Overcoming Data Gaps in Life Course Epidemiology by Matching Across Cohorts.

Life course epidemiology is hampered by the absence of large studies with exposures and outcomes measured at different life stages in the same individuals. We describe when the effect of an exposure ( A ) on an outcome ( Y ) in a target population is identifiable in a combined ("synthetic") cohort created by pooling an early-life cohort including measures of A with a late-life cohort including measures of Y . We enumerate causal assumptions needed for unbiased effect estimation in the synthetic cohort and illustrate by simulating target populations under four causal models. From each target population, we randomly sampled early- and late-life cohorts and created a synthetic cohort by matching individuals from the two cohorts based on mediators and confounders. We estimated the effect of A on Y in the synthetic cohort, varying matching variables, the match ratio, and the strength of association between matching variables and A . Finally, we compared bias in the synthetic cohort estimates when matching variables did not d-separate A and Y to the bias expected in the original cohort. When the set of matching variables includes all variables d-connecting exposure and outcome (i.e., variables blocking all backdoor and front-door pathways), the synthetic cohort yields unbiased effect estimates. Even when matching variables did not fully account for confounders, the synthetic cohort estimate was sometimes less biased than comparable estimates in the original cohort. Methods based on merging cohorts may hasten the evaluation of early- and mid-life determinants of late-life health but rely on available measures of both confounders and mediators.

Risk factors for COVID-19 in-hospital mortality in Argentina: A competing risk survival analysis.

COVID-19 poses dire threats for low and middle-income countries (LMICs). Yet, there remains limited rigorous evidence describing the characteristics and outcomes of hospitalized patients for LMICs, and often the evidence was based on small samples and/or unicentric. The objective of this study was to examine risk factors of COVID-19 mortality in Argentina, a hard-hit middle-income Latin American country. We analyze data on 5,146 COVID-19 patients from 11 centers across 10 cities in Argentina, making this one of the largest multi-centric retrospective observational descriptive studies in the LMICs. Information on demographics and co-morbidities was extracted from medical records. Outcomes of relevance consisted of whether the patient was discharged or deceased (as established in medical records), along with date of each event. We use survival models that account for competing risks. Median age was 60 years (IQR: 48-72), there were fewer women (40.8%) hospitalized than men (59.2%), and the most prevalent comorbidities were hypertension (40.9%), diabetes (20.0%) and obesity (19.1%). Patients were hospitalized for a median duration of 8 days (IQR: 5-13), and in-hospital mortality was 18.1%, though it varied substantially across health centers (95%CI: 17.1%-19.2%). Baseline characteristics most associated with in-hospital mortality were respiratory rate (adjusted HR = 3.6, 95%CI: 2.5-5.4 for ≥ 26 breathes/min), older age (adjusted HR = 2.5, 95%CI: 2.0-3.3 for the 80+ age group), and chronic kidney disease (adjusted HR = 2.2, 95%CI: 1.8-2.8). Associations were attenuated when survival models did not account for the competing risk of being discharged. We document lower mortality rates than those in prior studies, likely due to a lower prevalence of comorbidities amongst patients in our sample. Compared with standard Cox models, we find that, when using competing risk models, risk factors have a larger role in explaining COVID-19 mortality. Overall, we provide rigorous evidence describing the characteristics and outcomes of hospitalized patients for LMICs. Thus, our findings are useful to conduct a more accurate in-hospital monitoring of patient subgroups who may be at greater risk. They also provide valuable guidance for public health and policy efforts in Argentina and other developing countries.

Regression discontinuity design to evaluate the effect of statins on myocardial infarction in electronic health records.

Regression discontinuity design (RDD) is a quasi-experimental method intended for causal inference in observational settings. While RDD is gaining popularity in clinical studies, there are limited real-world studies examining the performance on estimating known trial casual effects. The goal of this paper is to estimate the effect of statins on myocardial infarction (MI) using RDD and compare with propensity score matching and Cox regression. For the RDD, we leveraged a 2008 UK guideline that recommends statins if a patient's 10-year cardiovascular disease (CVD) risk score > 20%. We used UK electronic health record data from the Health Improvement Network on 49,242 patients aged 65 + in 2008-2011 (baseline) without a history of CVD and no statin use in the two years prior to the CVD risk score assessment. Both the regression discontinuity (n = 19,432) and the propensity score matched populations (n = 24,814) demonstrated good balance of confounders. Using RDD, the adjusted point estimate for statins on MI was in the protective direction and similar to the statin effect observed in clinical trials, although the confidence interval included the null (HR = 0.8, 95% CI 0.4, 1.4). Conversely, the adjusted estimates using propensity score matching and Cox regression remained in the harmful direction: HR = 2.42 (95% CI 1.96, 2.99) and 2.51 (2.12, 2.97). RDD appeared superior to other methods in replicating the known protective effect of statins with MI, although precision was poor. Our findings suggest that, when used appropriately, RDD can expand the scope of clinical investigations aimed at causal inference by leveraging treatment rules from everyday clinical practice.

Propensity Scores in Health Disparities Research: The Example of Cognitive Aging and the Hispanic Paradox.

BACKGROUND: Individuals of Mexican ancestry in the United States experience substantial socioeconomic disadvantages compared with non-Hispanic white individuals; however, some studies show these groups have similar dementia risk. Evaluating whether migration selection factors (e.g., education) associated with risk of Alzheimer disease and related dementia (ADRD) explain this paradoxical finding presents statistical challenges. Intercorrelation of risk factors, common with social determinants, could make certain covariate patterns very likely or unlikely to occur for particular groups, which complicates their comparison. Propensity score (PS) methods could be leveraged here to diagnose nonoverlap and help balance exposure groups. METHODS: We compare conventional and PS-based methods to examine differences in cognitive trajectories between foreign-born Mexican American, US-born Mexican American, and US-born non-Hispanic white individuals in the Health and Retirement Study (1994-2018). We examined cognition using a global measure. We estimated trajectories of cognitive decline from linear mixed models adjusted for migration selection factors also associated with ADRD risk conventionally or with inverse probability weighting. We also employed PS trimming and match weighting. RESULTS: In the full sample, where PS overlap was poor, unadjusted analyses showed both Mexican ancestry groups had worse baseline cognitive scores but similar or slower rates of decline compared with non-Hispanic white adults; adjusted findings were similar, regardless of method. Focusing analyses on populations where PS overlap was improved (PS trimming and match weighting) did not alter conclusions. CONCLUSIONS: Attempting to equalize groups on migration selection and ADRD risk factors did not explain paradoxical findings for Mexican ancestry groups in our study.

History of Low Hourly Wage and All-Cause Mortality Among Middle-aged Workers.

Importance: Earning a low wage is an increasingly recognized public health concern, yet little research exists on the long-term health consequences of sustained low-wage earning. Objective: To examine the association of sustained low-wage earning and mortality in a sample of workers with hourly wage reported biennially during peak midlife earning years. Design, Setting, and Participants: This longitudinal study included 4002 US participants, aged 50 years or older, from 2 subcohorts of the Health and Retirement Study (1992-2018) who worked for pay and reported earning hourly wages at 3 or more time points during a 12-year period during their midlife (1992-2004 or 1998-2010). Outcome follow-up occurred from the end of the respective exposure periods until 2018. Exposures: Low-wage-less than the hourly wage for full-time, full-year work at the federal poverty line-earning history was categorized as never earning a low wage, intermittently earning a low wage, and sustained earning a low wage. Main Outcomes and Measures: Cox proportional hazards and additive hazards regression models sequentially adjusted for sociodemographics, and economic and health covariates were used to estimate associations between low-wage history and all-cause mortality. We examined interaction with sex or employment stability on multiplicative and additive scales. Results: Of the 4002 workers (aged 50-57 years at the beginning of exposure period and 61-69 years at the end), 1854 (46.3%) were female; 718 (17.9%) experienced employment instability; 366 (9.1%) had a history of sustained low-wage earning; 1288 (32.2%) had intermittent low-wage earning periods; and 2348 (58.7%) had never earned a low wage. In unadjusted analyses, those who had never earned low wages experienced 199 deaths per 10 000 person-years, those with intermittent low wages, 208 deaths per 10 000 person-years, and those with sustained low wages, 275 deaths per 10 000 person-years. In models adjusted for key sociodemographic variables, sustained low-wage earning was associated with mortality (hazard ratio [HR], 1.35; 95% CI, 1.07-1.71) and excess deaths (66; 95% CI, 6.6-125); these findings were attenuated with additional adjustments for economic and health covariates. Significant excess death and elevated mortality risk were observed for workers with sustained low-wage exposure and employment fluctuations (eg, for sustained low-wage × employment fluctuated, HR, 2.18; 95% CI, 1.35-3.53; for sustained low-wage × stable employment, HR, 1.17; 95% CI, 0.89,-1.54; P for interaction = .003). Conclusions and Relevance: Sustained low-wage earning may be associated with elevated mortality risk and excess deaths, especially when experienced alongside unstable employment. If causal, our findings suggest that social and economic policies that improve the financial standing of low-wage workers (eg, minimum wage laws) could improve mortality outcomes.

Practical Guide to Honest Causal Forests for Identifying Heterogeneous Treatment Effects.

"Heterogeneous treatment effects" is a term which refers to conditional average treatment effects (i.e., CATEs) that vary across population subgroups. Epidemiologists are often interested in estimating such effects because they can help detect populations that may particularly benefit from or be harmed by a treatment. However, standard regression approaches for estimating heterogeneous effects are limited by preexisting hypotheses, test a single effect modifier at a time, and are subject to the multiple-comparisons problem. In this article, we aim to offer a practical guide to honest causal forests, an ensemble tree-based learning method which can discover as well as estimate heterogeneous treatment effects using a data-driven approach. We discuss the fundamentals of tree-based methods, describe how honest causal forests can identify and estimate heterogeneous effects, and demonstrate an implementation of this method using simulated data. Our implementation highlights the steps required to simulate data sets, build honest causal forests, and assess model performance across a variety of simulation scenarios. Overall, this paper is intended for epidemiologists and other population health researchers who lack an extensive background in machine learning yet are interested in utilizing an emerging method for identifying and estimating heterogeneous treatment effects.

Association Between Supplemental Nutrition Assistance Program Use and Memory Decline: Findings From the Health and Retirement Study.

BACKGROUND AND OBJECTIVES: Studies on the effect of the Supplemental Nutrition Assistance Program (SNAP) on the cognitive health of older adults are scarce. We sought to examine the associations between SNAP use and memory decline among SNAP-eligible US older adults. METHODS: Participants aged 50+ years and SNAP-eligible in 1996 from the Health and Retirement Study were included. Participants' SNAP eligibility was constructed using federal criteria. Participants also self-reported whether they used SNAP. Memory function was assessed biennially from 1996 through 2016 using a composite score. To account for preexisting differences in characteristics between SNAP users and nonusers, we modeled the probability of SNAP use using demographic and health covariates. Using linear mixed-effects models, we then modeled trajectories of memory function for SNAP users and nonusers using inverse probability (IP) weighting and propensity score (PS) matching techniques. In all models, we accounted for study attrition. RESULTS: Of the 3,555 SNAP-eligible participants, a total of 15.7% were SNAP users. At baseline, SNAP users had lower socioeconomic status and a greater number of chronic conditions than nonusers and were more likely to be lost to follow-up. Our multivariable IP-weighted models suggested that SNAP users had worse memory scores at baseline but slower rates of memory decline compared with nonusers (the annual decline rate is -0.038 standardized units [95% CI = -0.044 to -0.032] for users and -0.046 [95% CI = -0.049 to -0.043] for nonusers). Results were slightly stronger from the PS-matched sample (N = 1,014) (the annual decline rate was -0.046 units [95% CI = -0.050 to -0.042] for users and -0.060 units [95% CI = -0.064 to -0.056] for nonusers). Put in other words, our findings suggested that SNAP users had approximately 2 fewer years of cognitive aging over a 10-year period compared with nonusers. DISCUSSION: After accounting for preexisting differences between eligible SNAP users and nonusers as well as differential attrition, we find SNAP use to be associated with slower memory function decline.

Racial Residential Segregation in Young Adulthood and Brain Integrity in Middle Age: Can We Learn From Small Samples?

Racial residential segregation is associated with multiple adverse health outcomes in Black individuals. Yet, the influence of structural racism and racial residential segregation on brain aging is less understood. In this study, we investigated the association between cumulative exposure to racial residential segregation over 25 years (1985-2010) in young adulthood, as measured by the Getis-Ord Gi* statistic, and year 25 measures of brain volume (cerebral, gray matter, white matter, and hippocampal volumes) in midlife. We studied 290 Black participants with available brain imaging data who were enrolled in the Coronary Artery Risk Development in Young Adults (CARDIA) Study, a prospective cohort study. CARDIA investigators originally recruited 2,637 Black participants aged 18-30 years from 4 field centers across the United States. We conducted analyses using marginal structural models, incorporating inverse probability of treatment weighting and inverse probability of censoring weighting. We found that compared with low/medium segregation, greater cumulative exposure to a high level of racial residential segregation throughout young adulthood was associated with smaller brain volumes in general (e.g., for cerebral volume, ß = -0.08, 95% confidence interval: -0.15, -0.02) and with a more pronounced reduction in hippocampal volume, though results were not statistically significant. Our findings suggest that exposure to segregated neighborhoods may be associated with worse brain aging.

Statins and Cognitive Decline in the Cardiovascular Health Study: A Comparison of Different Analytical Approaches.

BACKGROUND: Despite their well-established benefits for the prevention of cardiovascular disease, robust evidence on the effects of statins on cognition is largely inconclusive. We apply various study designs and analytical approaches to mimic randomized controlled trial effects from observational data. METHODS: We used observational data from 5 580 participants enrolled in the Cardiovascular Health Study from 1989/1990 to 1999/2000. We conceptualized the cohort as an overlapping sequence of nonrandomized trials. We compared multiple selection (eligible population, prevalent users, new users) and analytic approaches (multivariable adjustment, inverse-probability treatment weights, propensity score matching) to evaluate the association between statin use and 5-year change in global cognitive function, assessed using the Modified Mini-Mental State Examination (3MSE). RESULTS: When comparing prevalent users to nonusers (N = 2 772), statin use was associated with slower cognitive decline over 5 years (adjusted annual change in 3MSE = 0.34 points/year; 95% CI: 0.05-0.63). Compared to prevalent user design, estimates from new user designs (eg, comparing eligible statin initiators to noninitiators) were attenuated showing either null or negative association, though not significant. For example, in a propensity score-matched sample of statin-eligible individuals (N = 454), the annual 3MS change comparing statin initiators to noninitiators was -0.21 points/year (95% CI: -0.81 to 0.39). CONCLUSIONS: The association of statin use and cognitive decline is attenuated toward the null when using rigorous analytical approaches that more closely mimic randomized controlled trials. Point estimates, even within the same study, may vary depending on the analytical methods used. Further studies that leverage natural or quasi experiments around statin use are needed to replicate our findings.

Association of negative financial shocks during the Great Recession with depressive symptoms and substance use in the USA: the CARDIA study.

BACKGROUND: The Great Recession of 2008 was marked by large increases in unemployment and decreases in the household wealth of many Americans. In the 21st century, there have also been increases in depressive symptoms, alcohol use and drug use among some groups in the USA. The objective of this analysis is to evaluate the influence of negative financial shocks incurred during the Great Recession on depressive symptoms, alcohol and drug use. METHODS: We employed a quasi-experimental fixed-effects design, using data from adults enrolled in the Coronary Artery Risk Development in Young Adults (CARDIA) study. Our financial shock predictors were within-person change in employment status, income and debt to asset ratio between 2005 and 2010. Our outcomes were within-person change in depressive symptoms score, alcohol use and past 30-day drug use. RESULTS: In adjusted models, we found that becoming unemployed and experiencing a drop in income and were associated with an increase in depressive symptoms. Incurring more debts than assets was also associated with an increase in depressive symptoms and a slight decrease in daily alcohol consumption (mL). CONCLUSION: Our findings suggest that multiple types of financial shocks incurred during an economic recession negatively influence depressive symptoms among black and white adults in the USA, and highlight the need for future research on how economic recessions are associated with health.

Racial discrimination in medical care settings and opioid pain reliever misuse in a U.S. cohort: 1992 to 2015.

BACKGROUND: In the United States whites are more likely to misuse opioid pain relievers (OPRs) than blacks, and blacks are less likely to be prescribed OPRs than whites. Our objective is to determine whether racial discrimination in medical settings is protective for blacks against OPR misuse, thus mediating the black-white disparities in OPR misuse. METHODS: We used data from 3528 black and white adults in the Coronary Artery Risk Development in Young Adults (CARDIA) study, an ongoing multi-site cohort. We employ causal mediation methods, with race (black vs white) as the exposure, lifetime discrimination in medical settings prior to year 2000 as the mediator, and OPR misuse after 2000 as the outcome. RESULTS: We found black participants were more likely to report discrimination in a medical setting (20.3% vs 0.9%) and less likely to report OPR misuse (5.8% vs 8.0%, OR = 0.71, 95% CI = 0.55, 0.93, adjusted for covariates). Our mediation models suggest that when everyone is not discriminated against, the disparity is wider with black persons having even lower odds of reporting OPR misuse (OR = 0.63, 95% CI = 0.45, 0.89) compared to their white counterparts, suggesting racial discrimination in medical settings is a risk factor for OPR misuse rather than protective. CONCLUSIONS: These results suggest that racial discrimination in a medical setting is a risk factor for OPR misuse rather than being protective, and thus could not explain the seen black-white disparity in OPR misuse.

Associations of Income Volatility With Incident Cardiovascular Disease and All-Cause Mortality in a US Cohort.

BACKGROUND: Income volatility is on the rise and presents a growing public health problem. Because in many epidemiological studies income is measured at a single point in time, the association of long-term income volatility with incident cardiovascular disease (CVD) and mortality has not been adequately explored. The goal of this study was to examine associations of income volatility from 1990 to 2005 with incident CVD and all-cause mortality in the subsequent 10 years. METHODS: The Coronary Artery Risk Development in Young Adults Study is an ongoing prospective cohort study conducted within urban field centers in Birmingham, AL; Chicago, IL; Minneapolis, MN; and Oakland, CA. We studied 3937 black and white participants 23 to 35 years of age in 1990 (our study baseline). Income volatility was defined as the intraindividual SD of the percent change in income across 5 assessments from 1990 to 2005. An income drop was defined as a decrease of ≥25% from the previous visit and less than the participant's average income from 1990 to 2005. CVD events (fatal and nonfatal) and all-cause mortality between 2005 and 2015 were adjudicated with the use of medical records and death certificates. CVD included primarily acute events related to heart disease and stroke. RESULTS: A total of 106 CVD events and 164 deaths occurred between 2005 and 2015 (incident rate, 2.76 and 3.66 per 1000 person-years, respectively). From Cox models adjusted for sociodemographic, behavioral, and CVD risk factors, higher income volatility and more income drops were associated with greater CVD risk (high versus low volatility: hazard ratio, 2.07; 95% CI, 1.10-3.90; ≥2 versus 0 income drops: hazard ratio, 2.54; 95% CI, 1.24-5.19) and all-cause mortality (high versus low volatility: hazard ratio, 1.78; 95% CI,1.03-3.09; ≥2 versus 0 income drops: hazard ratio, 1.92; 95% CI, 1.07-3.44). CONCLUSIONS: In a cohort of relatively young adults, income volatility and drops during a 15-year period of formative earning years were independently associated with a nearly 2-fold risk of CVD and all-cause mortality.

New evidence on the persistence of health spending.

Surprisingly little is known about long-term spending patterns in the under-65 population. Such information could inform efforts to improve coverage and control costs. Using the MarketScan claims database, we characterize the persistence of health care spending in the privately insured, under-65 population. Over a 6-year period, 69.8% of enrollees never had annual spending in the top 10% of the distribution and the bottom 50% of spenders accounted for less than 10% of spending. Those in the top 10% in 2003 were almost as likely (34.4%) to be in the top 10% five years later as one year later (43.4%). Many comorbid conditions retained much of their predictive power even 5 years later. The persistence at both ends of the spending distribution indicates the potential for adverse selection and cream skimming and supports the use of disease management, particularly for those with the conditions that remained strong predictors of high spending throughout the follow-up period.