BACKGROUND:Remediating vacant lots is associated with reductions in neighborhood violence and other adverse health outcomes, potentially via increasing visible cues-to-care that signal stewardship. However, accurately measuring these cues over time remains challenging, limiting our ability to measure the sustained effects of lot remediation. METHODS:We developed a standardized audit protocol using subject-matter and community experts and assessed vacant lot cues-to-care from Google Street View imagery of vacant lots in Philadelphia from 2007-2023. We fit a two-parameter item response theory model to combine the observable cues-to-care measured consistently into a single latent score representing the lot care condition. We validated the latent score by testing its sensitivity to a randomized controlled greening trial and a local lot maintenance program. RESULTS:Five raters audited 3419 images (518 vacant lots; 6315 ratings) captured from July 2007-October 2023. We selected reliable and conceptually aligned items to fit an item response theory (IRT) model representing a latent score of lot care condition and evaluated validity of that score. Average pairwise Cohen's kappa for audited items was 0.31 (SD=0.14) across all raters, and 0.55 (SD=0.18) among our most concordant raters. Eight items with above-moderate reliability (K>0.40) were retained for the final IRT model, which demonstrated high internal reliability (0.96). This IRT-derived care score was sensitive to improvements in lot conditions following greening interventions and enrollment in a lot maintenance program. CONCLUSION:Our IRT-derived score can be used to measure vacant lot care condition accurately over time. This can support long-term evaluation of vacant lot remediation interventions.
BACKGROUND:Air conditioning data are typically available only at coarse spatial resolutions. We examined modifications of heat-morbidity relationships using finer spatial resolution - specifically, ZIP code-level air conditioning penetration rates derived from household-level smart meter electricity records in southern California (2015-2019). METHODS:We first estimated ZIP code-specific relative risk (RR) between heat-related emergency department visits and 4-day lagged temperatures at the 95th to 50th temperature quantile. Meta-regression was used to estimate effect modification, as measured by the ratio of RR. RESULTS:Air conditioning penetration rates ranged from 23% to 96% (median = 79%). Among non-coastal regions, a 10% increase in air conditioning penetration was associated with lower heat and emergency department visit risk: ratio of RR = 0.79 (95% confidence interval: 0.68, 0.91) for dry bulb temperature and 0.80 (95% confidence interval: 0.70, 0.92) for wet bulb temperature. Air conditioning penetration did not modify heat-morbidity associations in coastal regions. CONCLUSION:Our results indicate that neighborhoods with higher air conditioning penetration rates were associated with lower heat-related health risks in noncoastal regions.
BACKGROUND:Parametric g-computation with competing events typically involves fitting multiple pooled logistic regression models. We outline an alternative approach based on fitting a single pooled multinomial logistic model. METHODS:Data from the Women's Interagency HIV Study (n = 1,164) were used to estimate the marginal 2-year risk of highly active antiretroviral therapy (HAART) initiation and AIDS/death before HAART initiation with two parametric g-computation approaches: multiple pooled logistic regression and pooled multinomial logistic regression. The total effect of historical injection drug use was estimated for both event types using the multinomial approach. RESULTS:Both g-computation implementations produced identical results. The 2-year risk difference comparing a scenario where all participants had historical injection drug use to one with no historical injection drug use was -12.5% (-18.0%, -7.0%) for HAART initiation and 13.2% (6.8%, 19.7%) for AIDS/death. CONCLUSIONS:Incorporating a pooled multinomial logit nuisance model for parametric g-computation simplifies estimation of total effects while accounting for right-censoring and competing events.
Many research findings are based on a nested subset of an original cohort followed over time, but results are rarely generalized back to the original population. Here, we conduct a "proof of concept" analysis demonstrating the application of simple methods to generalize the association between prepregnancy obesity and preeclampsia in women from the Nulliparous Pregnancy Outcomes Study: Monitoring Mothers to Be (nuMoM2b, target sample), using a subset of women whose follow-up was extended to 3 years postpartum in the Heart Health Study (nuMoM2b-HHS, source sample). We constructed inverse probability of selection weights and estimated three risk ratios for the association between obesity and preeclampsia: in the target nuMoM2b sample (N = 9,920), in the source nuMoM2b-HHS sample (N = 4,486), and in the weighted source nuMoM2b-HHS sample generalized back to the target sample using inverse probability of selection weights (pseudo N = 4,468). In the target, source, and weighted samples, the estimated risk ratios (95% confidence intervals) were 2.0 (1.7, 2.3), 2.2 (1.8, 2.6), and 2.0 (1.6, 2.4), respectively. We discuss the assumptions involved in generalizing study findings and methods available for doing so. When relevant, researchers should deploy methods for generalizing study nested cohort findings back to a target sample.
BACKGROUND:Participants in the All of Us research study differ from the U.S. population in myriad characteristics, limiting the generalizability of findings. A statistical reweighting tool to improve generalizability would enhance the scientific value of the data. METHODS:To account for differences between All of Us and the nationally representative 1999-2018 National Health and Nutrition Examination Survey (NHANES), we generated selection weights using four models incorporating sociodemographic, self-reported health, and clinical characteristics. We assessed covariate balance and compared predictors of all-cause mortality in weighted All of Us to NHANES using the ratio of hazard ratios (RHRs), where an RHR of one indicates unbiased estimates in (weighted) All of Us relative to NHANES. RESULTS:Weighting improved balance on measured variables between All of Us and NHANES. Among the four weighting models, the most complex model which included sociodemographic, health, and clinical variables and their interactions achieved HRs in All of Us most similar to those in NHANES. For example, the RHR for hypertension for unweighted All of Us vs. NHANES (RHR=1.5; 95% CI=1.4 to 1.7) was reduced to 1.2 (95% CI=0.9 to 1.5) after applying weights from the clinical-interaction model. Even in this model, 17 of 35 HRs evaluated diverged by >20% (RHR <0.8 or >1.2) between weighted All of Us and NHANES. CONCLUSIONS:Predictors of mortality in All of Us differ from those in the U.S. population both in their distribution and in their associations with mortality. Reweighting can mitigate selection bias, but no model we tested comprehensively achieved generalizability.
BACKGROUND:Given established relationships between social class and mortality, the growing concentration of income, wealth, and power among business owners in the United States may have increased mortality inequities across classes. To investigate this hypothesis, we analyzed temporal changes in mortality inequities between owners and non-owners. METHODS:Our sample included respondents ages 25-64 in the 1984, 1989, 1994, and 1999-2013 Panel Study of Income Dynamics with mortality follow-up through 2022 (respondents: 22,103; observations: 103,965). Business owners were individuals with personal or family ownership of, or direct financial interest in, a business in the prior year. Using g-computation, we estimated how inequities between owners and non-owners in 10-year age-adjusted mortality risks changed from 1984 to 2013. Next, we analyzed whether any changes were attributable to shifting social stratification. Finally, we analyzed whether growing income and wealth disparities between owners and non-owners exacerbated inequities. RESULTS:In 1984, non-owners had 1.4 times (95% confidence interval [CI]: 1.1, 1.8) greater 10-year age-adjusted mortality risks than owners. In 2013, the figure was 2.3 (95% CI: 1.8, 3.0), yielding a ratio of risk ratios of 1.7 (95% CI: 1.1, 2.5). After social-stratification adjustment, within-year inequities lessened; however, increases across years attenuated only somewhat (2013 vs. 1984 ratio of risk ratios: 1.5 [95% CI: 0.99, 2.2]). Finally, we did not find that increases in inequities across years would have lessened if income and wealth distributions had remained at 1984 levels. CONCLUSION:Mortality inequities between owners and non-owners have increased and cannot be fully explained by social stratification and individual-level income and wealth distributions.
Evidence-informed infectious disease policy requires estimates of the health effects of infections. However, pathogenic variation, whereby the risk of adverse outcomes depends on the strain-specific characteristics of the pathogen, might complicate the study of many infectious agents. Here, we consider the interpretation of epidemiologic studies of infectious diseases when there is such heterogeneity in strain-specific effects and when information on strain composition is unavailable. We use potential outcomes and causal inference theory for analyses in the presence of multiple versions of treatment to argue that oft-reported quantities in these studies have a causal interpretation that depends on population frequencies of infecting strains. Moreover, as in other contexts where the treatment-variation-irrelevance assumption might be violated, transportability requires additional considerations, beyond those needed for noncompound exposures. This discussion, considering potential heterogeneity in strain-specific effects, will facilitate the interpretation of epidemiologic studies and also highlight the value of pathogen subtype data.
