Search

In this paper, we present a practical approach for computing the sandwich variance estimator in 2-stage regression model settings. As a motivating example for 2-stage regression, we consider regression calibration, a popular approach for addressing covariate measurement error. The sandwich variance approach has rarely been applied in regression calibration, despite its requiring less computation time than popular resampling approaches for variance estimation, specifically the bootstrap. This is probably because it requires specialized statistical coding. Here we first outline the steps needed to compute the sandwich variance estimator. We then develop a convenient method of computation in R for sandwich variance estimation, which leverages standard regression model outputs and existing R functions and can be applied in the case of a simple random sample or complex survey design. We use a simulation study to compare the sandwich estimator to a resampling variance approach for both settings. Finally, we further compare these 2 variance estimation approaches in data examples from the Women's Health Initiative (1993–2005) and the Hispanic Community Health Study/Study of Latinos (2008–2011). In our simulations, the sandwich variance estimator typically had good numerical performance, but simple Wald bootstrap confidence intervals were unstable or overcovered in certain settings, particularly when there was high correlation between covariates or large measurement error. [ABSTRACT FROM AUTHOR]

The Preconception Period Analysis of Risks and Exposures Influencing Health and Development (PrePARED) Consortium creates a novel resource for addressing preconception health by merging data from numerous cohort studies. In this paper, we describe our data harmonization methods and results. Individual-level data from 12 prospective studies were pooled. The crosswalk-cataloging-harmonization procedure was used. The index pregnancy was defined as the first postbaseline pregnancy lasting more than 20 weeks. We assessed heterogeneity across studies by comparing preconception characteristics in different types of studies. The pooled data set included 114,762 women, and 25,531 (22%) reported at least 1 pregnancy of more than 20 weeks' gestation during the study period. Babies from the index pregnancies were delivered between 1976 and 2021 (median, 2008), at a mean maternal age of 29.7 (standard deviation, 4.6) years. Before the index pregnancy, 60% of women were nulligravid, 58% had a college degree or more, and 37% were overweight or obese. Other harmonized variables included race/ethnicity, household income, substance use, chronic conditions, and perinatal outcomes. Participants from pregnancy-planning studies had more education and were healthier. The prevalence of preexisting medical conditions did not vary substantially based on whether studies relied on self-reported data. Use of harmonized data presents opportunities to study uncommon preconception risk factors and pregnancy-related events. This harmonization effort laid the groundwork for future analyses and additional data harmonization. [ABSTRACT FROM AUTHOR]

The use of longitudinal finite mixture models such as group-based trajectory modeling has seen a sharp increase during the last few decades in the medical literature. However, these methods have been criticized, especially because of the data-driven modeling process, which involves statistical decision-making. In this paper, we propose an approach that uses the bootstrap to sample observations with replacement from the original data to validate the number of groups identified and to quantify the uncertainty in the number of groups. The method allows investigation of the statistical validity and uncertainty of the groups identified in the original data by checking to see whether the same solution is also found across the bootstrap samples. In a simulation study, we examined whether the bootstrap-estimated variability in the number of groups reflected the replicationwise variability. We evaluated the ability of 3 commonly used adequacy criteria (average posterior probability, odds of correct classification, and relative entropy) to identify uncertainty in the number of groups. Finally, we illustrate the proposed approach using data from the Quebec Integrated Chronic Disease Surveillance System to identify longitudinal medication patterns between 2015 and 2018 in older adults with diabetes. [ABSTRACT FROM AUTHOR]

Estimating the fraction of dementia cases in a population attributable to a risk factor or combination of risk factors (the population attributable fraction (PAF)) informs the design and choice of dementia risk-reduction activities. It is directly relevant to dementia prevention policy and practice. Current methods employed widely in the dementia literature to combine PAFs for multiple dementia risk factors assume a multiplicative relationship between factors and rely on subjective criteria to develop weightings for risk factors. In this paper we present an alternative approach to calculating the PAF based on sums of individual risk. It incorporates individual risk factor interrelationships and enables a range of assumptions about the way in which multiple risk factors will combine to affect dementia risk. Applying this method to global data demonstrates that the previous estimate of 40% is potentially too conservative an estimate of modifiable dementia risk and would necessitate subadditive interaction between risk factors. We calculate a plausible conservative estimate of 55.7% (95% confidence interval: 55.2, 56.1) based on additive risk factor interaction. [ABSTRACT FROM AUTHOR]

The number of operations that surgeons have previously performed is associated with their patients' outcomes. However, this association may not be causal, because previous studies have often been cross-sectional and their analyses have not considered time-varying confounding or positivity violations. In this paper, using the example of surgeons who perform coronary artery bypass grafting, we describe (hypothetical) target trials for estimation of the causal effect of the surgeons' operative volumes on patient mortality. We then demonstrate how to emulate these target trials using data from US Medicare claims and provide effect estimates. Our target trial emulations suggest that interventions on physicians' volume of coronary artery bypass grafting operations have little effect on patient mortality. The target trial framework highlights key assumptions and draws attention to areas of bias in previous observational analyses that deviated from their implicit target trials. The principles of the presented methodology may be adapted to other scenarios of substantive interest in health services research. [ABSTRACT FROM AUTHOR]

Cancer is an important cause of childhood mortality, yet the etiology is largely unknown. A combination of pre- and postnatal factors is thought to be implicated, including maternal medication use. We aimed to provide: 1) a systematic review of peer-reviewed publications on associations between maternal medication use and childhood cancer, with a focus on study design and methodology; and 2) suggestions for how to increase transparency, limit potential biases, and improve comparability in studies on maternal medication use and childhood cancer. We conducted a systematic search in the PubMed, Embase, Scopus, Cochrane, and Web of Science databases to June 8, 2020. Altogether, 112 studies were identified. The reviewed studies were heterogeneous in study design, exposure, and outcome classification. In 21 studies (19%), the outcome was any childhood cancer. Of the 91 papers that reported on specific types of cancer, 62% did not report the cancer classification system. The most frequently investigated medication groups were sex hormones (46 studies, excluding fertility medications), and antiinfectives (37 studies). Suggestions for strengthening future pharmacoepidemiologic studies on maternal medication use and childhood cancer relate to choice of cancer classification system, exposure windows, and methods for identification of, and control for, potential confounders. [ABSTRACT FROM AUTHOR]

Two frequently encountered but underrecognized challenges for causal inference in studying the long-term health effects of disasters among survivors include 1) time-varying effects of disasters on a time-to-event outcome and 2) selection bias due to selective attrition. In this paper, we review approaches for overcoming these challenges and demonstrate application of the approaches to a real-world longitudinal data set of older adults who were directly affected by the 2011 Great East Japan Earthquake and Tsunami (n  = 4,857). To illustrate the problem of time-varying effects of disasters, we examined the association between degree of damage due to the tsunami and all-cause mortality. We compared results from Cox regression analysis assuming proportional hazards with those derived using adjusted parametric survival curves allowing for time-varying hazard ratios. To illustrate the problem of selection bias, we examined the association between proximity to the coast (a proxy for housing damage from the tsunami) and depressive symptoms. We corrected for selection bias due to attrition in the 2 postdisaster follow-up surveys (conducted in 2013 and 2016) using multivariable adjustment, inverse probability of censoring weighting, and survivor average causal effect estimation. Our results demonstrate that analytical approaches which ignore time-varying effects on mortality and selection bias due to selective attrition may underestimate the long-term health effects of disasters. [ABSTRACT FROM AUTHOR]

The development of tools for power and sample-size calculations for mediation analysis has lagged far behind the development of methods. The accompanying paper by Rudolph et al. (Am J Epidemiol. 2020;189(12):1559–1567) is a helpful contribution in using simulations as a tool for power calculations for more complex methods and settings. Much work remains to be done in the development of easy-to-use packages and simple online websites for carrying out power and sample-size calculations for mediation analysis. Much remains to be learned with respect to the relative power of different methods in different settings. There will likely be feedback between these 2 important frontiers of the tools we have available and of our understanding of power when conducting mediation analysis. [ABSTRACT FROM AUTHOR]

Epidemiologists often compare the observed number of deaths in a cohort with the expected number of deaths, obtained by multiplying person-time accrued in the cohort by mortality rates for a reference population (ideally, a reference that represents the mortality rate in the cohort in the absence of exposure). However, if exposure is hazardous (or salutary), this calculation will not consistently estimate the number of deaths expected in the absence of exposure because exposure will have affected the distribution of person-time observed in the study cohort. While problems with interpretation of this standard calculation of expected counts were discussed more than 2 decades ago, these discussions had little impact on epidemiologic practice. The logic of counterfactuals may help clarify this topic as we revisit these issues. In this paper, we describe a simple way to consistently estimate the expected number of deaths in such settings, and we illustrate the approach using data from a cohort study of mortality among underground miners. [ABSTRACT FROM AUTHOR]

In this paper, we examine patterns of self-reported diagnosis of noncommunicable diseases (NCDs) and prevalences of algorithm/measured test-based, undiagnosed, and untreated NCDs in China, Ghana, India, Mexico, Russia, and South Africa. Nationally representative samples of older adults aged ≥50 years were analyzed from wave 1 of the World Health Organization's Study on Global Ageing and Adult Health (2007-2010; n = 34,149). Analyses focused on 6 conditions: angina, arthritis, asthma, chronic lung disease, depression, and hypertension. Outcomes for these NCDs were: 1) self-reported disease, 2) algorithm/measured test-based disease, 3) undiagnosed disease, and 4) untreated disease. Algorithm/measured test-based prevalence of NCDs was much higher than self-reported prevalence in all 6 countries, indicating underestimation of NCD prevalence in low- and middle-income countries. Undiagnosed prevalence of NCDs was highest for hypertension, ranging from 19.7% (95% confidence interval (CI): 18.1, 21.3) in India to 49.6% (95% CI: 46.2, 53.0) in South Africa. The proportion untreated among all diseases was highest for depression, ranging from 69.5% (95% CI: 57.1, 81.9) in South Africa to 93.2% (95% CI: 90.1, 95.7) in India. Higher levels of education and wealth significantly reduced the odds of an undiagnosed condition and untreated morbidity. A high prevalence of undiagnosed NCDs and an even higher proportion of untreated NCDs highlights the inadequacies in diagnosis and management of NCDs in local health-care systems. [ABSTRACT FROM AUTHOR]

Distributed networks of health-care data sources are increasingly being utilized to conduct pharmacoepidemio-logic database studies. Such networks may contain data that are not physically pooled but instead are distributed horizontally (separate patients within each data source) or vertically (separate measures within each data source) in order to preserve patient privacy. While multivariable methods for the analysis of horizontally distributed data are frequently employed, few practical approaches have been put forth to deal with vertically distributed healthcare databases. In this paper, we propose 2 propensity score-based approaches to vertically distributed data analysis and test their performance using 5 example studies. We found that these approaches produced point estimates close to what could be achieved without partitioning. We further found a performance benefit (i.e., lower mean squared error) for sequentially passing a propensity score through each data domain (called the "sequential approach") as compared with fitting separate domain-specific propensity scores (called the "parallel approach"). These results were validated in a small simulation study. This proof-of-concept study suggests a new multivariable analysis approach to vertically distributed health-care databases that is practical, preserves patient privacy, and warrants further investigation for use in clinical research applications that rely on health-care databases. [ABSTRACT FROM AUTHOR]

In cohort mortality studies, there often is interest in associations between an exposure of primary interest and mortality due to a range of different causes. A standard approach to such analyses involves fitting a separate regression model for each type of outcome. However, the statistical precision of some estimated associations may be poor because of sparse data. In this paper, we describe a hierarchical regression model for estimation of parameters describing outcome-specific relative rate functions and associated credible intervals. The proposed model uses background stratification to provide flexible control for the outcome-specific associations of potential confounders, and it employs a hierarchical “shrinkage” approach to stabilize estimates of an exposure’s associations with mortality due to different causes of death. The approach is illustrated in analyses of cancer mortality in 2 cohorts: a cohort of dioxin-exposed US chemical workers and a cohort of radiation-exposed Japanese atomic bomb survivors. Compared with standard regression estimates of associations, hierarchical regression yielded estimates with improved precision that tended to have less extreme values. The hierarchical regression approach also allowed the fitting of models with effect-measure modification. The proposed hierarchical approach can yield estimates of association that are more precise than conventional estimates when one wishes to estimate associations with multiple outcomes. [ABSTRACT FROM AUTHOR]

Increasingly often in epidemiologic research, associations between survival time and predictors of interest are measured by differences between distribution functions rather than hazard functions. For example, differences in percentiles of survival time, expressed in absolute time units (e.g., weeks), may complement the popular risk ratios, which are unitless measures. When analyzing time to an event of interest (e.g., death) in prospective cohort studies, the time scale can be set to start at birth or at study entry. The advantages of one time origin over the other have been thoroughly explored for the estimation of risks but not for the estimation of survival percentiles. In this paper, we analyze the use of different time scales in the estimation of survival percentiles with Laplace regression. Using this regression method, investigators can estimate percentiles of survival time over levels of an exposure of interest while adjusting for potential confounders. Our findings may help to improve modeling strategies and ease interpretation in the estimation of survival percentiles in prospective cohort studies. [ABSTRACT FROM AUTHOR]

Although the impact of interviewers on survey measurement has been studied for more than 85 years, such impacts are rarely considered in the analysis of health surveys. This issue is particularly important with single-stage surveys such as that used in the Behavioral Risk Factor Surveillance System(BRFSS),where there are no sampling clusters in which interviewer effects could be captured. The BRFSS involves an ongoing telephone survey of the health behaviors of US adults and was established in 1984 by the Centers for Disease Control and Prevention. Public-use BRFSS data files are widely used by epidemiologists and public health researchers to describe the health behaviors of adults in the United States. Since its onset, the BRFSS has provided identification codes for telephone interviewers completing BRFSS interviews in its public-use data files; however, a review of BRFSS publications shows no evidence that these codes have been used in estimating standard errors. In this paper we analyze data from the 2012 BRFSS, illustrate both design-based and model-based approaches to incorporating interviewer effects in variance estimation, and find evidence of substantial interviewer effects for 5 key estimates across states. These results suggest that BRFSS analysts should consider accounting for interviewer effects, and we provide example code enabling analysts to do so. We conclude with suggestions regarding possible directions for future research. [ABSTRACT FROM AUTHOR]

As pellagra reached epidemic proportions in the United States in the early 20th century, 2 teams of investigators assessed its incidence in cotton mill villages in South Carolina. The first, the Thompson-McFadden Commission, concluded that pellagra was likely infectious. The second, a Public Health Service investigation led by Joseph Goldberger, concluded that pellagra was caused by a dietary deficiency. In this paper, we recount the history of the 2 investigations and consider how the differences between the 2 studies' designs, measurements, analyses, and interpretations led to different conclusions. Because the novel dietary assessment strategy was a key feature of the Public Health Service's study design, we incorporated simulated measurement error in a reanalysis of the Public Health Service's data to assess whether this specific difference affected the divergent conclusions. [ABSTRACT FROM PUBLISHER]

Ozone is associated with adverse health; however, less is known about vulnerable/sensitive populations, which we refer to as sensitive populations. We systematically reviewed epidemiologic evidence (1988-2013) regarding sensitivity to mortality or hospital admission from short-term ozone exposure. We performed meta-analysis for overall associations by age and sex; assessed publication bias; and qualitatively assessed sensitivity to socioeconomic indicators, race/ethnicity, and air conditioning. The search identified 2,091 unique papers, with 167 meeting inclusion criteria (73 on mortality and 96 on hospitalizations and emergency department visits, including 2 examining both mortality and hospitalizations). The strongest evidence for ozone sensitivity was forage. Per 10-parts per billion increase in daily 8-hour ozone concentration, mortality risk for younger persons, at 0.60% (95% confidence interval (Cl): 0.40, 0.80), was statistically lower than that for older persons, at 1.27% (95% Cl: 0.76,1.78). Findings adjusted for publication bias were similar. Limited/suggestive evidence was found for higher associations among women; mortality risks were 0.39% (95% Cl: -0.22,1.00) higher than those for men. We identified strong evidence for higher associations with unemployment or lower occupational status and weak evidence of sensitivity for racial/ethnic minorities and persons with low education, in poverty, or without central air conditioning. Findings show that some populations, especially the elderly, are particularly sensitive to short-term ozone exposure. [ABSTRACT FROM AUTHOR]

Skin color has been proposed to contribute to race-based health disparities in the United States because of differences in ultraviolet (UV) light–induced vitamin D synthesis. The prediction of this hypothesis, herein named the UVD hypothesis, is that racial disparities in health outcomes are correlated with UV light availability. This paper investigates whether UV light availability is associated with disparities in the rates of low birth weight (LBW) and preterm birth (PTB) between whites and blacks, because these outcomes are thought to be influenced by vitamin D status and to shape disease risk in later life. Data on LBW and PTB from 2007 (n = 2,825,620 births) were compared with data on UV light exposure across the United States. Contrary to the predictions of the UVD hypothesis, LBW and PTB rate disparities were greatest in states with the highest UV light exposure. Notably, income inequality was positively and significantly related to LBW and PTB disparities, even after controlling for UV light availability. The results of this analysis demonstrate that there is a significant environmental gradient in racial disparities in birth outcomes in the United States, but other social or environmental factors associated with living in the southern United States are likely stronger contributors to disparities in birth outcomes than UV light–induced vitamin D status. [ABSTRACT FROM PUBLISHER]