Epidemiology最新文献

Background: Despite being the leading cause of death, the global tuberculosis (TB) burden is ill-defined. Existing methods to estimate incidence are time and/or resource-intensive and often inaccurate. Back-calculation was developed to estimate HIV incidence by considering reported cases to be a convolution of the disease duration and the incidence of new cases. New estimates of TB natural history parameters allow us to develop Bayesian back-calculation methods for TB to assign case notification data to the time point of onset of disease.

Methods: Recorded counts of TB cases are underestimates of the true burden of disease, so we include a multiplier derived from prevalence to notification ratios to account for underreporting. We assume a Poisson distribution for notifications and incidence and use a penalized-likelihood before smooth estimates. We estimate sex-stratified TB incidence for Vietnam, Cambodia, and the Philippines via Markov chain Monte Carlo.

Results: Annual estimated TB incidence was, on average 19% greater than recorded notifications. TB incidence among males was on average 3.8% higher than females in Vietnam, 1.3% in Cambodia, and 2.5% higher in the Philippines.

Conclusions: These estimates account for the delay between bacteriologically positive subclinical disease and notification and, as such, may be more temporally accurate than existing methods.

Epidemiologic studies are often concerned with estimating causal effects of a sequence of treatment decisions on survival outcomes. In many settings, treatment decisions do not occur at fixed, prespecified follow-up times. Rather, timing varies across subjects in ways that may be informative of subsequent treatment decisions and potential outcomes. Awareness of the issue and potential solutions is lacking in the literature, which motivates this work. Here, we formalize the issue of informative timing, problems associated with ignoring it, and show how g-methods can be used to analyze sequential treatments that are informatively timed. As we describe, in such settings, the waiting times between successive treatment decisions may be properly viewed as time-varying confounders. Using synthetic examples, we illustrate how g-methods that do not adjust for these waiting times may be biased and how adjustment can be done in scenarios where patients may die or be censored in between treatments. Finally, we provide implementation guidance and examples using publicly available software. Our concluding message is that (1) considering timing is important for valid inference and (2) correcting for informative timing can be done with g-methods that adjust for waiting times between treatments as time-varying confounders.

Background: Structural racism can manifest in contemporary neighborhoods via historical policies or programs. For example, the Home Owners' Loan Corporation, a government-backed program from the 1930s, systematically diverted wealth away from Black neighborhoods. The reproductive health consequences of this racist program remain understudied. We evaluated associations between residence in a historically redlined neighborhood and fecundability, the per-cycle probability of conception.

Methods: We used data from the Black Women's Health Study, a US cohort of Black women who were aged 21-69 years in 1995 and were followed up with biannual questionnaires. Experiences of infertility (i.e., tried for ≥12 months to conceive without success) were captured on several questionnaires. A 2011 supplemental module collected pregnancy histories between 1995 and 2011, including planning status and time to conception. We linked geocoded addresses to historical Home Owners' Loan Corporation grades (A ["best"] to D ["hazardous," i.e., redlined]). Using proportional probabilities regression models with generalized estimating equations, we estimated fecundability ratios and 95% confidence intervals (CIs).

Results: Our analysis included 818 planned pregnancy attempts from 674 participants (mean age = 33.9 years). Relative to participants residing in neighborhoods with the highest grades (A or B), adjusted models showed reduced fecundability among participants who resided in lower graded neighborhoods (C: 0.91, 95% CI: 0.77, 1.09; D: 0.82, 95% CI: 0.68, 0.99).

Conclusions: In this cohort of US Black women, contemporary residence in a historically redlined neighborhood was associated with reduced fecundability. Our findings highlight the importance of exploring how historical neighborhood disinvestment affects reproductive health.

Background: While the Supplemental Nutrition Assistance Program (SNAP) aims to reduce food insecurity among low-income households, nearly half of recipients remain food insecure. Increasing state minimum wages could help improve food security, but because SNAP benefits are income-dependent, net effects are unclear.

Methods: Using the U.S. Current Population Survey Food Security Supplement (2002-2019), we linked households interviewed in two consecutive Decembers to create a two-year panel. The primary sample included SNAP recipient households with at least one adult working in year 1. The exposure was the average effective state minimum wage (2019 dollars), for each state and year. We estimated prevalence differences (PD) in food insecurity per $1 increase in minimum wage using a within-household linear fixed-effects model adjusting for time-varying economic confounders and concurrent safety-net policies. We investigated variation in the effect by household and family structure, race/ethnicity, and educational attainment using stratified models.

Results: Overall estimates were most compatible with protective effects (PD per 10,000 households: -298, 95% CI: -673, 77). The strongest protective estimates were for senior-headed (-1,472, 95% CI: -2,869, -76), Hispanic (-865, 95% CI: -1,638, -92), and some-college households (-988, 95% CI: -1,664, -312). Estimates for Indigenous households were imprecise and possibly harmful (900, 95% CI: -736, 2,537). Most other subgroup estimates were near zero.

Conclusions: Increased minimum wages may modestly support food security for working SNAP households. As SNAP benefit rules evolve, these findings suggest that minimum-wage policies can complement and reinforce the program's goals to protect low-income households from food hardship.

Background: Algorithms to estimate gestational age (GA) for non-live births have been developed in other administrative healthcare databases, but their applicability in Korea remains unknown. We adapted algorithms developed in the United States and evaluated their validity in Korean healthcare claims data.

Methods: Using the National Health Information Database (NHID) of South Korea, we linked GA information at influenza vaccination from the national vaccination registry to establish a reference standard. Non-live births were stratified into spontaneous/induced abortions and stillbirths. Four algorithms were tested: (1) assigning outcome-specific GA, (2) adjusting GA based on ultrasound scan records, (3) a regression model using gestational markers as predictors, and (4) a random forest model. Algorithms were evaluated by the proportion of estimates within 1-4 weeks of the reference standard and the mean squared error (MSE). External validation was conducted using an independent dataset.

Results: Random forests performed best for both spontaneous/induced abortions (MSE 1.68 weeks²) and stillbirths (MSE 0.97 weeks²), with 92.6% (95% CI 91.6-93.4) and 97.4% (96.2-98.3) of predictions falling within two weeks of the reference standard, respectively. However, in the external validation set, the ultrasound record-based adjustment approach performed similarly to the random forest approach for both spontaneous/induced abortions (MSE 8.37 vs. 8.15 weeks²) and stillbirths (MSE 12.42 vs. 12.52 weeks²).

Conclusions: Deterministic approaches may be preferable for estimating GA of non-live births in the NHID, as they are simpler to implement and perform comparably to model-based algorithms. These algorithms can support pregnancy research in the Korean population.

Background: Longitudinal data often include gaps in observation when outcomes (and other variables) are unmeasured, due to missed study visits or drop out. We explore the fundamentals of data gaps and use simulation to compare approaches for handling data gaps when estimating outcome incidence.

Methods: We generated a simulation of 1000 individuals across 10 study visits. We used 4 data generating mechanisms: (1) missingness was independent of the outcome; (2) there was a baseline common cause of missingness and the outcome; (3) there was a time-varying common cause; and (4) the outcome directly affected future missingness. We estimated the risk and rate of the first outcome occurrence (generated as a transient; repeated; and permanent outcome), using crude and adjusted approaches, across 1000 iterations and compared bias and empirical standard error.

Results: Under Scenario 1, in crude analyses, results were unbiased when censoring prior to a data gap but not when allowing participants to return. Under Scenarios 2-4, all crude approaches were biased. Inverse probability of censoring weights and multiple imputation were relatively unbiased across scenarios and outcome types; multiple imputation was more precise. Inverse probability of observation weights were biased when the outcome was permanent and were less precise than either of the other two approaches.

Conclusions: Crude approaches allowing participants to return following a data gap are not recommended because they can be biased even when missingness and the outcome are independent. Instead, one should either censor or handle the data gap using multiple imputation.

Substantial observational evidence exists in support of bariatric surgery being associated with reduction in risk for a wide range of outcomes, including cardiovascular disease (CVD) in patients with diabetes. Two recent studies, however, argued that much of that prior work suffers from various sources of underappreciated bias as well as design decisions that compromise whether one can conceive of a corresponding target trial. Furthermore, results based on analyses of claims data from Optum and electronic health record data from the Veterans Administration (VA) are presented as providing evidence of no CVD benefit for bariatric surgery in patients with diabetes. In this paper, we use data from a prior Kaiser Permanente study to emulate a trial that mimics the methods employed in the VA study. This new analysis finds a reduction in risk of CVD in patients with diabetes, consistent with pre-existing evidence. We discuss possible mechanisms by which the discrepant results can be reconciled, including issues of statistical validity that arise from small samples, whether recent work on transportability indicates that we should not always expect results to always be concordant and the role of conservatism associated with "clinical trial thinking". We conclude with a discussion of what standards should be used when considering the work of others in the literature and the role that evidence triangulation may be play in the future.

During the COVID-19 pandemic, estimating the total deaths averted by vaccination was of great public health interest. Instead of estimating total deaths averted by vaccination among both vaccinated and unvaccinated individuals, some studies empirically estimated only "directly averted" deaths among vaccinated individuals, typically suggesting that vaccines prevented more deaths among unvaccinated and vaccinated individuals than directly among vaccinated individuals only, due to the indirect effect. Here, we define the causal estimand to quantify outcomes "directly averted" by vaccination-i.e., the impact of vaccination for vaccinated individuals, holding vaccination coverage fixed-for vaccination at multiple time points, which is a lower bound on the total outcomes averted when the indirect effect is non-negative. We develop an unbiased estimator for the causal estimand in a one-stage randomized controlled trial (RCT) and explore the bias of a popular "hazard difference" estimator frequently used in empirical studies. We show that even in an RCT, the hazard difference estimator is biased if vaccination has a non-null effect, as it fails to incorporate the greater depletion of susceptibles among the unvaccinated individuals. In simulations, the overestimation is small for averted deaths when infection-fatality rate is low, as for many important pathogens. However, the overestimation can be large for averted infections given a high basic reproduction number and a high vaccine efficacy against infection. Additionally, we define and compare estimand and estimators for avertible outcomes (i.e., outcomes that could have been averted by vaccination, but were not due to failure to vaccinate). Future studies can explore the identifiability of the causal estimand in observational settings.

Objective: To compare reporting of severe maternal morbidity (SMM) in birth certificate versus hospital claims data and evaluate differences by patient race/ethnicity.

Methods: We compared incidence rates of blood transfusion, hysterectomy, intensive care unit admission, uterine rupture, and 3rd/4th degree perineal laceration between 2019 deliveries in the US birth certificate and the Premier Healthcare Database, overall and stratifying by maternal race/ethnicity. We then computed incidence rate ratios (IRRs) computed between datasets, and fit logistic regression models of race/ethnicity on SMM.

Results: Comparing 3,467,934 birth certificate deliveries with 3,450,569 Premier deliveries (n=905,766 pre-weighting for national representativeness), incidence rates of SMMs were lower in birth certificate compared with Premier data, and these rate differentials varied by maternal race/ethnicity. For example, among non-Hispanic white patients, the incidence rate of blood transfusions in birth certificate data was 50% that of the incidence rate in the Premier claims dataset (IRR: 0.50, 95% CI: 0.47, 0.52). Among all other races/ethnicities, the incidence rate of blood transfusions was even lower relative to the claims data (IRR range: 0.29-0.39). Adjusted odds ratios (aOR) for SMM in non-Hispanic Black and Hispanic patients versus non-Hispanic white patients were closer to the null in birth certificate than Premier data (e.g., compared with non-Hispanic white patients, non-Hispanic Black patients had a 16% higher adjusted odds in the birth certificate (95% CI: 1.10, 1.21) data versus an 84% higher adjusted odds of blood transfusion in Premier data (95% CI: 1.79, 1.89)).

Conclusions: Birth certificates report substantially less SMM than claims data, with greater differential in reporting for non-Hispanic Black and Hispanic patients that may bias birth certificate-based research findings.

Background: Linking cancer cohort participants to state cancer registries typically relies on personally identifiable information, including social security numbers (SSN), which uniquely identify individuals. However, complete SSN collection can be limited due to privacy concerns. This study evaluates the sensitivity of cancer registry linkage using partial or missing SSN and examines differences by demographic characteristics.

Methods: Using data from 284,361 participants in the Cancer Prevention Study-3, we conducted probabilistic linkages with cancer registries in Georgia, Ohio, and Texas using Match*Pro software. Participants were linked using combinations of personally identifiable information: complete SSN, partial SSN (last four digits), and missing SSN. We compared the sensitivity of linkages before and after manual review and stratified by sex, age, and race-ethnicity.

Results: Before manual review, the sensitivity for missing and partial SSNs was 92.5%. Sensitivity improved to 98.6% for missing SSN and 98.8% for partial SSN after manual review. We observed no notable heterogeneity by sex, age, or race-ethnicity, with sensitivity exceeding 87% across all subgroups. Manual review substantially reduced uncertain matches, contributing to high linkage accuracy.

Conclusion: This study demonstrates that high sensitivity in cancer registry linkage can be achieved without a complete SSN, provided other personally identifiable information (e.g., name, date of birth, longitudinal address) is available. These findings support the feasibility of accurate cancer case identification in cohorts with limited SSN data, particularly for historically marginalized populations, and underscore the importance of designing inclusive population-based cancer studies.