Epidemiology最新文献

Causal mediation analysis aims to disentangle the pathways through which an exposure influences an outcome. In the presence of mediator-outcome confounders affected by exposure (intermediate confounders), the natural indirect effect (NIE) is not identifiable under nonparametric structural equation models (SEMs) with independent errors. To address this challenge, we focus on the indirect pathway and introduce a novel class of indirect effect measures, referred to as generalized NIEs, of which the NIE is a special case. In particular, we introduce a case of generalized NIE defined through a randomized intervention, which, under the nonparametric SEM with independent errors, coincides with the interventional indirect effect (IIE)-even though identifying the IIE generally does not rely on the cross-world assumptions implied by nonparametric SEM with independent errors. Furthermore, when an additional no-heterogeneity assumption is imposed, the NIE becomes equal to this generalized NIE and hence is identifiable. Unlike prior approaches, we propose new indirect effect measures criteria that ensure valid mediation interpretation even in the presence of intermediate confounders. Under traditional identification assumptions alone, the IIE fails to satisfy these criteria. In contrast, all proposed generalized NIEs meet them, providing a wide range of options beyond the existing measures. Our findings highlight the generalized NIEs as a more pragmatic and reasonable alternative in settings where intermediate confounders are inevitable.

Background: We assessed the safety and effectiveness of the first- and second-dose BNT162b2 COVID-19 vaccination, offered as part of the national COVID-19 vaccine roll-out from September 2021, in children and adolescents in England.

Methods: Our observational study using OpenSAFELY-TPP, included adolescents aged 12-15 years and children aged 5-11 years. It compared individuals receiving (1) the first vaccination to unvaccinated controls and (2) the second vaccination to single-vaccinated controls. We matched vaccinated individuals with controls on age, sex, and other important characteristics. Outcomes were positive SARS-CoV-2 test (adolescents only), COVID-19 accident and emergency (A&E) attendance, COVID-19 hospitalization, COVID-19 critical care admission, and COVID-19 death; with safety outcomes, A&E attendance, unplanned hospitalization, pericarditis, and myocarditis.

Results: Among 820,926 previously unvaccinated adolescents, 20-week incidence rate ratios (IRRs) comparing vaccination with no vaccination were 0.74 for positive SARS-CoV-2 test, 0.60 for COVID-19 A&E attendance, and 0.58 for COVID-19 hospitalization. Among 441,858 adolescents who had received the first vaccination, IRRs comparing second dose with single-vaccination were 0.67 for positive SARS-CoV-2 test, 1.00 for COVID-19 A&E attendance, and 0.60 for COVID-19 hospitalization. In both children groups, COVID-19-related outcomes were too rare to allow IRRs to be estimated precisely. Across all analyses, there were no COVID-19-related deaths, and fewer than seven COVID-19-related critical care admissions. Myocarditis and pericarditis were documented only in the vaccinated groups, with rates of 27 and 10 cases/million after the first and second doses, respectively.

Conclusions: BNT162b2 vaccination in adolescents reduced COVID-19 A&E attendance and hospitalization, although these outcomes were rare. Protection against positive SARS-CoV-2 tests was transient.

There are concerns about the fairness of clinical prediction models. 'Fair' models are defined as those for which their performance or predictions are not inappropriately influenced by protected attributes such as ethnicity, gender, or socio-economic status. Researchers have raised concerns that current algorithmic fairness paradigms enforce strict egalitarianism in healthcare, leveling down the performance of models in higher-performing subgroups instead of improving it in lower-performing ones. We propose assessing the fairness of a prediction model by expanding the concept of net benefit, using it to quantify and compare the clinical impact of a model in different subgroups. We use this to explore how a model distributes benefit across a population, its impact on health inequalities, and its role in the achievement of health equity. We show how resource constraints might introduce necessary trade-offs between health equity and other objectives of healthcare systems. We showcase our proposed approach with the development of two clinical prediction models: 1) a prognostic type 2 diabetes model used by clinicians to enrol patients into a preventive care lifestyle intervention programme, and 2) a lung cancer screening algorithm used to allocate diagnostic scans across the population. This approach helps modelers better understand if a model upholds health equity by considering its performance in a clinical and social context.

Background: Pooled logistic regression is a popular tool for survival analyses in epidemiology but can face computational challenges. Commonly these challenges are addressed through widening time intervals or using a parametric functional form for time. We propose a third option to reduce the computational burden without constraining the functional form for time.

Methods: The proposed algorithm operates by restricting the long data set to rows that correspond to unique event times. However, our approach is only compatible when modeling time most flexibly with disjoint indicators. We compared the standard implementation to the proposed implementation in SAS, R, and Python using a publicly available data set.

Results: For the example considered, both implementations provided the same point estimates, but the proposed implementation was between 6 and 68 times faster depending on the software.

Conclusions: The proposed implementation can greatly simplify estimation of pooled logistic regression models, which is especially important when relying on the bootstrap for inference.

Background: Observational studies of the association between antibiotics and preterm delivery report conflicting findings. We assessed the effects of third trimester antibiotic use on preterm delivery and low birthweight, using a study design that accounts for immortal time bias.

Methods: We used the UK's Clinical Practice Research Datalink to identify pregnant females aged 12-50, over the period 2002 to 2016, reaching 27 weeks of gestation without antibiotic use until that point. We applied the prevalent new-user design, matching each third trimester antibiotic initiator with a reference non-user at the same gestational day, using time-conditional propensity scores. The two matched groups were compared on the incidence of preterm delivery and low birthweight. The full cohort was also analyzed with antibiotic use considered as time-fixed and time-varying exposures.

Results: The cohort included 207,027 pregnancies, with 16,865 initiating antibiotics matched to 16,865 non-users. The hazard ratio (HR) of preterm delivery with third trimester antibiotic use was 1.14 (95% CI: 1.04-1.24), compared with non-use. With time-fixed exposure, subject to immortal time bias, the HR was 0.78 (95% CI: 0.73-0.83), while with time-varying exposure, the HR was 1.23 (95% CI: 1.16-1.32). The HR of low birthweight with antibiotic initiation was 1.07 (95% CI: 0.93-1.25) compared with 0.91 (95% CI: 0.83-1.00) under the time-fixed approach.

Conclusion: Using the prevalent new-user design, which emulates a randomized trial, antibiotic use late in pregnancy was associated with a modest increased incidence of preterm delivery. Previous inconclusive studies may have resulted from observational methods that introduced, or insufficiently addressed, immortal time bias.

An impactful epidemiologic question is one that, if answered, could inform meaningful action to reduce the burden of disease in the population it concerns. We propose a set of factors that could be used for discussing, evaluating, and communicating the public health impact of epidemiologic studies. These factors pertain to the burden and distribution of disease, the potential for an intervention to alter the disease burden, and the context in which the study is conducted. The disease burden is characterized by the number of cases, severity or cost of disease, and distribution of disease across the population. The potential for intervention is characterized by the mutability of the exposure itself, the prevalence and distribution of other causes of the disease in the population, the prevalence of the exposure and risk of the outcome under the natural course (prior to any intervention),1 and the feasibility of intervening. An epidemiologic question need not be impactful along all these factors to make answering it worthwhile. However, answering epidemiologic questions with more of these factors present will likely have greater public health impact than answering questions for which these factors are absent. We hope that collecting these factors into a single framework may aid students and senior epidemiologists alike when organizing arguments for the value of their own work or attempting to evaluate the impact of others' work.

Background: Sibling comparison designs are increasingly used to strengthen causal claims about social exposures and health outcomes, yet methodologic challenges in interpreting their results remain insufficiently addressed. This study develops empirical approaches to help assess whether sibling comparison estimates provide reliable evidence for causal relationships.

Methods: We used childhood family income and severe mental disorders in a Danish nationwide cohort (n=643,623; 403,963 siblings born 1986-1996) as an example. We applied three complementary approaches: negative control analyses using pseudo-siblings (unrelated individuals with similar income differences as real siblings) to isolate exposure variability effects from shared familial confounding effects; assessment of sibling age structure, exposure correlation, and variation patterns to establish whether meaningful contrasts exist between siblings; and critical period assumption evaluation through age-specific income measurement.

Results: Family income at age 14 was associated with decreased mental disorder risk in the population-wide analysis (adjusted hazard ratio [aHR]=0.78; 95% CI:0.76-0.81) but showed no association using a sibling comparison design (aHR=1.02; 95% CI:0.94-1.11). The pseudo-sibling cohort matched on income also showed substantial attenuation (aHR=0.93; 95% CI:0.85-1.01), while pseudo-siblings not matched on income showed no attenuation. Income associations were similar across childhood measurement ages 0-14 (aHR range = 0.67-0.82).

Conclusions: In this example, estimates from the sibling comparison design may reflect limited exposure variability within families and unmet life course model assumptions, rather than or in addition to the removal of shared familial confounding. The empirical approaches we developed help researchers distinguish methodologic factors from genuine null findings, and are available with R code for implementation.