our own agent-based simulation model and other published literature. Gaussian processes were used to fit the emulator to simulator estimates, assuming presence of noise for design points. Bayesian posterior predictive checks and leave-one-out cross validation confirmed the emulator's descriptive accuracy. RESULTS: In this example, our emulator resulted in a 13-fold (95% Confidence Interval (CI): 8-22) improvement in computing time compared to that of more complex chronic disease simulation models. One emulator run took 3.00 seconds (95% CI: 1.65-5.28) on a 64-bit operating system laptop with 8.00 gigabytes (GB) of Random Access Memory (RAM), compared to > 11 hours for 1000 simulator runs on a high-performance computing cluster with 1500 GBs of RAM. Pareto k estimates were < 0.70 for all emulations, which demonstrates sufficient predictive accuracy of the emulator. CONCLUSIONS: The emulator presented in this tutorial offers a practical and flexible modelling tool that can help inform health policy-making in countries with a generalized HIV epidemic and growing NCD burden. Future emulator applications could be used to forecast the changing burden of HIV, hypertension and depression over an extended (> 10 year) period, estimate longer-term prevalence of other co-occurring conditions (e.g., postpartum depression among women living with HIV), and project the impact of nationally-prioritized interventions such as national health insurance schemes and differentiated care models.

Economic Evaluation of Extended-Release Buprenorphine for Persons With Opioid Use Disorder.

Importance: In 2017, the US Food and Drug Administration (FDA) approved a monthly injectable form of buprenorphine, extended-release buprenorphine; published data show that extended-release buprenorphine is effective compared with no treatment, but its current cost is higher and current retention is lower than that of transmucosal buprenorphine. Preliminary research suggests that extended-release buprenorphine may be an important addition to treatment options, but the cost-effectiveness of extended-release buprenorphine compared with transmucosal buprenorphine remains unclear. Objective: To evaluate the cost-effectiveness of extended-release buprenorphine compared with transmucosal buprenorphine. Design, Setting, and Participants: This economic evaluation used a state transition model starting in 2019 to simulate the lifetime of a closed cohort of individuals with OUD presenting for evaluation for opioid agonist treatment with buprenorphine. The data sources used to estimate model parameters included cohort studies, clinical trials, and administrative data. The model relied on pharmaceutical costs from the Federal Supply Schedule and health care utilization costs from published studies. Data were analyzed from September 2021 to January 2023. Interventions: No treatment, treatment with transmucosal buprenorphine, or treatment with extended-release buprenorphine. Main Outcomes and Measures: Mean lifetime costs per person, discounted quality-adjusted life-years (QALYs), and incremental cost-effectiveness ratios (ICERs). Results: The simulated cohort included 100 000 patients with OUD receiving (61% male; mean [SD] age, 38 [11] years) or not receiving medication treatment (58% male, mean [SD] age, 48 [18] years). Compared with no medication treatment, treatment with transmucosal buprenorphine yielded an ICER of $19 740 per QALY. Compared with treatment with transmucosal buprenorphine, treatment with extended-release buprenorphine yielded lower effectiveness by 0.03 QALYs per person at higher cost, suggesting that treatment with extended-release buprenorphine was dominated and not preferred. In probabilistic sensitivity analyses, treatment with transmucosal buprenorphine was the preferred strategy 60% of the time. Treatment with extended-release buprenorphine was cost-effective compared with treatment with transmucosal buprenorphine at a $100 000 per QALY willingness-to-pay threshold only after substantial changes in key parameters. Conclusions and Relevance: In this economic evaluation of extended-release buprenorphine compared with transmucosal buprenorphine for the treatment of OUD, extended-release buprenorphine was not associated with efficient allocation of limited resources when transmucosal buprenorphine was available. Future initiatives should aim to improve retention rates or decrease costs associated with extended-release buprenorphine.

Predictors and impact of survivorship care plans and survivorship care visits.

PURPOSE: We performed this study to characterize the population at the Lifespan Cancer Institute (LCI) who received a survivorship care plan (SCP) with or without a survivorship care visit (SCV) to determine both the impact on specialty referrals and the demographic and clinical predictors of SCPs and SCVs. METHODS: We retrospectively reviewed EMR records on 1960 patients at LCI between 2014 and 2017 for SCPs and SCVs and extracted demographics, distress thermometer (DT) scores collected at the time of initial presentation, and subsequent referrals. We evaluated the bivariate associations of SCP and SCV with continuous and categorical factors and assessed the adjusted effect of these factors on receipt of SCP and SCV independently. All analyses were performed in R v4.0.2. RESULTS: SCPs were completed in 740 (37.8%) patients, and of those, 65.9% had a SCV. The mean age was 63.9, 67% were female, and 51.2% were married or partnered. Patients treated for breast, lung, and prostate cancers most received an SCP. Compared to SCP alone, the SCV was associated with more specialty referrals. Those who were younger and had breast cancer were more likely to receive a SCP, and those who were younger and female and had breast cancer were more likely to receive a SCV. CONCLUSIONS: Gender, age, and type of cancer are significant predictors of receipt of SCP and SCV. Patients who received either SCP, SCV, or both were more likely to receive specialty referrals than those who received neither. IMPLICATIONS FOR CANCER SURVIVORS: Identifying predictive factors of SCP and SCV can help facilitate earlier receipt of specialty services and specialty referrals as needed.

Community Mitigation Strategies, Mobility, and COVID-19 Incidence Across Three Waves in the United States in 2020.

BACKGROUND: Summarizing the impact of community-based mitigation strategies and mobility on COVID-19 infections throughout the pandemic is critical for informing responses and future infectious disease outbreaks. Here, we employed time-series analyses to empirically investigate the relationships between mitigation strategies and mobility on COVID-19 incident cases across US states during the first three waves of infections. METHODS: We linked data on daily COVID-19 incidence by US state from March to December 2020 with the stringency index, a well-known index capturing the strictness of mitigation strategies, and the trip ratio, which measures the ratio of the number of trips taken per day compared with the same day in 2019. We utilized multilevel models to determine the relative impacts of policy stringency and the trip ratio on COVID-19 cumulative incidence and the effective reproduction number. We stratified analyses by three waves of infections. RESULTS: Every five-point increase in the stringency index was associated with 2.89% (95% confidence interval = 1.52, 4.26%) and 5.01% (3.02, 6.95%) reductions in COVID-19 incidence for the first and third waves, respectively. Reducing the number of trips taken by 50% compared with the same time in 2019 was associated with a 16.2% (-0.07, 35.2%) decline in COVID-19 incidence at the state level during the second wave and 19.3% (2.30, 39.0%) during the third wave. CONCLUSIONS: Mitigation strategies and reductions in mobility are associated with marked health gains through the reduction of COVID-19 infections, but we estimate variable impacts depending on policy stringency and levels of adherence.

Microsimulation Model Calibration with Approximate Bayesian Computation in R: A Tutorial.

Mathematical health policy models, including microsimulation models (MSMs), are widely used to simulate complex processes and predict outcomes consistent with available data. Calibration is a method to estimate parameter values such that model predictions are similar to observed outcomes of interest. Bayesian calibration methods are popular among the available calibration techniques, given their strong theoretical basis and flexibility to incorporate prior beliefs and draw values from the posterior distribution of model parameters and hence the ability to characterize and evaluate parameter uncertainty in the model outcomes. Approximate Bayesian computation (ABC) is an approach to calibrate complex models in which the likelihood is intractable, focusing on measuring the difference between the simulated model predictions and outcomes of interest in observed data. Although ABC methods are increasingly being used, there is limited practical guidance in the medical decision-making literature on approaches to implement ABC to calibrate MSMs. In this tutorial, we describe the Bayesian calibration framework, introduce the ABC approach, and provide step-by-step guidance for implementing an ABC algorithm to calibrate MSMs, using 2 case examples based on a microsimulation model for dementia. We also provide the R code for applying these methods.

Stigma Determines Antiretroviral Adherence in Adults With HIV in Myanmar.

INTRODUCTION: Understanding social and structural barriers that determine antiretroviral therapy (ART) adherence can improve care. Assessment of such factors is limited in Myanmar, a country with high HIV prevalence and increasing number of people living with HIV initiating ART. METHODS: Questionnaires were administered to adults with HIV across 4 Myanmar cities to estimate adherence and its potential determinants, including HIV knowledge, social support, barriers to care, enacted and internalized stigma, and engagement in peer-to-peer HIV counseling (PC). Associations were determined using logistic mixed-effects modeling. RESULTS: Among 956 participants, the mean age was 39 years, 52% were female, 36% had CD4 <350 cells/mm3, and 50% received pre-ART PC. Good adherence was reported by 74% of participants who had better HIV knowledge than those reporting nonadherence. Among nonadherent, 44% were forgetful and 81% were careless about taking ART. Among all participants, most (53%) were very satisfied with their social support and 79% reported lack of financial resources as barriers to care. Participants most frequently reported being viewed differently by others (30%) and feeling as if they were paying for past karma or sins because of their HIV diagnosis (66%). Enacted stigma (odds ratio 0.86; 95% confidence interval 0.79 to 0.92, P < 0.01) and internalized stigma (odds ratio 0.73; 95% confidence interval: 0.56 to 0.95, P = 0.023) were associated with worse adherence. CONCLUSIONS: Increased self-reported ART adherence in Myanmar is associated with less enacted and internalized stigma. These findings suggest the benefit of developing and promoting adherence interventions, which are focused on mitigating HIV-related stigma in the county.

Egocentric social network characteristics and cardiovascular risk among patients with hypertension or diabetes in western Kenya: a cross-sectional analysis from the BIGPIC trial.

OBJECTIVES: Management of cardiovascular disease (CVD) is an urgent challenge in low-income and middle-income countries, and interventions may require appraisal of patients' social networks to guide implementation. The purpose of this study is to determine whether egocentric social network characteristics (SNCs) of patients with chronic disease in western Kenya are associated with overall CVD risk and individual CVD risk factors. DESIGN: Cross-sectional analysis of enrollment data (2017-2018) from the Bridging Income Generation with GrouP Integrated Care trial. Non-overlapping trust-only, health advice-only and multiplex (trust and health advice) egocentric social networks were elicited for each participant, and SNCs representing social cohesion were calculated. SETTING: 24 communities across four counties in western Kenya. PARTICIPANTS: Participants (n=2890) were ≥35 years old with diabetes (fasting glucose ≥7 mmol/L) or hypertension. PRIMARY AND SECONDARY OUTCOMES: We hypothesised that SNCs would be associated with CVD risk status (QRISK3 score). Secondary outcomes were individual CVD risk factors. RESULTS: Among the 2890 participants, 2020 (70%) were women, and mean (SD) age was 60.7 (12.1) years. Forty-four per cent of participants had elevated QRISK3 score (≥10%). No relationship was observed between QRISK3 level and SNCs. In unadjusted comparisons, participants with any individuals in their trust network were more likely to report a good than a poor diet (41% vs 21%). SNCs for the trust and multiplex networks accounted for a substantial fraction of variation in measures of dietary quality and physical activity (statistically significant via likelihood ratio test, adjusted for false discovery rate). CONCLUSION: SNCs indicative of social cohesion appear to be associated with individual behavioural CVD risk factors, although not with overall CVD risk score. Understanding how SNCs of patients with chronic diseases relate to modifiable CVD risk factors could help inform network-based interventions. TRIAL REGISTRATION NUMBER: ClinicalTrials.gov identifier: NCT02501746; https://clinicaltrials.gov/ct2/show/NCT02501746.

Cost-Effectiveness of Care Environments for Improving the Mental Health of Orphaned and Separated Children and Adolescents in Kenya

BACKGROUND: Institutionalization has shown contradictory effects on the mental health of orphaned and separated children and adolescents (OSCA) in sub-Saharan Africa. There is a paucity of data surrounding the cost-effectiveness of different care environments for improving OSCA's mental health. AIMS OF THE STUDY: The goal of this analysis was to evaluate the cost-effectiveness of Charitable Children's Institutions (orphanages) compared to family-based settings serving OSCA in East Africa in terms of USD/unit reduction in mental health diagnoses (depression, anxiety, post-traumatic stress disorder, suicidality) and quality-adjusted life-year (QALY) gained. METHODS: This economic analysis was conducted from a societal perspective as part of the Orphaned and Separated Children's Assessments Related to their (OSCAR's) Health and Well-Being Project, a 10-year longitudinal cohort study evaluating the effects of different care environments on OSCA's physical and psychological health in western Kenya. Cost data were ascertained from 9 institutions and 225 family-based settings in the OSCAR cohort via survey assessments, budget reports, and expert interviews. Monthly per-child costs were calculated as the sum of recurrent and capital costs divided by the environment's maximum residential capacity, and cost differences between care environments were estimated using two-part models. Mental health effectiveness outcomes were derived from prior survival regression analyses conducted among the OSCAR cohort. We used Child Depression Inventory Short-Form scores at baseline and follow-up to calculate the number of depression-free days (DFDs) over the follow-up period, and translated DFDs into QALYs using established utility weights. Incremental cost-effectiveness ratios (ICERs) were calculated as the difference in monthly per-child cost divided by the difference in each mental health outcome, comparing institutions to family-based settings. Sampling uncertainty in the ICERs was handled using nonparametric bootstrapping with 1,000 replications. We assumed a willingness-to-pay threshold of three times Kenya's per capita gross domestic product. RESULTS: Charitable Children's Institutions cost USD 123 more on average than family-based settings in terms of monthly per-child expenditures (p<0.001). Compared to family-based care, institutional care resulted in an ICER of USD 236, USD 280, USD 397, and USD 456 per unit reduction in depression, anxiety, PTSD, and suicidal diagnosis among OSCA, respectively. The incremental cost per additional QALY was USD 4,929 (95% CI: USD 3096 -- USD 6740). The probability of Charitable Children's Institutions being more cost-effective than family-based settings was greater than 90% for willingness-to-pay thresholds above USD 7,000/QALY. DISCUSSION: Only a subset of institutions in the cohort were willing to provide budgetary information for this assessment, which potentially biased our cost estimates. However, institutions who did not provide budget data likely had lower expenditures than those for whom cost data were collected, leading to more conservative cost estimates. Furthermore, our QALY estimates were based solely on depression-free days such that OSCA in institutions may experience added mental health benefits for no additional costs. IMPLICATIONS FOR HEALTH POLICY: Compared to family-based settings, institutions may be more cost-effective for improving mental health outcomes among orphaned and separated children and adolescents. Our findings suggest that policy-makers should prioritize resources to strengthen family-based care but that formal institutions can offer cost-effective, mental health support as a last resort.

Bayesian versus Empirical Calibration of Microsimulation Models: A Comparative Analysis.

Calibration of a microsimulation model (MSM) is a challenging but crucial step for the development of a valid model. Numerous calibration methods for MSMs have been suggested in the literature, most of which are usually adjusted to the specific needs of the model and based on subjective criteria for the selection of optimal parameter values. This article compares 2 general approaches for calibrating MSMs used in medical decision making, a Bayesian and an empirical approach. We use as a tool the MIcrosimulation Lung Cancer (MILC) model, a streamlined, continuous-time, dynamic MSM that describes the natural history of lung cancer and predicts individual trajectories accounting for age, sex, and smoking habits. We apply both methods to calibrate MILC to observed lung cancer incidence rates from the Surveillance, Epidemiology and End Results (SEER) database. We compare the results from the 2 methods in terms of the resulting parameter distributions, model predictions, and efficiency. Although the empirical method proves more practical, producing similar results with smaller computational effort, the Bayesian method resulted in a calibrated model that produced more accurate outputs for rare events and is based on a well-defined theoretical framework for the evaluation and interpretation of the calibration outcomes. A combination of the 2 approaches is an alternative worth considering for calibrating complex predictive models, such as microsimulation models.

Group Medical Visit and Microfinance Intervention for Patients With Diabetes or Hypertension in Kenya.

BACKGROUND: Incorporating social determinants of health into care delivery for chronic diseases is a priority. OBJECTIVES: The goal of this study was to evaluate the impact of group medical visits and/or microfinance on blood pressure reduction. METHODS: The authors conducted a cluster randomized trial with 4 arms and 24 clusters: 1) usual care (UC); 2) usual care plus microfinance (MF); 3) group medical visits (GMVs); and 4) GMV integrated into MF (GMV-MF). The primary outcome was 1-year change in systolic blood pressure (SBP). Mixed-effects intention-to-treat models were used to evaluate the outcomes. RESULTS: A total of 2,890 individuals (69.9% women) were enrolled (708 UC, 709 MF, 740 GMV, and 733 GMV-MF). Average baseline SBP was 157.5 mm Hg. Mean SBP declined -11.4, -14.8, -14.7, and -16.4 mm Hg in UC, MF, GMV, and GMV-MF, respectively. Adjusted estimates and multiplicity-adjusted 98.3% confidence intervals showed that, relative to UC, SBP reduction was 3.9 mm Hg (-8.5 to 0.7), 3.3 mm Hg (-7.8 to 1.2), and 2.3 mm Hg (-7.0 to 2.4) greater in GMV-MF, GMV, and MF, respectively. GMV and GMV-MF tended to benefit women, and MF and GMV-MF tended to benefit poorer individuals. Active participation in GMV-MF was associated with greater benefit. CONCLUSIONS: A strategy combining GMV and MF for individuals with diabetes or hypertension in Kenya led to clinically meaningful SBP reductions associated with cardiovascular benefit. Although the significance threshold was not met in pairwise comparison hypothesis testing, confidence intervals for GMV-MF were consistent with impacts ranging from substantive benefit to neutral effect relative to UC. Incorporating social determinants of health into care delivery for chronic diseases has potential to improve outcomes. (Bridging Income Generation With Group Integrated Care [BIGPIC]; NCT02501746).

The impact of care environment on the mental health of orphaned, separated and street-connected children and adolescents in western Kenya: a prospective cohort analysis.

INTRODUCTION: The effect of care environment on orphaned and separated children and adolescents' (OSCA) mental health is not well characterised in sub-Saharan Africa. We compared the risk of incident post-traumatic stress disorder (PTSD), depression, anxiety and suicidality among OSCA living in Charitable Children's Institutions (CCIs), family-based care (FBC) and street-connected children and youth (SCY). METHODS: This prospective cohort followed up OSCA from 300 randomly selected households (FBC), 19 CCIs and 100 SCY in western Kenya from 2009 to 2019. Annual data were collected through standardised assessments. We fit survival regression models to investigate the association between care environment and mental health diagnoses. RESULTS: The analysis included 1931 participants: 1069 in FBC, 783 in CCIs and 79 SCY. At baseline, 1004 participants (52%) were male with a mean age (SD) of 13 years (2.37); 54% were double orphans. In adjusted analysis (adjusted HR, AHR), OSCA in CCIs were significantly less likely to be diagnosed with PTSD (AHR 0.69, 95% CI 0.49 to 0.97), depression (AHR 0.48 95% CI 0.24 to 0.97), anxiety (AHR 0.56, 95% CI 0.45 to 0.68) and suicidality (AHR 0.73, 95% CI 0.56 to 0.95) compared with those in FBC. SCY were significantly more likely to be diagnosed with PTSD (AHR 4.52, 95% CI 4.10 to 4.97), depression (AHR 4.72, 95% CI 3.12 to 7.15), anxiety (AHR 4.71, 95% CI 1.56 to 14.26) and suicidality (AHR 3.10, 95% CI 2.14 to 4.48) compared with those in FBC. CONCLUSION: OSCA living in CCIs in this setting were significantly less likely to have incident mental illness, while SCY were significantly more, compared with OSCA in FBC.

Projected Estimates of Opioid Mortality After Community-Level Interventions.

Importance: The United States is experiencing a crisis of opioid overdose. In response, the US Department of Health and Human Services has defined a goal to reduce overdose mortality by 40% by 2022. Objective: To identify specific combinations of 3 interventions (initiating more people to medications for opioid use disorder [MOUD], increasing 6-month retention with MOUD, and increasing naloxone distribution) associated with at least a 40% reduction in opioid overdose in simulated populations. Design, Setting, and Participants: This decision analytical model used a dynamic population-level state-transition model to project outcomes over a 2-year horizon. Each intervention scenario was compared with the counterfactual of no intervention in simulated urban and rural communities in Massachusetts. Simulation modeling was used to determine the associations of community-level interventions with opioid overdose rates. The 3 examined interventions were initiation of more people to MOUD, increasing individuals' retention with MOUD, and increasing distribution of naloxone. Data were analyzed from July to November 2020. Main Outcomes and Measures: Reduction in overdose mortality, medication treatment capacity needs, and naloxone needs. Results: No single intervention was associated with a 40% reduction in overdose mortality in the simulated communities. Reaching this goal required use of MOUD and naloxone. Achieving a 40% reduction required that 10% to 15% of the estimated OUD population not already receiving MOUD initiate MOUD every month, with 45% to 60%% retention for at least 6 months, and increased naloxone distribution. In all feasible settings and scenarios, attaining a 40% reduction in overdose mortality required that in every month, at least 10% of the population with OUD who were not currently receiving treatment initiate an MOUD. Conclusions and Relevance: In this modeling study, only communities with increased capacity for treating with MOUD and increased MOUD retention experienced a 40% decrease in overdose mortality. These findings could provide a framework for developing community-level interventions to reduce opioid overdose death.

Cash Transfers and Contraceptive Use: A Regression Discontinuity Analysis.

Conditional cash transfers (CCTs) have become important components of social protection policies in Latin America. By establishing coresponsibilities tied to health and education, CCTs may reduce poverty and encourage human capital investment. While CCT programs can have unintended effects on sexual and reproductive health outcomes, such effects have been mixed and poorly documented in South America. This study examines the impact of Ecuador's CCT program, Bono de Desarrollo Humano, on contraceptive behavior among women of childbearing age who are sexually active and do not wish to become pregnant. We analyze nationally representative data in a regression-discontinuity quasi-experimental design. Using an instrumental variable approach and a set of robustness checks, our study finds no significant effects of the CCT program on contraceptive use. Our results offer important considerations for the ongoing policy debate in South America regarding the effects of cash transfer programs on beneficiaries.

Comparison of machine learning methods for predicting viral failure: a case study using electronic health record data.

Background: Human immunodeficiency virus (HIV) viral failure occurs when antiretroviral therapy fails to suppress and sustain a person's viral load count below 1,000 copies of viral ribonucleic acid per milliliter. For those newly diagnosed with HIV and living in a setting where healthcare resources are limited, such as a low- and middle-income country, the World Health Organization recommends viral load monitoring six months after initiation of antiretroviral treatment and yearly thereafter. Deviations from this schedule are made in cases where viral failure occurs or at the discretion of the clinician. Failure to detect viral failure in a timely fashion can lead to delayed administration of essential interventions. Clinical prediction models based on information available in the patient medical record are increasingly being developed and deployed for decision support in clinical medicine and public health. This raises the possibility that prediction models can be used to detect potential for viral failure in advance of viral measurements, particularly when those measurements occur infrequently. Objective: Our goal is to use electronic health record data from a large HIV care program in Kenya to characterize and compare the predictive accuracy of several statistical machine learning methods for predicting viral failure at the first and second measurements following initiation of antiretroviral therapy. Predictive accuracy is measured in terms of sensitivity, specificity and area under the receiver-operator characteristic curve. Methods: We trained and cross-validated 10 statistical machine learning models and algorithms on data from over 10,000 patients in the Academic Model Providing Access to Healthcare care program in western Kenya. These included parametric, non-parametric, ensemble, and Bayesian methods. The input variables included 50 items from the clinical record, hand picked in consultation with clinician experts. Predictive accuracy measures were calculated using 10-fold cross validation. Results: Viral load failure rate is about 20% in this patient cohort at both the first and second measurements. Ensemble techniques generally outperformed other methods. For predicting viral failure at the first follow up measure, specificity was over 90% for these methods, but sensitivity was typically in the 50-60% range. Predictive accuracy was greater for the second follow up measure, with sensitivities over 80%. Super Learner, gradient boosting and Bayesian additive regression trees consistently outperformed other methods. For a viral failure rate of 20%, the positive predictive value for the top-performing methods is between 75 and 85%, while the negative predictive value is over 95%. Conclusion: Evidence from this study suggests that machine learning techniques have potential to identify patients at risk for viral failure prior to their scheduled measurements. Ultimately, prognostic virologic assessment can help guide the administration of earlier targeted intervention such as enhanced drug resistance monitoring, rigorous adherence counseling, or appropriate next-line therapy switching. External validation studies should be used to confirm the results found here.

Community Health Workers Improve Linkage to Hypertension Care in Western Kenya.

BACKGROUND: Elevated blood pressure (BP) is the leading global risk factor for mortality. Delay in seeking hypertension care is associated with increased mortality. OBJECTIVES: This study investigated whether community health workers, equipped with behavioral communication strategies and smartphone technology, can increase linkage of individuals with elevated BP to a hypertension care program in western Kenya and significantly reduce BP. METHODS: The study was a cluster randomized trial with 3 arms: 1) usual care (standard training); 2) "paper-based" (tailored behavioral communication, using paper-based tools); and 3) "smartphone" (tailored behavioral communication, using smartphone technology). The co-primary outcomes were: 1) linkage to care; and 2) change in systolic BP (SBP). A covariate-adjusted mixed-effects model was used, adjusting for differential time to follow-up. Bootstrap and multiple imputation were used to handle missing data. RESULTS: A total of 1,460 individuals (58% women) were enrolled (491 usual care, 500 paper-based, 469 smartphone). Average baseline SBP was 159.4 mm Hg. Follow-up measures of linkage were available for 1,128 (77%) and BP for 1,106 (76%). Linkage to care was 49% overall, with significantly greater linkage in the usual care and smartphone arms of the trial. Average overall follow-up SBP was 149.9 mm Hg. Participants in the smartphone arm experienced a modestly greater reduction in SBP versus usual care (-13.1 mm Hg vs. -9.7 mm Hg), but this difference was not statistically significant. Mediation analysis revealed that linkage to care contributed to SBP change. CONCLUSIONS: A strategy combining tailored behavioral communication and mobile health (mHealth) for community health workers led to improved linkage to care, but not statistically significant improvement in SBP reduction. Further innovations to improve hypertension control are needed. (Optimizing Linkage and Retention to Hypertension Care in Rural Kenya [LARK]; NCT01844596).

Missing Data in Marginal Structural Models: A Plasmode Simulation Study Comparing Multiple Imputation and Inverse Probability Weighting.

BACKGROUND: The use of marginal structural models (MSMs) to adjust for time-varying confounding has increased in epidemiologic studies. However, in the setting of MSMs, recommendations for how best to handle missing data are contradictory. We present a plasmode simulation study to compare the validity and precision of MSMs estimates using complete case analysis (CC), multiple imputation (MI), and inverse probability weighting (IPW) in the presence of missing data on time-independent and time-varying confounders. MATERIALS AND METHODS: Simulations were based on a cohort substudy using data from the Osteoarthritis Initiative which estimated the marginal causal effect of intra-articular injection use on yearly changes in knee pain. We simulated 81 scenarios with parameter values varied on missing mechanisms (MCAR, MAR, and MNAR), percentages of missing (10%, 20%, and 30%), type of confounders (time-independent, time-varying, either or both), and analytical approaches (CC, IPW, and MI). The performance of CC, IPW, and MI methods was compared using relative bias, mean squared error of the estimates of interest, and empirical power. RESULTS: Across scenarios defined by missing data mechanism, extent of missing data, and confounder type, MI generally produced less biased estimates (range: 1.2%-6.7%) with better precision (range: 0.17-0.18) compared with IPW (relative bias: -5.3% to 8.0%; precision: 0.19-0.53). Empirical power was constant across the scenarios using MI. CONCLUSIONS: Under simple yet realistically constructed scenarios, MI seems to confer an advantage over IPW in MSMs applications.

The use of latent class analysis for identifying subtypes of depression: A systematic review.

Depression is a significant public health problem but symptom remission is difficult to predict. This may be due to substantial heterogeneity underlying the disorder. Latent class analysis (LCA) is often used to elucidate clinically relevant depression subtypes but whether or not consistent subtypes emerge is unclear. We sought to critically examine the implementation and reporting of LCA in this context by performing a systematic review to identify articles detailing the use of LCA to explore subtypes of depression among samples of adults endorsing depression symptoms. PubMed, PsycINFO, CINAHL, Scopus, and Google Scholar were searched to identify eligible articles indexed prior to January 2016. Twenty-four articles reporting 28 LCA models were eligible for inclusion. Sample characteristics varied widely. The majority of articles used depression symptoms as the observed indicators of the latent depression subtypes. Details regarding model fit and selection were often lacking. No consistent set of depression subtypes was identified across studies. Differences in how models were constructed might partially explain the conflicting results. Standards for using, interpreting, and reporting LCA models could improve our understanding of the LCA results. Incorporating dimensions of depression other than symptoms, such as functioning, may be helpful in determining depression subtypes.

Prevalence of Long-Term Opioid Use in Long-Stay Nursing Home Residents.

BACKGROUND/OBJECTIVES: Overall and long-term opioid use among older adults have increased since 1999. Less is known about opioid use in older adults in nursing homes (NHs). DESIGN: Cross-sectional. SETTING: U.S. NHs (N = 13,522). PARTICIPANTS: Long-stay NH resident Medicare beneficiaries with a Minimum Data Set 3.0 (MDS) assessment between April 1, 2012, and June 30, 2012, and 120 days of follow-up (N = 315,949). MEASUREMENTS: We used Medicare Part D claims to measure length of opioid use in the 120 days from the index assessment (short-term: ≤30 days, medium-term: >30-89 days, long-term: ≥90 days), adjuvants (e.g., anticonvulsants), and other pain medications (e.g., corticosteroids). MDS assessments in the follow-up period were used to measure nonpharmacological pain management use. Modified Poisson models were used to estimate adjusted prevalence ratios (aPR) and 95% confidence intervals (CI) for age, gender, race and ethnicity, cognitive and physical impairment, and long-term opioid use. RESULTS: Of all long-stay residents, 32.4% were prescribed any opioid, and 15.5% were prescribed opioids long-term. Opioid users (versus nonusers) were more commonly prescribed pain adjuvants (32.9% vs 14.9%), other pain medications (25.5% vs 11.0%), and nonpharmacological pain management (24.5% vs 9.3%). Long-term opioid use was higher in women (aPR = 1.21, 95% CI = 1.18-1.23) and lower in racial and ethnic minorities (non-Hispanic blacks vs whites: APR = 0.93, 95% CI = 0.90-0.94) and those with severe cognitive impairment (vs no or mild impairment, aPR = 0.82, 95% CI = 0.79-0.83). CONCLUSION: One in seven NH residents was prescribed opioids long-term. Recent guidelines on opioid prescribing for pain recommend reducing long-term opioid use, but this is challenging in NHs because residents may not benefit from nonpharmacological and nonopioid interventions. Studies to address concerns about opioid safety and effectiveness (e.g., on pain and functional status) in NHs are needed.

Survivors of an Acute Coronary Syndrome With Lower Patient Activation Are More Likely to Experience Declines in Health-Related Quality of Life.

BACKGROUND: Patient activation comprises the knowledge, skills, and confidence for self-care and may lead to better health outcomes. OBJECTIVES: We examined the relationship between patient activation and changes in health-related quality of life (HRQOL) after hospitalization for an acute coronary syndrome (ACS). METHODS: We studied patients from 6 medical centers in central Massachusetts and Georgia who had been hospitalized for an ACS between 2011 and 2013. At 1 month after hospital discharge, the patients completed the 6-item Patient Activation Measure and were categorized into 4 levels of activation. Multinomial logistic regression analyses compared activation level with clinically meaningful changes (≥3.0 points, generic; ≥10.0 points, disease-specific) in generic physical (SF-36v2 Physical Component Summary [PCS]), generic mental (SF-36v2 Mental Component Summary [MCS]), and disease-specific (Seattle Angina Questionnaire [SAQ]) HRQOL from 1 to 3 and 1 to 6 months after hospitalization, adjusting for potential sociodemographic and clinical confounders. RESULTS: The patients (N = 1042) were, on average, 62 years old, 34% female, and 87% non-Hispanic white. A total of 10% were in the lowest level of activation. The patients with the lowest activation had 1.95 times (95% confidence interval, 1.05-3.62) and 2.18 times (95% confidence interval, 1.17-4.05) the odds of experiencing clinically significant declines in MCS and SAQ HRQOL, respectively, between 1 and 6 months than the most activated patients. The patient activation level was not associated with meaningful changes in PCS scores. CONCLUSIONS: Hospital survivors of an ACS with lower activation may be more likely to experience declines in mental and disease-specific HRQOL than more-activated patients, identifying a group at risk of poor outcomes.