Application of elastic net regression for modeling COVID-19 sociodemographic risk factors.

OBJECTIVES: COVID-19 has been at the forefront of global concern since its emergence in December of 2019. Determining the social factors that drive case incidence is paramount to mitigating disease spread. We gathered data from the Social Vulnerability Index (SVI) along with Democratic voting percentage to attempt to understand which county-level sociodemographic metrics had a significant correlation with case rate for COVID-19. METHODS: We used elastic net regression due to issues with variable collinearity and model overfitting. Our modelling framework included using the ten Health and Human Services regions as submodels for the two time periods 22 March 2020 to 15 June 2021 (prior to the Delta time period) and 15 June 2021 to 1 November 2021 (the Delta time period). RESULTS: Statistically, elastic net improved prediction when compared to multiple regression, as almost every HHS model consistently had a lower root mean square error (RMSE) and satisfactory R2 coefficients. These analyses show that the percentage of minorities, disabled individuals, individuals living in group quarters, and individuals who voted Democratic correlated significantly with COVID-19 attack rate as determined by Variable Importance Plots (VIPs). CONCLUSIONS: The percentage of minorities per county correlated positively with cases in the earlier time period and negatively in the later time period, which complements previous research. In contrast, higher percentages of disabled individuals per county correlated negatively in the earlier time period. Counties with an above average percentage of group quarters experienced a high attack rate early which then diminished in significance after the primary vaccine rollout. Higher Democratic voting consistently correlated negatively with cases, coinciding with previous findings regarding a partisan divide in COVID-19 cases at the county level. Our findings can assist regional policymakers in distributing resources to more vulnerable counties in future pandemics based on SVI.

Spatial Modeling of Sociodemographic Risk for COVID-19 Mortality.

In early 2020, the Coronavirus Disease 19 (COVID-19) rapidly spread across the United States (US), exhibiting significant geographic variability. While several studies have examined the predictive relationships of differing factors on COVID-19 deaths, few have looked at spatiotemporal variation at refined geographic scales. The objective of this analysis is to examine this spatiotemporal variation in COVID-19 deaths with respect to association with socioeconomic, health, demographic, and political factors. We use multivariate regression applied to Health and Human Services (HHS) regions as well as nationwide county-level geographically weighted random forest (GWRF) models. Analyses were performed on data from three separate time frames which correspond to the spread of distinct viral variants in the US: pandemic onset until May 2021, May 2021 through November 2021, and December 2021 until April 2022. Multivariate regression results for all regions across three time windows suggest that existing measures of social vulnerability for disaster preparedness (SVI) are predictive of a higher degree of mortality from COVID-19. In comparison, GWRF models provide a more robust evaluation of feature importance and prediction, exposing the value of local features for prediction, such as obesity, which is obscured by coarse-grained analysis. Overall, GWRF results indicate that this more nuanced modeling strategy is useful for determining the spatial variation in the importance of sociodemographic risk factors for predicting COVID-19 mortality.

Estimating County Level Health Indicators Using Spatial Microsimulation.

Given the importance of understanding health outcomes at fine spatial scales, iterative proportional fitting (IPF), a form of small area estimation, was applied to a fixed number of health-related variables (obesity, overweight, diabetes) taken from regionalized 2019 survey responses (n = 5474) from the Idaho Behavioral Risk Factor Surveillance System (BRFSS). Using associated county-level American Community Survey (ACS) census data, a set of constraints, which included age categorization, race, sex, and education level, were used to create county-level weighting matrices for each variable, for each of the seven (7) Idaho public health districts. Using an optimized modeling construction technique, we identified significant constraints and grouping splits for each variable/region, resulting in estimates that were internally and externally validated. Externally validated model results for the most populated counties showed correlations ranging from .79 to .85, with p values all below .05. Estimates indicated higher levels of obesity and overweight individuals for midsouth and southwestern Idaho counties, with a cluster of higher diabetes estimates in the center of the state (Gooding, Lincoln, Minidoka, and Jerome counties). Alternative external sources for health outcomes aligned extremely well with our estimates, with wider confidence intervals in more rural counties with sparse populations.

Application of Elastic Net Regression for Modeling COVID-19 Sociodemographic Risk Factors.

Objectives: COVID-19 has been at the forefront of global concern since its emergence in December of 2019. Determining the social factors that drive case incidence is paramount to mitigating disease spread. We gathered data from the Social Vulnerability Index (SVI) along with Democratic voting percentage to attempt to understand which county-level sociodemographic metrics had a significant correlation with case rate for COVID-19. Methods: We used elastic net regression due to issues with variable collinearity and model overfitting. Our modelling framework included using the ten Health and Human Services regions as submodels for the two time periods 22 March 2020 to 15 June 2021 (prior to the Delta time period) and 15 June 2021 to 1 November 2021 (the Delta time period). Results: Statistically, elastic net improved prediction when compared to multiple regression, as almost every HHS model consistently had a lower root mean square error (RMSE) and satisfactory R2 coefficients. These analyses show that the percentage of minorities, disabled individuals, individuals living in group quarters, and individuals who voted Democratic correlated significantly with COVID-19 attack rate as determined by Variable Importance Plots (VIPs). Conclusions: The percentage of minorities per county correlated positively with cases in the earlier time period and negatively in the later time period, which complements previous research. In contrast, higher percentages of disabled individuals per county correlated negatively in the earlier time period. Counties with an above average percentage of group quarters experienced a high attack rate early which then diminished in significance after the primary vaccine rollout. Higher Democratic voting consistently correlated negatively with cases, coinciding with previous findings regarding a partisan divide in COVID-19 cases at the county level. Our findings can assist policymakers in distributing resources to more vulnerable counties in future pandemics based on SVI.

Effects of trust, risk perception, and health behavior on COVID-19 disease burden: Evidence from a multi-state US survey.

Early public health strategies to prevent the spread of COVID-19 in the United States relied on non-pharmaceutical interventions (NPIs) as vaccines and therapeutic treatments were not yet available. Implementation of NPIs, primarily social distancing and mask wearing, varied widely between communities within the US due to variable government mandates, as well as differences in attitudes and opinions. To understand the interplay of trust, risk perception, behavioral intention, and disease burden, we developed a survey instrument to study attitudes concerning COVID-19 and pandemic behavioral change in three states: Idaho, Texas, and Vermont. We designed our survey (n = 1034) to detect whether these relationships were significantly different in rural populations. The best fitting structural equation models show that trust indirectly affects protective pandemic behaviors via health and economic risk perception. We explore two different variations of this social cognitive model: the first assumes behavioral intention affects future disease burden while the second assumes that observed disease burden affects behavioral intention. In our models we include several exogenous variables to control for demographic and geographic effects. Notably, political ideology is the only exogenous variable which significantly affects all aspects of the social cognitive model (trust, risk perception, and behavioral intention). While there is a direct negative effect associated with rurality on disease burden, likely due to the protective effect of low population density in the early pandemic waves, we found a marginally significant, positive, indirect effect of rurality on disease burden via decreased trust (p = 0.095). This trust deficit creates additional vulnerabilities to COVID-19 in rural communities which also have reduced healthcare capacity. Increasing trust by methods such as in-group messaging could potentially remove some of the disparities inferred by our models and increase NPI effectiveness.

Do Invasive and Naturalized Aphid Pest Populations Respond Differently to Climatic and Landscape Factors?

Ongoing environmental change affects pest populations, migration, and propensity to damage crops, but the responses to climatic drivers could vary among newly invasive and already naturalized closely related species. To compare these responses of a newly invasive aphid, Metopolophium festucae cerealium (Stroyan), with its naturalized congeneric [M. dirhodum (Walker)] and confamilial [Sitobian avenae (Fab.)], we conducted annual surveys over four years across a total of 141 winter wheat fields in the inland Pacific Northwest, USA. Key climatic factors (cumulative precipitation for each calendar year to sampling date, cumulative degree days), landscape factors (proportion of wheat and landscape diversity within the sample year), and Julian day were calculated for each sampling event, and aphid abundance by species, total aphid abundance, overall species richness, diversity, and aphid community composition were assessed. Metopolophium f. cerealium, the second most abundant species, was positively associated with precipitation, suggesting a projected increase in precipitation in winter and spring in the region could favor its establishment and expansion. Although M. dirhodum and S. avenae linearly (positively) associated with temperature, M. f. cerealium did not, indicating that continued warming may be detrimental to the species. Despite the weak impacts of landscape factors, our study indicated that more wheat generally facilitates cereal aphid abundance. Metopolophium f. cerealium abundance tended to be higher in earlier (May/early June vs. late June/July) samples when wheat crop could be vulnerable to aphid feeding. This study suggests that the new presence of M. f. cerealium has important pest management implications in the region.

Parenting matters: Moderation of biological and community risk for obesity.

Contributions of parental limit setting, negativity, scaffolding, warmth, and responsiveness to Body Mass Index (BMI) were examined. Parenting behaviors were observed in parent-child interactions, and child BMI was assessed at 5 years of age. Mothers provided demographic information and obtained child saliva samples used to derive cortisol concentration indicators (N = 250). Geospatial crime indices were computed based on publically available information for a subsample residing within the boundaries of a Pacific Northwest city (N = 114). Maternal warmth and limit setting moderated the association between child HPA-axis regulation and BMI. BMI was higher for children at lower cortisol concentrations with greater maternal warmth and lower for youngsters with mid-range cortisol values under high maternal limit setting. Maternal scaffolding moderated the effects of crime exposure, so that lower scaffolding translated into higher child BMI with greater neighborhood crime exposure. These parenting behaviors could be leveraged in obesity prevention/intervention efforts.

Featured Article: Community Crime Exposure and Risk for Obesity in Preschool Children: Moderation by the Hypothalamic-Pituitary-Adrenal-Axis Response.

Objective: Identification of early risk factors related to obesity is critical to preventative public health efforts. In this study, we investigated links between the Hypothalamic-Pituitary-Adrenal (HPA)-axis activity (diurnal cortisol pattern), geospatially operationalized exposure to neighborhood crime, and body mass index (BMI) for a sample of 5-year-old children. Greater community crime exposure and lower HPA-axis activity were hypothesized to contribute to higher BMI, with child HPA-axis moderating the association between crime exposure and BMI. Method: Families residing within the boundaries of the City of Seattle (N = 114) provided information concerning demographic/psychosocial risk factors, used to calculate a Cumulative Risk Index, indicating the number of contextual adversities present. Child BMI and diurnal cortisol pattern (derived from assays of saliva samples) were examined, along with neighborhood crime indices computed with publically available information, based on participants' locations. Results: Hierarchical multiple regression analyses, adjusted for covariates (cumulative risk, age, and sex), indicated that crime proximity made a unique contribution to child BMI, in the direction signaling an increase in the risk for obesity. Consistent with our hypothesis, a significant interaction was observed, indicative of moderation by diurnal cortisol pattern. Follow-up simple slope analyses demonstrated that crime exposure was significantly related to higher BMI for children with low-flat (blunted) diurnal cortisol patterns, where community crime and BMI were not significantly associated at higher levels of cortisol. Conclusion: Community crime exposure contributes to higher BMI as early as the preschool period, and blunted diurnal cortisol patterns may place children experiencing neighborhood adversity at greater risk for obesity.

Latent profile and cluster analysis of infant temperament: Comparisons across person-centered approaches.

There is renewed interest in person-centered approaches to understanding the structure of temperament. However, questions concerning temperament types are not frequently framed in a developmental context, especially during infancy. In addition, the most common person-centered techniques, cluster analysis (CA) and latent profile analysis (LPA), have not been compared with respect to derived temperament types. To address these gaps, we set out to identify temperament types for younger and older infants, comparing LPA and CA techniques. Multiple data sets (N = 1,356; 672 girls, 677 boys) with maternal ratings of infant temperament obtained using the Infant Behavior Questionnaire-Revised (Gartstein & Rothbart, 2003) were combined. All infants were between 3 and 12 months of age (M = 7.85; SD = 3.00). Due to rapid development in the first year of life, LPA and CA were performed separately for younger (n = 731; 3 to 8 months of age) and older (n = 625; 9 to 12 months of age) infants. Results supported 3-profile/cluster solutions as optimal for younger infants, and 5-profile/cluster solutions for the older subsample, indicating considerable differences between early/mid and late infancy. LPA and CA solutions produced relatively comparable types for younger and older infants. Results are discussed in the context of developmental changes unique to the end of the first year of life, which likely account for the present findings. (PsycINFO Database Record

Geospatial Ecology of Adolescent Problem Behavior: Contributions of Community Factors and Parental Monitoring.

Addressed the ecology of deviant peer involvement, antisocial behavior and alcohol use, utilizing publically available information for indices of community risk/protective factors. A geospatial model was developed, combining geographic data (census, crime proximity, race/ethnicity, transportation accessibility) with information gathered for individual adolescents/household, geo-coded by home address. Adolescent-report of delinquency, association with deviant peers, substance use, and parental monitoring was obtained, along with parent-report of demographic characteristics. Deviant peer involvement was predicted by the Crime Proximity Index, with closeness of crime being associated with more deviant peer affiliation, as well as the Transportation Index, with greater accessibility leading to more involvement with troubled peers. Antisocial behaviors also increased with greater access to transportation. Adolescent alcohol use was lower in communities with a higher proportion of a non-Caucasian population, and increased with greater transportation access. Adolescent outcomes were associated with different prediction models, yet parental monitoring emerged as a consistent contributing factor.