American journal of epidemiology最新文献

Background: Identifying transmission events is important in understanding infectious disease dynamics. Such events are typically unobservable, particularly in respiratory diseases such as tuberculosis (TB). We apply network techniques to identify transmission clusters and features shared within clusters.

Methods: We estimate directed pairwise transmission probabilities via an existing iterative algorithm that employs a modified Naïve Bayes classifier and use these probabilities to create a network. We explore noise reduction techniques to trim low probability edges. We group individuals with TB based on edges informed by transmission probabilities via network clustering algorithms. We apply our framework to simulated data and assess clustering algorithm performance. We then apply this approach to data from a cohort study in Lima, Peru and examine homogeneity of the clusters using a binary entropy measure.

Results: We find cluster performance to be consistent across all edge trimming scenarios and clustering methods. We find high levels of entropy, implying heterogeneity, for age, sex, socioeconomic status, individuals who work outside the house, and people using public transit.

Conclusions: We analyze estimated directed pairwise transmission probabilities with network techniques. The approach is consistent across network construction and clustering methods and can be applied to any disease outbreak to understand its dynamics.

Machine learning (ML) methods have the potential to improve precision medicine by estimating personalized treatment effects. However, formal validation of these methods remains limited, leaving their reliability in empirical settings largely uncertain. In this study, we evaluated the internal and external validity of 17 causal heterogeneity ML methods-including metalearners, tree-based methods, and deep learning methods-using data from two large randomized controlled trials: the International Stroke Trial (n = 19 435) and the Chinese Acute Stroke Trial (n = 21 106). We assessed performance using three visual-based metrics and three quantitative metrics. Our analysis found that none of the ML methods consistently demonstrated reliable performance, neither internal nor external. Heterogeneous treatment effects estimated from training data failed to generalize to the test data, even in the absence of distribution shifts. These results raise concerns about the current applicability of ML models in precision medicine and highlight the need for more robust validation techniques to ensure generalizability.

Plasma HIV-1 RNA viral loads (VL) are measured via laboratory assays with changing lower limits of quantification over time. We described an approach to produce an analytic-ready dataset of VLs over time and across longitudinal cohorts of adults. A three-step approach was used: 1) initial data cleaning; 2) data checking with visualization; and 3) final data cleaning. Assumptions, data-driven decisions, and information from cohort-specific data managers produce an analytic-ready dataset of VLs with minimal missing data for date of blood draw, HIV-1 RNA result (copies/mL), below the lower limit of quantification (BLLQ) indicator, and the lower limit of quantification (LLQ). Among 3 663 786 VLs from 186 990 participants (median number of VLs per participant = 12, interquartile range 4-27) measured from 1988 to 2021, 61% of VL records were harmonized via the three-step approach. The proportion of VLs below the lower limit of quantification increased from 39% to 60% after application of this approach. Changes to LLQ, VL result, and BLLQ indicator variables were made to 45%, 36%, and 22% of VLs, respectively. Stated assumptions, visualized data distributions, and a documented approach to preparing an analytic-ready dataset of pooled individual-level longitudinal data revealed data idiosyncrasies, informed assumptions, and improved the data for research inference.

Transgender and gender diverse youth (TGDY) report high levels of distress. While access to gender affirming care is associated with improved wellbeing among TGDY, limited research has examined medium to long term outcomes of this care. PARLONSMAAT (the Participatory longitudinal study of the impacts of gender-affirming care among TGDY) will explore the characteristics, trajectories of care, and health and psychosocial outcomes into young adulthood of youth receiving services at a large Canadian interdisciplinary pediatric gender identity clinic, housed within a French language tertiary pediatric hospital centre. Using a participatory research approach, we will conduct a single-center prospective cohort study. All new clinic patients will be eligible to participate, with rolling recruitment (estimate of 100 new participants/year). Self-reported questionnaires on gender, sexual orientation, mental health, wellbeing, and social support will be collected, in addition to medical records. PARLONSMAAT brings many novel design features to examining the impact of pediatric gender-affirming care, including a large prospective sample, a long-term follow-up into adulthood, integration of self-report and medical data, and participatory design for lived expertise. PARLONSMAAT will offer important insights around the impact and safety of gender-affirming care to improve clinical care and support better health outcomes among TGDY.

As AI systems are increasingly used to guide decisions, it is essential that they follow ethical principles. A core principle in medicine is non-maleficence, often equated with "do no harm". A formal definition of harm based on counterfactual reasoning has been proposed and popularized. This notion of harm has been promoted in simple settings with binary treatments and outcomes. Here, we highlight a problem with this definition in settings involving multiple treatment options. Illustrated by an example with three tuberculosis treatments (say, A, B, and C), we demonstrate that the counterfactual definition of harm can produce intransitive results: B is less harmful than A, C is less harmful than B, yet C is more harmful than A when compared pairwise. This intransitivity poses a challenge as it may lead to practical (clinical) decisions that are difficult to justify or defend. In contrast, an interventionist definition of harm based on expected utility forgoes counterfactual comparisons and ensures transitive treatment rankings.

Spatial patterning of environmental hazards often leads to concerns about spatial confounding: that the exposures we study share similar spatial distributions with other causes of disease. Recent efforts to address spatial confounding have approached it using clever specification of spatial models, or models that adjust for aspects of spatial location itself. In the article by Li et al. (Am J Epidemiol. XXXX;XXX(XX):XXXX-XXXX)), the authors describe and demonstrate several models for addressing spatial confounding for binary exposures. These important results demonstrate an aspect of environmental exposures that should concern all environmental epidemiologists: inadequate adjustment for spatial confounding can increase, rather than decrease, bias from spatial confounding. In this commentary, we disagree with some terminology and enthusiastically agree with the importance of the problem and the utility of the approaches described by these authors. Causal inference in environmental epidemiology is fraught with many challenges, and Li et al. give hope for progress on one of the lesser understood, yet potentially ubiquitous, problems: spatial confounding.

Composite major congenital malformation (MCM) outcomes are commonly used to assess teratogenic effects of prenatal medication exposure, but this approach dilutes effect estimates when the risk is confined to a specific MCM. Tree-based scan statistics address this by screening outcomes using a hierarchical tree, enabling detection of specific risks without predefined hypotheses. To apply this method across ICD-9-CM and ICD-10-CM eras, we developed a unified hierarchical outcomes tree for MCM. We selected ICD-9-CM and ICD-10-CM codes classified as congenital anomalies, removing minor malformations, chromosomal anomalies, and single-gene conditions. A multi-level tree was built based on the Multilevel Clinical Classification Software, General Equivalence Mappings, and expert review. We validated the tree using birth cohorts from MarketScan and Medicaid databases (2011-2013; 2016-2018), assessing balance of MCM prevalences within one year of birth via standardized mean differences (SMDs). The final tree included 1023 codes, organized into 244 clinical MCM groups at the most granular level. We identified 572 107 (2011-2013) and 360 167 infants (2016-2018) in MarketScan and 362 820 and 3 500 589 infants in Medicaid. All SMDs were below 0.1, indicating consistency across coding eras. This hierarchical MCM tree bridges ICD-9-CM and ICD-10-CM, enabling consistent outcome definitions and enhancing detection of specific teratogenic risks.

Overdose prevention centers (OPCs) are associated with improved community health and decreased crime, but opponents argue that OPCs depress nearby property values. We estimated the association of the opening of the first two public recognized OPC in the United States with neighborhood residential rents and real estate sales in the East Harlem and Washington Heights neighborhoods of New York City (NYC). Using augmented synthetic controls, we analyzed quarterly and semiannual rental listings and annual and semiannual sales within 300- and 500-meter buffers around the OPCs. Donor units were buffers around syringe service programs without OPCs and opioid treatment programs. Primary outcomes were median quarterly rental listing price and median annual sales price. Overall, we found no changes in neighborhood rental or sales prices. For quarterly rentals at 300 m, we estimated (ATT, 95% CI) $145 (-$780, $1070) in East Harlem and -$505 (-$1279, $269) in Washington Heights. For annual sales at 500 m, we estimated -$542 993 (-$1 228 024, $142038) in East Harlem and $1 121 706 (-$431 285, $2674697) in Washington Heights. Conformal inference identified no detectable time-point effects. Overall, OPC implementation in NYC was not associated with changes in rents or sales, suggesting these facilities may not generate appreciable effects on local housing values.

Many older adults experience financial insecurity. While prior studies link lower later-life SES, financial stress, and financial shocks to worse cognitive outcomes, limited research has examined how dynamic changes in financial well-being-a multidimensional measure of financial circumstances-influence cognitive aging. Here, we examined associations between changes in financial well-being and memory outcomes among 7676 adults aged 50+ in the Health and Retirement Study ("HRS," 2010-2020). We developed and validated an 8-item index of poor financial well-being using existing HRS survey items aligned with domains from the Consumer Financial Protection Bureau's Financial Well-Being Scale. In confounder-adjusted linear mixed-effects models, we estimated associations of average financial well-being and significant improvements or worsening in financial well-being over four years with changes in memory z-scores calculated biennially from 2016-2020. Each 1-point worsening in average financial well-being was associated with poorer memory function (β = -0.009 SD, 95% CI, -0.020 to 0.003) and accelerated decline (β = -0.007 SD/year, 95% CI, -0.010 to -0.003). Associations were largest for participants with significant worsening of financial well-being and for those aged ≥65 at baseline. Results were robust to sensitivity analyses addressing potential reverse causation and attrition. These findings suggest that midlife and later-life declines in financial well-being may contribute to accelerated cognitive aging.

Epidemiologic studies often evaluate the association between an exposure and an event risk. When time-varying, exposure updates usually occur at discrete visits although changes are in continuous time and survival models require values to be constantly known. Moreover, exposures are likely measured with error, and their observation truncated at the event time. We aimed to quantify in a Cox regression the bias in the association resulting from intermittent measurements of an error-prone exposure. Using simulations under various scenarios, we compared five methods: last observation carried-forward (LOCF), classical two-stage regression-calibration using measurements up to the event (RC) or also after (PE-RC), multiple imputation (MI) and joint modeling of the exposure and the event (JM). The LOCF, and to a lesser extent the classical RC, showed substantial bias in almost all 45 scenarios. The RC bias was avoided when considering post-event information. The MI performed relatively well, as did the JM. Illustrations exploring the association of Body Mass Index and Executive Functioning with dementia risk showed consistent conclusions. Accounting for measurement error and discrete updates is critical when studying time-varying exposures. MI and JM techniques may be applied in this context, while classical RC should be avoided due to the informative truncation.