Epidemiology
· 2025 May · PMID 39996615
·
Full text
BACKGROUND: Mediation analyses of the preeclampsia-perinatal outcome association through preterm birth (PTB) have produced paradoxical findings. For example, preeclamptic births at preterm gestations show a lower risk of...BACKGROUND: Mediation analyses of the preeclampsia-perinatal outcome association through preterm birth (PTB) have produced paradoxical findings. For example, preeclamptic births at preterm gestations show a lower risk of adverse outcomes than normotensive births. These results have been explained by unmeasured baseline confounding between PTB and outcomes, with PTB as the sole mediator. However, other intermediate variables, such as placental abruption, small for gestational age (SGA) births, and chorioamnionitis, are confounders yet are excluded because they occur after preeclampsia. METHODS: Using data from the Consortium on Safe Labor (2002-2008; ), we utilized interventional indirect effects to examine whether adjusting for causal intermediates mitigates confounding bias to resolve the perinatal paradox. We compared two approaches to handle intermediate confounding by abruption, SGA, and chorioamnionitis when PTB is the focal mediator: as exposure-induced confounders or as multiple mediators. We developed bias formulas to assess unmeasured confounding for interventional effects. RESULTS: When PTB was the sole mediator, the estimated protective direct effect of preeclampsia (risk ratio = 0.60; 95% confidence interval = 0.52, 0.71) was in line with previous paradoxical findings. The estimated protective effect persisted even after adjusting for intermediate confounders. Sensitivity analyses suggested an unmeasured confounder must strongly influence the outcome to resolve the paradox. CONCLUSION: Adjusting for causal intermediates such as abruption, SGA, and chorioamnionitis is inadequate to eliminate unmeasured PTB-perinatal mortality confounding. The paradox of preeclampsia's protective direct effect on mortality remains unresolved. Sensitivity analyses to unmeasured confounding are effective in bolstering conclusions from causal mediation analyses and should be more widely applied.
Geronimus AT, Waidmann TA, Bound J
… +2 more, Pancini V, Yang M
Epidemiology
· 2025 May · PMID 39929188
·
Full text
BACKGROUND: The nature and timing of increasing educational inequity in US life expectancy before the coronavirus disease 2019 pandemic suggest that long-term adverse labor market conditions secondary to globalization an...BACKGROUND: The nature and timing of increasing educational inequity in US life expectancy before the coronavirus disease 2019 pandemic suggest that long-term adverse labor market conditions secondary to globalization and technological change played a role for less-educated workers, but this has not been tested. METHODS: We exploit spatiotemporal variation in mortality and long-term economic conditions at the year and commuting zone level to estimate the relationship between macroeconomic restructuring and diverging mortality trends, 1990-2017, by race, sex, and education. Our measure of macroeconomic restructuring is based on the baseline industrial mix of an area, a measure that is plausibly exogenous to mortality. RESULTS: Mortality trends were substantially worse in commuting zones experiencing long-term economic stagnation than in others. For both White and Black adults, this relationship was strongest in the lowest quartile of the education distribution. Residence in commuting zones in the top quartile of our measure of economic conditions was associated with an additional 1-2 years lived between ages 25 and 84 compared with living in a commuting zone in the bottom quartile. The primary mediators of these divergent mortality trends were cancer, cardiovascular and metabolic diseases, and diseases of other internal body systems. Deaths from suicide or substance abuse did not contribute importantly toward accounting for the estimated impact of long-term economic stagnation on mortality. CONCLUSION: In our study, diverging trends in US life expectancy were associated with macroeconomic changes witnessed over the last half-century. The causes of death mediating this link were largely found in rates of death from stress-related internal diseases.
Epidemiology
· 2025 May · PMID 39929186
·
Publisher ↗
Plant-capture is a specialized variant of traditional capture-recapture methods used to estimate the size of a population. In epidemiologic literature, a notable application of this method is the estimation of the size o...Plant-capture is a specialized variant of traditional capture-recapture methods used to estimate the size of a population. In epidemiologic literature, a notable application of this method is the estimation of the size of homeless populations through point-in-time street surveys. With this approach, decoys referred to as "plants" are introduced into the population to estimate the capture probability. Previous plant-capture studies have not systematically accounted for uncertainty in the capture status of individual plants. To address this, we propose three increasingly complex hierarchical modeling approaches to formally incorporate uncertainty into the plant-capture model arising from the capture status of plants and heterogeneity between survey sites. We then apply our methods to estimate the size of the homeless population in large US cities in the context of the "S-Night" study conducted by the US Census Bureau. Details on the frequentist and Bayesian implementations of our models, along with empirical evaluations of their statistical performance, are provided in the supplementary materials.
Epidemiology
· 2025 May · PMID 39898536
·
Full text
BACKGROUND: Public health and social measures are crucial for controlling the spread of pathogens. However, well-tailored assessments of their impact remain elusive, particularly considering time-varying immunity establi...BACKGROUND: Public health and social measures are crucial for controlling the spread of pathogens. However, well-tailored assessments of their impact remain elusive, particularly considering time-varying immunity established from prior exposures and its waning. METHODS: We developed a mathematical model to estimate the time-varying basic reproduction number, accounting for the dynamics of underlying immunity. Applying this framework, we retrospectively assessed the impact of public health and social measures implemented from November 2021 to April 2022 on SARS-CoV-2 transmission in Korea and discussed potential biases from ignoring underlying immunity. RESULTS: Our proposed model estimated a notable attenuation in the impact of public health on social measures on SARS-CoV-2 transmission in Korea with the emergence of the Omicron variants while remaining effective throughout the Delta and Omicron periods. These changes during the Omicron period became evident only upon adjusting for underlying immunity and were correlated with observed human mobility patterns in Korea. CONCLUSIONS: Our findings support the importance of incorporating underlying immunity in evaluating public health and social measures, particularly in the presence of substantial changes over a short period, such as widespread infections or vaccination. This model would stand as a tool for informing public health planning, capable of mitigating the overall disease burden in future epidemics.
Bohmann P, Stein MJ, Weber A
… +10 more, Konzok J, Fontvieille E, Peruchet-Noray L, Gan Q, Fervers B, Viallon V, Baurecht H, Leitzmann MF, Freisling H, Sedlmeier AM
Epidemiology
· 2025 Mar · PMID 39887119
·
Publisher ↗
BACKGROUND: Individual traditional anthropometric measures such as body mass index and waist circumference may not fully capture the relation of adiposity to mortality. Investigating multitrait body shapes could overcome...BACKGROUND: Individual traditional anthropometric measures such as body mass index and waist circumference may not fully capture the relation of adiposity to mortality. Investigating multitrait body shapes could overcome this limitation, deepening insights into adiposity and mortality. METHODS: Using UK Biobank data from 462,301 adults (40-69 years at baseline: 2006-2010), we derived four body shapes from principal component analysis on body mass index, height, weight, waist and hip circumference, and waist-to-hip ratio. We then used multivariable-adjusted Cox proportional hazard models to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for associations between body shapes and mortality for principal component scores of +1 and -1. RESULTS: During 6,114,399 person-years of follow-up, 28,807 deaths occurred. A generally obese body shape exhibited a U-shaped mortality association. A tall and centrally obese body shape showed increased mortality risk in a dose-response manner (comparing a score of +1 and 0: HR = 1.16, 95% CI = 1.14, 1.18). Conversely, tall and lean or athletic body shapes displayed no increased mortality risks when comparing a score of +1 and 0, with positive relations for the comparison between a score of -1 and 0 in these shapes (short and stout shape: HR = 1.12, 95% CI = 1.10, 1.14; nonathletic shape: HR = 1.15, 95% CI = 1.13, 1.17). CONCLUSION: Four distinct body shapes, reflecting heterogeneous expressions of obesity, were differentially associated with all-cause and cause-specific mortality. Multitrait body shapes may refine our insights into the associations between different adiposity subtypes and mortality.
Epidemiology
· 2025 Mar · PMID 39887118
·
Publisher ↗
Age-period-cohort models have a long history in epidemiology, social science, and econometrics. An important feature of these models is that they suffer from an inherent identifiability problem, due to the deterministic...Age-period-cohort models have a long history in epidemiology, social science, and econometrics. An important feature of these models is that they suffer from an inherent identifiability problem, due to the deterministic linear relation between age, period, and cohort. A proposed solution to this problem is the mechanism-based approach, which uses sets of mediators to identify the causal age, period, and cohort effects. Although this approach is conceptually general, previous literature has been limited to special cases and parametric identification. We derive a general nonparametric identification result, which is valid under explicit assumptions about the underlying data-generating mechanism and the set of mediators used for identification. We show how this identification result lends itself naturally to parametric estimation of the causal age, period, and cohort effects similar to the parametric G-formula estimation in causal inference.
Brown JP, Yland JJ, Williams PL
… +2 more, Huybrechts KF, Hernández-Díaz S
Epidemiology
· 2025 Mar · PMID 39887117
·
Full text
The analysis of perinatal studies is complicated by twins and other multiple births even when multiples are not the exposure, outcome, or a confounder of interest. In analyses of infant outcomes restricted to live births...The analysis of perinatal studies is complicated by twins and other multiple births even when multiples are not the exposure, outcome, or a confounder of interest. In analyses of infant outcomes restricted to live births, common approaches to handling multiples include restriction to singletons, counting outcomes at the pregnancy level (i.e., by counting if at least one twin experienced a binary outcome), or infant-level analysis including all infants and accounting for clustering of outcomes, such as by using generalized estimating equations or mixed effects models. Several healthcare administration databases only support restriction to singletons or pregnancy-level approaches. For example, in MarketScan insurance claims data, diagnoses in twins are often assigned to a single infant identifier, thereby preventing ascertainment of infant-level outcomes among multiples. Different approaches correspond to different questions, produce different estimands, and often rely on different assumptions. We demonstrate the differences that can arise from these different approaches using Monte Carlo simulations, algebraic formulas, and an applied example.
Knapp EA, Kress AM, Ghidey R
… +28 more, Gorham TJ, Galdo B, Petrill SA, Aris IM, Bastain TM, Camargo CA, Coccia MA, Cragoe N, Dabelea D, Dunlop AL, Gebretsadik T, Hartert T, Hipwell AE, Johnson CC, Karagas MR, LeWinn KZ, Maldonado LE, McEvoy CT, Mirzakhani H, O'Connor TG, O'Shea TM, Wang Z, Wright RJ, Ziegler K, Zhu Y, Bartlett CW, Lau B, Program Collaborators for Environmental influences on Child Health Outcomes
Epidemiology
· 2025 May · PMID 39884749
·
Full text
BACKGROUND: Collaborative research consortia provide an efficient method to increase sample size, enabling evaluation of subgroup heterogeneity and rare outcomes. In addition to missing data challenges faced by all cohor...BACKGROUND: Collaborative research consortia provide an efficient method to increase sample size, enabling evaluation of subgroup heterogeneity and rare outcomes. In addition to missing data challenges faced by all cohort studies like nonresponse and attrition, collaborative studies have missing data due to differences in study design and measurement of the contributing studies. METHODS: We extend ROSETTA, a latent variable method that creates common measures across datasets collecting the same latent constructs with only partial overlap in measures, to define a common measure of socioeconomic status (SES) across cohorts with varying indicators in the Environmental influences on Child Health Outcomes Cohort, a consortium of pregnancy and pediatric cohorts. RESULTS: Starting with 52 indicators of prenatal SES from 39,372 participants across 53 cohorts, ROSETTA created three factors representing key domains of SES: income and education, insurance and poverty, and unemployment. At least one factor score was available for 34,528 participants and two factors were available for more participants than any single indicator. Factors fit the data well, had content validity, and were correlated with alternative measures of SES (for income and education factor, r = 0.40-0.89). Higher SES as measured by the factor scores was associated with lower odds of prenatal smoking: odds ratio income and education : 0.42 (95% confidence interval: 0.38, 0.45). Missing data were reduced compared with most methods, except for multiple imputation. CONCLUSION: ROSETTA aids in pooled analysis of individual participant data by creating measures on a common scale and maximizing data in the presence of missing and mismatched measures.
Asheim A, Nilsen SM, Opdahl S
… +5 more, Risnes K, Balstad Magnussen E, Carlsen F, Davies NM, Bjørngaard JH
Epidemiology
· 2025 May · PMID 39874482
·
Full text
BACKGROUND: Hospital regionalization involves balancing hospital volume and travel time. We investigated how hospital volume and travel time affect perinatal mortality and the risk of delivery in transit using three diff...BACKGROUND: Hospital regionalization involves balancing hospital volume and travel time. We investigated how hospital volume and travel time affect perinatal mortality and the risk of delivery in transit using three different study designs. METHODS: This nationwide cohort study used data from the Medical Birth Registry of Norway (1999-2016) and Statistics Norway. We compared estimates across three designs: (1) Observed confounder adjustment: Comparing women giving birth at hospitals of different sizes and travel times (1,066,332 births), (2) Sibling comparison: Comparing women who moved between hospital catchment areas between births (203,464 births), and (3) Neighbor comparison: comparing women living in neighboring municipalities, but in different hospital catchment areas (460,776 births). RESULTS: The study population included 5080 (0.48%) perinatal deaths and 7063 deliveries in transit (0.66%). For hospitals with 2000 compared with 500 births/year, observed confounder adjustment showed 1.81 times higher perinatal mortality (95% confidence interval [CI]: 1.21, 2.73). However, sibling and neighbor comparisons showed a factor of 0.64 (95% CI: 0.43, 0.97) and 0.61% (95% CI: 0.43, 0.88) lower perinatal mortality, respectively. Increased travel time was strongly associated with higher perinatal mortality using observed confounder adjustment, but this was not supported by the other designs. Longer travel time was consistently linked to an increased risk of delivery in transit. CONCLUSIONS: Perinatal mortality was higher in high-volume hospitals when adjusting for observed confounders. However, triangulating inferences from the other designs suggested the opposite, estimating that observed confounder control was insufficient. This supports the idea that access to higher-volume hospitals could improve perinatal outcomes at the population level.
Keränen E, Rysä J, Tiihonen M
… +2 more, Hartikainen S, Tolppanen AM
Epidemiology
· 2025 May · PMID 39868701
·
Full text
BACKGROUND: Helicobacter pylori ( H. pylori ) has been inconsistently associated with the risk of Alzheimer disease. The exposure assessment period has often overlapped with the prodromal time of Alzheimer disease. Cogni...BACKGROUND: Helicobacter pylori ( H. pylori ) has been inconsistently associated with the risk of Alzheimer disease. The exposure assessment period has often overlapped with the prodromal time of Alzheimer disease. Cognitive disorders might increase vulnerability to infectious pathogens, complicating the ascertainment of the temporal relationship between H. pylori infection and Alzheimer disease. METHODS: This Finnish nested case-control study included 70,520 persons with incident Alzheimer disease diagnosed between 2005 and 2011 and 281,233 age-, sex-, and region of residence-matched controls. We obtained information on comorbidities and drug use from the national healthcare registers. We identified dispensed H. pylori eradication treatments from the Prescription Register. We considered exposure at least 5 years before Alzheimer disease diagnosis in the main analysis. We compared the risk of Alzheimer disease between H. pylori eradication treatment users and nonusers using confounder-adjusted (comorbidities and other drug use) conditional logistic regression. We assessed cumulative exposure by calculating the number of eradication treatments. RESULTS: The prevalence of exposure to H. pylori eradication treatment at least 5 years before the outcome was 4.1% in cases and 3.9% in controls. The odds ratio (95% confidence interval) was 1.06 (1.02, 1.11) in the crude and 1.03 (0.99, 1.07) in the confounder-adjusted model. We observed no association between cumulative exposure and risk of Alzheimer disease. CONCLUSION: Our results, reflecting diagnosed and treated H. pylori infection late in life, do not support the hypothesis of H. pylori as an independent risk factor for Alzheimer disease. The previously reported association may be explained by reverse association and confounding.
Lousdal ML, Lash TL, Flanders WD
… +4 more, Brookhart MA, Kristiansen IS, Vedsted P, Støvring H
Epidemiology
· 2025 May · PMID 39855264
·
Publisher ↗
BACKGROUND: Improvements in breast cancer therapy since the randomized controlled trials of mammography screening might have reduced the screening benefit. Most observational studies of mammography effectiveness would be...BACKGROUND: Improvements in breast cancer therapy since the randomized controlled trials of mammography screening might have reduced the screening benefit. Most observational studies of mammography effectiveness would be confounded by these improvements and other factors. Using a design resistant to this confounding, we evaluated whether mammography in asymptomatic women reduces breast cancer mortality during the treatment era succeeding the trials. METHODS: We designed a quasi-experimental cohort study in regions of Denmark without organized screening. We predicted the number of expected mammograms for each general practice based on observed numbers of mammograms and individual risk factors. Regardless of a woman's individual exposure to mammography, we assigned her the ratio of observed to expected mammograms of her general practice as her instrumental variable. We employed this potential instrumental variable as mammography exposure status and followed women from 1 January 2006 until death, emigration, or 31 December 2014, whichever came first. RESULTS: We included 169,197 women aged 50-66 from 738 general practices and without previous breast cancer as of 1 January 2006. Women affiliated with a practice referring more women than expected, compared with less, had a lower hazard of breast cancer death (hazard ratio = 0.80; 95% confidence interval = 0.68, 0.95). Negative control associations were near null, suggesting no confounding bias. CONCLUSIONS: This quasi-experimental study estimated a continued protective effect of mammography in women where most were presumably asymptomatic. In contrast to conventional observational studies, the use of practice referral ratio as an instrumental variable may avoid bias from uncontrolled confounding.
Ish JL, Madrigal JM, Pearce JL
… +5 more, Keil AP, Fisher JA, Jones RR, Sandler DP, White AJ
Epidemiology
· 2025 May · PMID 39855262
·
Full text
BACKGROUND: We evaluated air emissions of industrial compounds, many of which have carcinogenic or endocrine-disrupting properties, in relation to breast cancer incidence. METHODS: Using the United States Environmental P...BACKGROUND: We evaluated air emissions of industrial compounds, many of which have carcinogenic or endocrine-disrupting properties, in relation to breast cancer incidence. METHODS: Using the United States Environmental Protection Agency's Toxics Release Inventory, we quantified air emissions of 28 compounds near Sister Study participants' residences during the 10 years leading up to study enrollment (2003-2006; n = 46,150). We used Cox proportional hazards regression to estimate adjusted hazard ratios (HR) and 95% confidence intervals (CI) for associations of residential emission levels of single pollutants with incident breast cancer. We assessed pollutant mixtures using an exposure continuum mapping (ECM) framework and characterized associations using a joint-exposure response function. RESULTS: During follow-up (median = 13.4 years), we identified 4155 breast cancer cases. We observed nonmonotonic but elevated associations with breast cancer for emissions within 3 km of the residence for nickel compounds (HR quintile5vs.none = 1.3; 95% CI = 1.0, 1.6) and trichloroethylene (HR quintile5vs.none = 1.3; 95% CI = 1.0, 1.6). ECM identified 25 mixture profiles that explained 72% of the variance in emissions patterns, with most participants experiencing relatively low emissions profiles. The joint-exposure response function suggested that a higher incidence of breast cancer occurred among individuals with relatively rare, high emissions profiles; however, the overall trend was not associated with breast cancer ( P = 0.09). CONCLUSIONS: In our study, breast cancer incidence was associated with air emissions of certain industrial carcinogens. Although the overall emissions mixture did not show a trend related to breast cancer, this may not reflect the importance of individual compounds or specific emissions sources.
Jia KM, Boyer CB, Wallinga J
… +1 more, Lipsitch M
Epidemiology
· 2025 May · PMID 39855261
·
Full text
During the coronavirus disease (COVID-19) pandemic, researchers attempted to estimate the number of averted and avertible outcomes due to vaccination campaigns to quantify public health impact. However, the estimands use...During the coronavirus disease (COVID-19) pandemic, researchers attempted to estimate the number of averted and avertible outcomes due to vaccination campaigns to quantify public health impact. However, the estimands used in these analyses have not been previously formalized. It is also unclear how these analyses relate to the broader framework of direct, indirect, total, and overall causal effects under interference. Here, using potential outcome notation, we adjust the direct and overall effects to accommodate analyses of averted and avertible outcomes. We use this framework to interrogate the commonly held assumption that vaccine-averted outcomes via direct impact among vaccinated individuals (or vaccine-avertible outcomes via direct impact among unvaccinated individuals) is a lower bound on vaccine-averted (or -avertible) outcomes overall. To do so, we describe a susceptible-infected-recovered-death model stratified by vaccination status. When vaccine efficacies wane, the lower bound fails for vaccine-avertible outcomes. When transmission or fatality parameters increase over time, the lower bound fails for both vaccine-averted and -avertible outcomes. Only in the simplest scenario where vaccine efficacies, transmission, and fatality parameters are constant over time, outcomes averted via direct impact among vaccinated individuals (or outcomes avertible via direct impact among unvaccinated individuals) is a lower bound on overall impact. In conclusion, the lower bound can fail under common violations to assumptions on time-invariant vaccine efficacy, pathogen properties, or behavioral parameters. In real data analyses, estimating what seems like a lower bound on overall impact through estimating direct impact may be inadvisable without examining the directions of indirect effects.
Kline DM, White BN, Lancaster KE
… +4 more, Egan KL, Murphy E, Miller WC, Hepler SA
Epidemiology
· 2025 May · PMID 39855258
·
Full text
BACKGROUND: The overdose epidemic remains largely driven by opioids, but the county-level prevalence of opioid misuse is unknown. Without this information, public health and policy responses are limited by a lack of know...BACKGROUND: The overdose epidemic remains largely driven by opioids, but the county-level prevalence of opioid misuse is unknown. Without this information, public health and policy responses are limited by a lack of knowledge on the scope of the problem. METHODS: Using an integrated abundance model, we estimate the annual county-level prevalence of opioid misuse for counties in North Carolina from 2016 to 2021. The model integrates county-level observed counts of illicit opioid overdose deaths, people receiving prescriptions for buprenorphine, and people served by treatment programs. It also incorporates state-level survey estimates of misuse prevalence. County-level social and environmental covariates are also accounted for in the model. Data are integrated through a Bayesian hierarchical model to estimate posterior distributions of the parameters. RESULTS: In general, the estimated prevalence of misuse was decreasing over the study period. Estimated prevalence was above average in the western and southeastern parts of the state. We also estimated that the proportion of people who misuse opioids who fatally overdosed increased sharply over the study period as the median estimated proportion in 2021 was more than 8 times greater than in 2016. The proportion of people who misuse opioids who received buprenorphine and were served by treatment programs increased over the study period. CONCLUSIONS: Estimates from our integrated abundance model fill an important gap in public health knowledge about the local prevalence of people who misuse opioids and can be used to inform an adequate and equitable allocation of resources to communities across the state.
Epidemiology
· 2025 May · PMID 39855255
·
Publisher ↗
Epidemiologic studies recruiting individuals with higher-than-population-average mortality can be affected by "truncation by death," whereby the outcome of interest (e.g., quality of life) is considered not to be defined...Epidemiologic studies recruiting individuals with higher-than-population-average mortality can be affected by "truncation by death," whereby the outcome of interest (e.g., quality of life) is considered not to be defined for individuals who die before the end of follow-up. Here, we use the potential outcomes framework and principal stratification to derive conditions under which the survivor average causal effect, an estimand defined for the "always-survivors" stratum, is modified by a variable that represents a possible common cause of survival and the outcome of interest and by a variable that only affects survival. Further, we show that this principal effect can be expressed as a weighted average of this treatment effect for individuals with each level of these variables, and that these weights depend not only on the relative frequencies of the levels in the total population but also on the "always-survivors" principal stratum. We also discuss the implications of this work for the transportability of the survivor average causal effect.