Epidemiology最新文献

We discuss generalizability analyses under a partially nested trial design, where part of the trial is nested within a cohort of trial-eligible individuals, while the rest of the trial is not nested. This design arises, for example, when only some centers participating in a trial are able to collect data on non-randomized individuals, or when data on non-randomized individuals cannot be collected for the full duration of the trial. Our work is motivated by the Necrotizing Enterocolitis Surgery Trial, which compared initial laparotomy versus peritoneal drain for infants with necrotizing enterocolitis or spontaneous intestinal perforation. During the first phase of the study, data were collected from randomized individuals as well as consenting non-randomized individuals; during the second phase of the study, however, data were only collected from randomized individuals, resulting in a partially nested trial design. We propose methods for generalizability analyses with partially nested trial designs. We describe identification conditions and propose estimators for causal estimands in the target population of all trial-eligible individuals, both randomized and non-randomized, in the part of the data where the trial is nested while using trial information spanning both parts. We evaluate the estimators in a simulation study and provide an illustration using the Necrotizing Enterocolitis Surgery Trial study.

Background: Data on lactation status for research are often collected through surveys. Information on human milk feeding collected at routine healthcare visits and stored in electronic health records (EHR) is an emerging source of data for lactation research. We compared information on milk feeding obtained from structured EHR fields with survey data.

Methods: We included participants from five US healthcare systems in the Managing Our Mood survey. Individuals had a live birth (March 2022-October 2023), depression diagnosis during pregnancy, and ≥1 record of human milk feeding information in their or their infant's EHR. We compared information from EHR data up to ten months after delivery with survey data collected 3-4 months after delivery as the reference. We assessed agreement on lactation status (human milk feeding ever and at survey) using percent agreement, Cohen's kappa, sensitivity, specificity, positive predictive value, and negative predictive value overall and by characteristics.

Results: According to survey data, the prevalence of human milk feeding ever was 93.2% and was 73.0% at the time of survey among 281 eligible individuals. Agreement between data sources for ever and for human milk feeding at the survey was ≥92% with kappas ≥0.77. EHR and survey data agreed on human milk feeding ever for 97.3% (95% confidence interval: 94.6%, 98.7%) and on human milk feeding at the time of the survey for 98.0% (95% confidence interval: 95.1%, 99.2%) of those who reported yes to these practices on the survey. These measurements were lower among individuals with fewer records.

Conclusions: There was substantial agreement on lactation status between EHR and survey data. These findings suggest that lactation information from structured EHR may be used for epidemiologic research.

Background: School closures during the coronavirus disease 2019 pandemic disrupted children's education, socialization, and access to mental health resources, raising concerns about long-term effects on childrens' mental health. The objective was to evaluate the impacts of pandemic-era school reopenings on children's mental health and healthcare expenditures. Variation in the timing of school reopenings created a unique quasi-experiment.

Methods: We used difference-in-differences analysis to examine how staggered implementation of school reopenings affected diagnoses with depression, anxiety, and attention-deficit/hyperactivity disorder, and related healthcare spending among school-aged children during March 2020-June 2021 across 24 California counties. Data were drawn from medical claims from the second largest private health insurer in the state (N = 185,735).

Results: School reopening was associated with a 1.2%-point drop in monthly prevalence of mental health diagnoses [95% confidence interval (CI) = -1.59, -0.74], and a 10.6% (95% CI = -13.4%, -7.8%) drop in related healthcare spending. The mental health conditions that saw the largest differential between in-person and remote school districts were anxiety and depression. Effects were strongest among girls.

Conclusions: In-person learning is an important component of children's mental health. These results are informative for future policymaking during public health crises, to balance infection risk with the need for socialization and other critical resources that schools provide to children.

Background: Drivers of persistent racial-ethnic inequities in severe maternal morbidity are poorly understood. This study examined how neighborhood-level structural racism measures shape risk of severe maternal morbidity among Black mothers.

Methods: Data are from live hospital births in California between 1997 and 2019 at ≥20 weeks' gestation (N = 555,511). We leveraged information from the U.S. Census Bureau and the American Community Survey to determine neighborhood (census tract) measures of structural racism across six domains (homeownership, unemployment, poverty, educational attainment, racialized economic deprivation, and racial residential segregation). We used (1) an additive composite index (quartile 1 [low]-quartile 4 [high]) and (2) latent class analysis to characterize four structural racism typologies. We examined associations across both measurement approaches using mixed-effects logistic regression models with neighborhood random intercepts, adjusting for maternal age, education, and hospital payer/insurance information.

Results: Black mothers living in neighborhoods scoring high (quartile 4) on the additive composite index had 13% higher risk of severe maternal morbidity than those in neighborhoods scoring low (quartile 1) (95% confidence interval = 1.04, 1.24). Models evaluating latent class typologies also revealed that Black mothers living in neighborhoods characterized by consistently high racial inequity in unemployment, racialized economic deprivation, and racial residential segregation across the study period had a 12% higher risk of severe maternal morbidity compared with those in neighborhoods consistently scoring low in the domains examined (95% confidence interval = 1.03, 1.23).

Conclusions: Our findings support the hypothesis that neighborhood-level measures of structural racism influence the risk of severe maternal morbidity among Black mothers.

Background: Between 1970 and 1991, when viral inhibition reduced the risk, people with a bleeding disorder in the United Kingdom had their missing clotting factors replaced with plasma products derived from donated plasma at risk of infection. We analyzed longer-term survival of people with bleeding disorders exposed to plasma products.

Methods: The National Haemophilia Database documents people with bleeding disorders registered, treated before, and alive on 1 January 1992. We estimated all-cause mortality proportional hazard ratios for exposure groups (Human immunodeficiency virus (HIV)/Hepatitis C virus (HCV) coinfected, HCV-diagnosed, and HCV-status unknown) versus HCV antibody negative, within distinct epochs: 1992-1999; 2000-2009; 2010-2019. We estimated years of life lost by epoch and exposure group versus UK general population lifetables or via parametric survival models compared with people with bleeding disorders negative or unknown for HCV antibodies. Models were adjusted for sex, age band at 1 January 1992, bleeding disorder, and severity.

Results: Of 6282 people with bleeding disorders who met inclusion criteria, 15% were HIV/HCV coinfected, 32% HCV antibody positive, and 28% HCV antibody negative. Compared with HCV-negative, those HIV/HCV coinfected had an all-cause mortality hazard ratio of 4.2 (95% confidence interval: 2.9, 6.0) and HCV+ of 2.2 (1.7, 2.8) in 2010 to 2019. Years of life lost for 2014 to 2019 were 740 (95% confidence interval: 440, 1030) for HCV+ persons and 270 (130, 400) for HIV/HCV coinfected persons, compared with HCV unknown or negative persons.

Conclusions: People with bleeding disorders in the United Kingdom infected before, but alive at, 1 January 1992, were still at increased risk of death 3 decades postimplementation of HCV screening of blood supplies.

Causal inference is the goal of randomized trials and many observational studies. The first step in a formal causal inference framework is to define the causal estimand, and in both types of study this can be done mathematically or, equivalently, by specifying an ideal trial: a hypothetical perfect randomized experiment (with representative sample, perfect adherence, etc). The target trial framework is increasingly used for causal inference in observational studies, but clarity is lacking in how a target trial should be specified and how it relates to an ideal trial. Here, we review the mathematical and ideal trial approaches to defining a causal estimand, highlighting their equivalence and the need to balance practical relevance and feasibility of estimation regardless of approach. We then consider the question of how a target trial should be specified, outlining the challenges of a recommended approach, commonly seen in applications, that puts the focus heavily on the feasibility of estimation: to specify the target trial such that it is closely aligned with the observational data (e.g., uses the same eligibility criteria). We argue that with this "aligned" approach, biases may remain relative to the estimand of ultimate practical interest, defined by the ideal trial, that mirror the often-overlooked biases of actual trials. We conclude that consideration of the ideal trial and of how the target trial and its emulation or the actual trial differ from it is necessary to identify and manage all bias sources in both settings. An example from respiratory epidemiology is used for illustration.

Background: Projections for major health outcomes are crucial for decision-making to enable early interventions. Pooling results from meta-analyses with estimates based on individual participant data can reduce uncertainty and improve generalizability. However, current methods for meta-meta-analyses of nonlinear functions have remained underdeveloped.

Methods: We proposed a novel meta-analysis method to pool pointwise nonlinear function literature estimates with parameter estimates, applied here to the association of dietary change scenarios with mortality and ischemic heart disease (IHD) based on Finnish individual participant data. We linked individual-level demographic and risk factor data from the Health 2000, FINRISK 2007, FINRISK 2012, and FinHealth 2017 surveys (n = 20,784) with follow-up data on outcomes obtained from national health registers. We applied a Poisson multistate model and microsimulation to project state probabilities.

Results: Pooling reduced uncertainty in the hazard ratio estimates and in the health projections. We estimated that a two-thirds reduction in red and processed meat consumption would decrease the prevalence of IHD by 2 (95% prediction intervals [PI] 1, 4) percentage points (%pt) in the 2017 cohort and deaths by 2%pt (95% PI 1, 4) by 2050. We estimated that a 100% increase in whole grain consumption would reduce IHD by 2%pt (95% PI 0, 3) and deaths by 2%pt (95% PI 0, 3).

Conclusions: Our flexible meta-analysis method allowed the pooling of nonlinear estimates reported in the literature without detailed technical information. Our estimates support the hypothesis that more plant-based diets reduce mortality and IHD prevalence. The most beneficial scenarios included reductions in red and processed meat and increments in whole grain consumption.

Background: Randomized trials of major adverse cardiovascular events found no effect of dipeptidyl-peptidase-4 inhibitors (DPP-4i) medications compared with second-generation sulfonylureas, while nonrandomized studies estimated a benefit of DPP-4i. Socioeconomic residual confounding was thought to be implicated. We compared area-level prescribing density and physician prescribing preference as candidate instrumental variables for the effect of DPP-4i medications on major adverse cardiovascular events.

Methods: Using Medicare claims data, we built two cohorts emulating randomized trials of sitagliptin or saxagliptin starters, each compared with sulfonylurea starters. The proportion of DPP-4i prescribing in a ZIP Code tabulation area defined the area-level prescribing density instrumental variable at various cutoffs (0% vs. 100% to <50% vs. ≥50%). Patients' physician prescribing history using the same proportion cutoffs was the physician prescribing preference candidate instrumental variable. An instantaneous physician preference instrumental variable used a physician's most recent prescription. We adjusted two-stage instrumental variable regression models for propensity score quintiles.

Results: Unadjusted analyses for sitagliptin and saxagliptin, each compared with sulfonylurea, estimated a reduced risk of major adverse cardiovascular events [sitagliptin hazard ratio (HR) = 0.86; 95% confidence interval = 0.83, 0.88]; saxagliptin (HR = 0.68; 0.64, 0.73). All instrumental variables were strong and reduced covariate imbalance. Analyses of area-level prescribing density found no meaningful difference for sitagliptin (0% vs. 100% HR = 1.1; 0.79, 1.6). Analyses of physician prescribing preference estimated reduced risk for sitagliptin (<50% vs. ≥50% HR = 0.69; 0.48, 0.98). Instantaneous physician prescribing preference analyses showed little to no difference for sitagliptin (HR = 0.86; 0.60, 1.1) and saxagliptin (HR = 0.98; 0.56, 1.7).

Conclusions: Candidate instrumental variables focusing on short-term prescribing preference hold promise over area-based variables but remain inefficient.

Test-negative designs (TNDs) are widely used for postmarket evaluation of vaccine effectiveness (VE), particularly in cases when randomized trials are not feasible. Unlike classical TNDs, which only include healthcare seekers with symptoms, recent TNDs have involved individuals with various reasons for testing, especially in an outbreak setting. While including these data can increase sample size and hence improve precision, concerns have been raised about whether they introduce bias into the current framework of TNDs, thereby demanding a formal statistical examination of this modified design. In this article, using statistical derivations, causal graphs, and numerical demonstrations, we show that the standard odds ratio estimator may be biased if various reasons for testing are not taken into account. To eliminate this bias, we identify three categories of reasons for testing, namely symptoms, mandatory screening, and case contact tracing, and characterize associated statistical properties and estimands. Based on our characterization, we show how to consistently estimate each estimand via stratification. Furthermore, we describe when these estimands correspond to the same VE parameter and, when appropriate, propose a stratified estimator that can incorporate multiple reasons for testing and improve precision. We demonstrate the performance of our proposed method through simulation studies and a real-data analysis.