Validation of Questionnaire-based Case Definitions for Chronic Obstructive Pulmonary Disease.

BACKGROUND: Various questionnaire-based definitions of chronic obstructive pulmonary disease (COPD) have been applied using the US representative National Health and Nutrition Examination Survey (NHANES), but few have been validated against objective lung function data. We validated two prior definitions that incorporated self-reported physician diagnosis, respiratory symptoms, and/or smoking. We also validated a new definition that we developed empirically using gradient boosting, an ensemble machine learning method. METHODS: Data came from 7,996 individuals 40-79 years who participated in NHANES 2007-2012 and underwent spirometry. We considered participants "true" COPD cases if their ratio of postbronchodilator forced expiratory volume in 1 second to forced vital capacity was below 0.7 or the lower limit of normal. We stratified all analyses by smoking history. We developed a gradient boosting model for smokers only; predictors assessed (25 total) included sociodemographics, inhalant exposures, clinical variables, and respiratory symptoms. RESULTS: The spirometry-based COPD prevalence was 26% for smokers and 8% for never smokers. Among smokers, using questionnaire-based definitions resulted in a COPD prevalence ranging from 11% to 16%, sensitivity ranging from 18% to 35%, and specificity ranging from 88% to 92%. The new definition classified participants based on age, bronchodilator use, body mass index (BMI), smoking pack-years, and occupational organic dust exposure, and resulted in the highest sensitivity (35%) and specificity (92%) among smokers. Among never smokers, the COPD prevalence ranged from 4% to 5%, and we attained good specificity (96%) at the expense of sensitivity (9-10%). CONCLUSION: Our results can be used to parametrize misclassification assumptions for quantitative bias analysis when pulmonary function data are unavailable.

Features of the Physical and Social Neighborhood Environment and Neighborhood-Level Alzheimer's Disease and Related Dementia in South Carolina.

BACKGROUND: Studies are increasingly examining the relationship between the neighborhood environment and cognitive decline; yet, few have investigated associations between multiple neighborhood features and Alzheimer's disease and related dementias (ADRD). OBJECTIVE: We investigated the relationship between neighborhood features and ADRD cumulative incidence from 2010 to 2014 in the South Carolina Alzheimer's Disease Registry (SCADR). METHODS: Diagnosed ADRD cases ≥50 years of age were ascertained from the SCADR by ZIP code and census tract. Neighborhood features from multiple secondary sources included poverty, air pollution [particulate matter with a diameter of 2.5 micrometers or less (PM2.5)], and rurality at the census-tract level and access to healthy food, recreation facilities, and diabetes screening at the county level. In addition to using Poisson generalized linear regression to estimate ADRD incident rate ratios (IRR) with 95% confidence intervals (CIs), we applied integrated nested Laplace approximations and stochastic partial differential equations (INLA-SPDE) to address disparate spatial scales. We estimated associations between neighborhood features and ADRD cumulative incidence. RESULTS: The average annual ADRD cumulative incidence was 690 per 100,000 people per census tract (95% CI: 660, 710). The analysis was limited to 98% of census tracts with a population ≥50 years old (i.e., 1,081 of 1,103). The average percent of families living below the federal poverty line per census tract was 18.8%, and ∼20% of census tracts were considered rural. The average percent of households with limited access to healthy food was 6.4%. In adjusted models, every 5µg/m3) increase of PM2.5 was associated with 65% higher ADRD cumulative incidence (IRR=1.65; 95% CI: 1.30, 2.09), where PM2.5 at or below 12 µg/m3 is considered healthy. Compared to large urban census tracts, rural and small urban tracts had 10% (IRR=1.10; 95% CI: 1.00, 1.23) and 5% (IRR=1.05; 95% CI: 0.96, 1.16) higher ADRD, respectively. For every percent increase of the county population with limited access to healthy food, ADRD was 2% higher (IRR=1.02; 95% CI: 1.01, 1.04). CONCLUSIONS: Neighborhood environment features, such as higher air pollution levels, were associated with higher neighborhood ADRD incidence. The INLA-SPDE method could have broad applicability to data collected across disparate spatial scales. https://doi.org/10.1289/EHP13183.

Implementing machine learning methods with complex survey data: Lessons learned on the impacts of accounting sampling weights in gradient boosting.

Despite the prominent use of complex survey data and the growing popularity of machine learning methods in epidemiologic research, few machine learning software implementations offer options for handling complex samples. A major challenge impeding the broader incorporation of machine learning into epidemiologic research is incomplete guidance for analyzing complex survey data, including the importance of sampling weights for valid prediction in target populations. Using data from 15, 820 participants in the 1988-1994 National Health and Nutrition Examination Survey cohort, we determined whether ignoring weights in gradient boosting models of all-cause mortality affected prediction, as measured by the F1 score and corresponding 95% confidence intervals. In simulations, we additionally assessed the impact of sample size, weight variability, predictor strength, and model dimensionality. In the National Health and Nutrition Examination Survey data, unweighted model performance was inflated compared to the weighted model (F1 score 81.9% [95% confidence interval: 81.2%, 82.7%] vs 77.4% [95% confidence interval: 76.1%, 78.6%]). However, the error was mitigated if the F1 score was subsequently recalculated with observed outcomes from the weighted dataset (F1: 77.0%; 95% confidence interval: 75.7%, 78.4%). In simulations, this finding held in the largest sample size (N = 10,000) under all analytic conditions assessed. For sample sizes <5,000, sampling weights had little impact in simulations that more closely resembled a simple random sample (low weight variability) or in models with strong predictors, but findings were inconsistent under other analytic scenarios. Failing to account for sampling weights in gradient boosting models may limit generalizability for data from complex surveys, dependent on sample size and other analytic properties. In the absence of software for configuring weighted algorithms, post-hoc re-calculations of unweighted model performance using weighted observed outcomes may more accurately reflect model prediction in target populations than ignoring weights entirely.

Association of distance to swine concentrated animal feeding operations with immune-mediated diseases: An exploratory gene-environment study.

BACKGROUND: Concentrated animal feeding operations (CAFOs) are a source of environmental pollution and have been associated with a variety of health outcomes. Immune-mediated diseases (IMD) are characterized by dysregulation of the normal immune response and, while they may be affected by gene and environmental factors, their association with living in proximity to a CAFO is unknown. OBJECTIVES: We explored gene, environment, and gene-environment (GxE) relationships between IMD, CAFOs, and single nucleotide polymorphisms (SNPs) of prototypical xenobiotic response genes AHR, ARNT, and AHRR and prototypical immune response gene PTPN22. METHODS: The exposure analysis cohort consisted of 6,464 participants who completed the Personalized Environment and Genes Study Health and Exposure Survey and a subset of 1,541 participants who were genotyped. We assessed the association between participants' residential proximity to a CAFO in gene, environment, and GxE models. We recombined individual associations in a transethnic model using METAL meta-analysis. RESULTS: In White participants, ARNT SNP rs11204735 was associated with autoimmune diseases and rheumatoid arthritis (RA), and ARNT SNP rs1889740 was associated with RA. In a transethnic genetic analysis, ARNT SNPs rs11204735 and rs1889740 and PTPN22 SNP rs2476601 were associated with autoimmune diseases and RA. In participants living closer than one mile to a CAFO, the log-distance to a CAFO was associated with autoimmune diseases and RA. In a GxE interaction model, White participants with ARNT SNPs rs11204735 and rs1889740 living closer than eight miles to a CAFO had increased odds of RA and autoimmune diseases, respectively. The transethnic model revealed similar GxE interactions. CONCLUSIONS: Our results suggest increased risk of autoimmune diseases and RA in those living in proximity to a CAFO and a potential role of the AHR-ARNT pathway in conferring risk. We also report the first association of ARNT SNPs rs11204735 and rs1889740 with RA. Our findings, if confirmed, could allow for novel genetically-targeted or other preventive approaches for certain IMD.

Feasibility of leveraging menstrual cycle tracking apps for preconception research recruitment.

Background: Mobile applications (apps) present a new opportunity to study menstrual cycles and time to pregnancy. Understanding the characteristics of cycle tracking app users is important to evaluate the feasibility of recruiting participants for preconception research. Methods: Users of a cycle tracking smartphone app, Ovia Fertility, aged 18 or older in the U.S. were randomly invited via email to complete a "fertility research" questionnaire that included demographic and reproductive characteristics. Among those attempting pregnancy without medical assistance, attempt duration, factors influencing pregnancy planning, health history and behaviors while attempting to conceive were queried. Respondents could choose to enter a raffle for a $50 gift card. Results: Initially, 639 people responded to the demographics portion of the survey representing 49 states and Washington DC. Of these, 344 (54%) were trying to conceive and of those, 297 (86%) were not using medical treatments. Of those not trying to conceive, 12% reported that they planned to start in the next 3 months. Most participants were ages 26-35 (63%), of White race (70%), reported non-Hispanic ethnicity (87%), had at least a bachelor's degree (56%) and an income between $50,000 and $200,000 (58%). One-third were of recommended BMI (35%), 24% overweight, and 41% obese. Most participants reported no fertility-related health conditions (58%). Forty-eight participants (17%) had been trying to conceive for 1 month or less, 88 (31%) had been trying for 2 months or less, and 122 (43%) for 3 months or less. Interruptions in pregnancy attempts were common, 31% reported periods without intercourse. Of those attempting pregnancy, 47% of partners completed their own questionnaire. Conclusion: This first-of its-kind analysis describes users of a cycle-tracking smartphone app who could be eligible for recruitment to a prospective time-to-pregnancy study. Survey respondents were diverse with respect to geographic location, BMI, and income. However, special recruitment efforts will be needed to recruit participants and partners who identify as other than non-Hispanic White. Participants with fertility-related conditions are not overly represented among app users who are trying to conceive. Targeting and pre-enrolling app users who are planning to begin a pregnancy attempt in the next 3 months may be an advantage of app-based recruitment.

Relation of repeated exposures to air emissions from swine industrial livestock operations to sleep duration and awakenings in nearby residential communities.

OBJECTIVES: Since waste from swine industrial livestock operations (ILOs) produces air pollutants associated with negative health outcomes among nearby residents, we assessed the impact of odorant emissions on sleep duration and awakenings. DESIGN: A repeated-measures design. SETTING: Sixteen residential communities in eastern North Carolina hosting swine ILOs. PARTICIPANTS: Eighty participants residing in eastern North Carolina from 2003 to 2005. INTERVENTION (IF ANY): Not applicable. MEASUREMENTS: Study participants completed twice-daily diaries in which they rated the strength of hog odors and indicated whether they were asleep or awake per hour for 2 weeks. Simultaneously, a monitoring trailer placed in a central location in each community measured the atmospheric concentration of hydrogen sulfide (H2S). Subject-conditional fixed-effects regression models were used to estimate associations between 2 markers of swine ILO pollutant exposures (H2S and swine odor) and 2 self-reported sleep outcomes (nightly sleep duration and awakening from sleep). RESULTS: Among 80 participants, nightly (across a 12-hour period) swine odor was associated with lower nightly sleep duration (mean difference = -14.3 minutes, 95% confidence interval -25.0 to -3.3 minutes) compared to odor-free nights and detection of nightly hydrogen sulfide was associated with an increased risk of awakening (hazard ratio = 1.23, 95% confidence interval 0.98 to 1.55) compared to nights with no detection of hydrogen sulfide. CONCLUSIONS: These results suggest that environmental odorants are important considerations for sleep health and highlight the importance of sleep as a potential mediator between environmental air pollution and health outcomes impacted by poor sleep.

Natural hazards and mental health among US Gulf Coast residents.

BACKGROUND: Individuals affected by disasters are at risk for adverse mental health sequelae. Individuals living in the US Gulf Coast have experienced many recent major disasters, but few studies have explored the cumulative burden of experiencing multiple disasters on mental health. OBJECTIVE: The objective of this study was to evaluate the relationship between disaster burden and mental health. METHODS: We used data from 9278 Gulf Long-term Follow-up Study participants who completed questionnaires on perceived stress, anxiety, depression, and post-traumatic stress disorder (PTSD) in 2011-2013. We linked 2005-2010 county-level data from the Spatial Hazard Events and Losses Database for the United States, a database of loss-causing events, to participant's home address. Exposure measures included total count of loss events as well as severity quantified as property/crop losses per capita from all hazards. We used multilevel modeling to estimate odds ratios (OR) and 95% confidence intervals (CI) for each exposure-outcome relationship. RESULTS: Total count of loss events was positively associated with perceived stress (ORQ4:1.40, 95% CI:1.21-1.61) and was inversely associated with PTSD (ORQ4:0.66, 95% CI:0.45-0.96). Total duration of exposure was also associated with stress (ORQ4:1.16, 95% CI:1.01-1.33) but not with other outcomes. Severity based on cumulative fatalities/injuries was associated with anxiety (ORQ4:1.31, 95% CI:1.05-1.63) and stress (ORQ4:1.34, 95% CI:1.15-1.57), and severity based on cumulative property/crop losses was associated with anxiety (ORQ4:1.42, 95% CI:1.12-1.81), depression (ORQ4:1.22, 95% CI:0.95-1.57) and PTSD (ORQ4:1.99, 95% CI:1.44-2.76).