Validation of multilevel regression and poststratification methodology for small area estimation of health indicators from the behavioral risk factor surveillance system

Correlations of Socioeconomic and Clinical Determinants with United States County-Level Stroke Prevalence

Socioeconomic status (SES) is a multi-faceted theoretical construct associated with stroke risk and outcomes. Knowing which SES measures best correlate with population stroke metrics would improve its accounting in observational research and inform interventions. Using the Centers for Disease Control and Prevention's (CDC) Population Level Analysis and Community Estimates (PLACES) and other publicly available databases, we conducted an ecological study comparing correlations of different United States county-level SES, health care access and clinical risk factor measures with age-adjusted stroke prevalence. The prevalence of adults living below 150% of the federal poverty level most strongly correlated with stroke prevalence compared to other SES and non-SES measures (correlation coefficient = 0.908, R2 = 0.825; adjusted partial correlation coefficient: 0.589, R2 = 0.347). ANN NEUROL 2024;96:739-744.

Australian Youth Self-Harm Atlas: spatial modelling and mapping of self-harm prevalence and related risk and protective factors to inform youth suicide prevention strategies

AIMS

Suicide prevention strategies have shifted in many countries, from a national approach to one that is regionally tailored and responsive to local community needs. Previous Australian studies support this approach. However, most studies have focused on suicide deaths which may not fully capture a complete understanding of prevention needs, and few have focused on the priority population of youth. This was the first nationwide study to examine regional variability of self-harm prevalence and related factors in Australian young people.

METHODS

A random sample of Australian adolescents (12-17-year-olds) were recruited as part of the Young Minds Matter (YMM) survey. Participants completed self-report questions on self-harm (i.e., non-suicidal self-harm and suicide attempts) in the previous 12 months. Using mixed effects regressions, an area-level model was built with YMM and Census data to produce out-of-sample small area predictions for self-harm prevalence. Spatial unit of analysis was Statistical Area Level 1 (average population 400 people), and all prevalence estimates were updated to 2019.

RESULTS

Across Australia, there was large variability in youth self-harm prevalence estimates. Northern Territory, Western Australia, and South Australia had the highest estimated state prevalence. Psychological distress and depression were factors which best predicted self-harm at an individual level. At an area-level, the strongest predictor was a high percentage of single unemployed parents, while being in an area where ≥30% of parents were born overseas was associated with reduced odds of self-harm.

CONCLUSIONS

This study identified characteristics of regions with lower and higher youth self-harm risk. These findings should assist governments and communities with developing and implementing regionally appropriate youth suicide prevention interventions and initiatives.

Spatial and Temporal Patterns of Chronic Disease Burden in the U.S., 2018-2021

INTRODUCTION

Chronic diseases are primary causes of mortality and disability in the U.S. Although individual-level indices to assess the burden of multiple chronic diseases exist, there is a lack of quantitative tools at the population level. This gap hinders the understanding of the geographical distribution and impact of chronic diseases, crucial for effective public health strategies. This study aims to construct a Chronic Disease Burden Index (CDBI) for evaluating county-level disease burden, to identify geographic and temporal patterns, and investigate the association between CDBI and social vulnerability.

METHODS

A total of 20 health measures from CDC's PLACES database (2018-2021) were used to construct annual county-level CDBIs through principal component analysis. Geographic hotspots of chronic disease burden were identified using Getis-Ord Gi*. Multinomial logistic regression models and bivariate maps were used to assess the association between CDBI and CDC's social vulnerability index. Analyses were conducted in 2023-2024.

RESULTS

Counties with high chronic disease burden were predominantly clustered in the southern U.S. High persistent chronic disease burden was prevalent in Kentucky and West Virginia, while increased burden was observed in Ohio and Texas. Chronic disease burden was highly associated with social vulnerability index (ORQ5 vs Q1=7.6, 95% CI: [6.6, 8.8]), with nonmetro-urban counties experiencing elevated CDBI (OR=14.6, 95% CI: [9.7, 21.9]).

CONCLUSIONS

The CDBI offers an effective tool for assessing chronic disease burden at the population level. Identifying high-burden and vulnerable communities is a crucial first step toward facilitating resource allocation to enhance equitable healthcare access and advancing understanding of health disparities.

Prevalence and regional distribution of obstructive sleep apnea in Canada: Analysis from the Canadian Longitudinal Study on Aging

OBJECTIVES

Obstructive sleep apnea (OSA) is a common chronic condition that is often undiagnosed or diagnosed after many years of symptoms and has an impact on quality of life and several health factors. We estimated the Canadian national prevalence of OSA using a validated questionnaire and physical measurements in participants in the Canadian Longitudinal Study on Aging (CLSA).

METHODS

The method used individual risk estimation based upon the validated STOP-BANG scale developed for OSA. This stratified population sample spans Canada to provide regional estimates.

RESULTS

In this sample of adults aged 45 to 85 years old, the overall prevalence in 2015 of combined moderate and severe OSA in the 51,337 participants was 28.1% (95% confidence intervals, 27.8‒28.4). The regional prevalence varied statistically between Atlantic Canada and Western Canada (p < 0.001), although clinically the variations were limited. The provincial prevalence for moderate and severe OSA ranged from 27.5% (New Brunswick and British Columbia) to 29.1% (Manitoba). Body mass index (BMI) was the dominant determinant of the variance between provinces (β = 0.33, p < 0.001). Only 1.2% of participants had a clinical diagnosis of OSA.

CONCLUSION

The great majority (92.9%) of the participants at high risk of OSA were unrecognized and had no clinical diagnosis of OSA.

Measuring variation in infant mortality and deaths of despair by US congressional district in Pennsylvania: a methodological case study

Many ecological studies examine health outcomes and disparities using administrative boundaries such as census tracts, counties, or states. These boundaries help us to understand the patterning of health by place, along with impacts of policies implemented at these levels. However, additional geopolitical units (units with both geographic and political meaning), such as congressional districts (CDs), present further opportunities to connect research with public policy. Here we provide a step-by-step guide on how to conduct disparities-focused analysis at the CD level. As an applied case study, we use geocoded vital statistics data from 2010-2015 to examine levels of and disparities in infant mortality and deaths of despair in the 19 US CDs of Pennsylvania for the 111th-112th (2009-2012) Congresses and 18 CDs for the 113th-114th (2013-2016) Congresses. We also provide recommendations for extending CD-level analysis to other outcomes, states, and geopolitical boundaries, such as state legislative districts. Increased surveillance of health outcomes at the CD level can help prompt policy action and advocacy and, hopefully, reduce rates of and disparities in adverse health outcomes.

Electronic Health Records Versus Survey Small Area Estimates for Public Health Surveillance

INTRODUCTION

Electronic health records (EHRs) are increasingly being leveraged for public health surveillance. EHR-based small area estimates (SAEs) are often validated by comparison to survey data such as the Behavioral Risk Factor Surveillance System (BRFSS). However, survey and EHR-based SAEs are expected to differ. In this cross-sectional study, SAEs were generated using MDPHnet, a distributed EHR-based surveillance network, for all Massachusetts municipalities and zip code tabulation areas (ZCTAs), compared to BRFSS PLACES SAEs, and reasons for differences explored.

METHODS

This study delineated reasons a priori for how SAEs derived using EHRs may differ from surveys by comparing each strategy's case classification criteria and reviewing the literature. Hypertension, diabetes, obesity, asthma, and smoking EHR-based SAEs for 2021 in all ZCTAs and municipalities in Massachusetts were estimated with Bayesian mixed effects modeling and poststratification in the summer/fall of 2023. These SAEs were compared to BRFSS PLACES SAEs published by the U.S. Centers for Disease Control and Prevention.

RESULTS

Mean prevalence was higher in EHR data versus BRFSS in both municipalities and ZCTAs for all outcomes except asthma. ZCTA and municipal symmetric mean absolute percentages ranged from 12.0 to 38.2% and 13.1 to 39.8%, respectively. There was greater variability in EHR-based SAEs versus BRFSS PLACES in both municipalities and ZCTAs.

CONCLUSIONS

EHR-based SAEs tended to be higher than BRFSS and more variable. Possible explanations include detection of undiagnosed cases and over-classification using EHR data, and under-reporting within BRFSS. Both EHR and survey-based surveillance have strengths and limitations that should inform their preferred uses in public health surveillance.

Access to Prostate-Specific Antigen Testing and Mortality Among Men With Prostate Cancer

Importance

Prostate-specific antigen (PSA) screening for prostate cancer is controversial but may be associated with benefit for certain high-risk groups.

Objectives

To evaluate associations of county-level PSA screening prevalence with prostate cancer outcomes, as well as variation by sociodemographic and clinical factors.

Design, Setting, and Participants

This cohort study used data from cancer registries based in 8 US states on Hispanic, non-Hispanic Black, and non-Hispanic White men aged 40 to 99 years who received a diagnosis of prostate cancer between January 1, 2000, and December 31, 2015. Participants were followed up until death or censored after 10 years or December 31, 2018, whichever end point came first. Data were analyzed between September 2023 and January 2024.

Exposure

County-level PSA screening prevalence was estimated using the Behavior Risk Factor Surveillance System survey data from 2004, 2006, 2008, 2010, and 2012 and weighted by population characteristics.

Main Outcomes and Measures

Multivariable logistic, Cox proportional hazards regression, and competing risks models were fit to estimate adjusted odds ratios (AOR) and adjusted hazard ratios (AHR) for associations of county-level PSA screening prevalence at diagnosis with advanced stage (regional or distant), as well as all-cause and prostate cancer-specific survival.

Results

Of 814 987 men with prostate cancer, the mean (SD) age was 67.3 (9.8) years, 7.8% were Hispanic, 12.2% were non-Hispanic Black, and 80.0% were non-Hispanic White; 17.0% had advanced disease. There were 247 570 deaths over 5 716 703 person-years of follow-up. Men in the highest compared with lowest quintile of county-level PSA screening prevalence at diagnosis had lower odds of advanced vs localized stage (AOR, 0.86; 95% CI, 0.85-0.88), lower all-cause mortality (AHR, 0.86; 95% CI, 0.85-0.87), and lower prostate cancer-specific mortality (AHR, 0.83; 95% CI, 0.81-0.85). Inverse associations between PSA screening prevalence and advanced cancer were strongest among men of Hispanic ethnicity vs other ethnicities (AOR, 0.82; 95% CI, 0.78-0.87), older vs younger men (aged ≥70 years: AOR, 0.77; 95% CI, 0.75-0.79), and those in the Northeast vs other US Census regions (AOR, 0.81; 95% CI, 0.79-0.84). Inverse associations with all-cause mortality were strongest among men of Hispanic ethnicity vs other ethnicities (AHR, 0.82; 95% CI, 0.78-0.85), younger vs older men (AHR, 0.81; 95% CI, 0.77-0.85), those with advanced vs localized disease (AHR, 0.80; 95% CI, 0.78-0.82), and those in the West vs other US Census regions (AHR, 0.89; 95% CI, 0.87-0.90).

Conclusions and Relevance

This population-based cohort study of men with prostate cancer suggests that higher county-level prevalence of PSA screening was associated with lower odds of advanced disease, all-cause mortality, and prostate cancer-specific mortality. Associations varied by age, race and ethnicity, and US Census region.

Hospital Capital Assets, Community Health, and the Utilization and Cost of Inpatient Care: A Population-Based Study of US Counties

BACKGROUND

The provision of high-quality hospital care requires adequate space, buildings, and equipment, although redundant infrastructure could also drive service overprovision.

OBJECTIVE

To explore the distribution of physical hospital resources-that is, capital assets-in the United States; its correlation with indicators of community health and nonhealth factors; and the association between hospital capital density and regional hospital utilization and costs.

RESEARCH DESIGN

We created a dataset of n=1733 US counties by analyzing the 2019 Medicare Cost Reports; 2019 State Inpatient Database Community Inpatient Statistics; 2020-2021 Area Health Resource File; 2016-2020 American Community Survey; 2022 PLACES; and 2019 CDC WONDER. We first calculated aggregate hospital capital assets and investment at the county level. Next, we examined the correlation between community's medical need (eg, chronic disease prevalence), ability to pay (eg, insurance), and supply factors with 4 metrics of capital availability. Finally, we examined the association between capital assets and hospital utilization/costs, adjusted for confounders.

RESULTS

Counties with older and sicker populations generally had less aggregate hospital capital per capita, per hospital day, and per hospital discharge, while counties with higher income or insurance coverage had more hospital capital. In linear regressions controlling for medical need and ability to pay, capital assets were associated with greater hospital utilization and costs, for example, an additional $1000 in capital assets per capita was associated with 73 additional discharges per 100,000 population (95% CI: 45-102) and $19 in spending per bed day (95% CI: 12-26).

CONCLUSIONS

The level of investment in hospitals is linked to community wealth but not population health needs, and may drive use and costs.

The Association of County-level Prostate-specific Antigen Screening with Metastatic Prostate Cancer and Prostate Cancer Mortality

BACKGROUND AND OBJECTIVE

There exists ongoing debate about the benefits and harms of prostate-specific antigen (PSA) screening for prostate cancer. This study sought to evaluate the association of county-level PSA screening rates with county-level incidence of metastatic prostate cancer and prostate cancer mortality in the USA.

METHODS

This ecological study used data from the 2004-2012 Behavioral Risk Factor Surveillance System (BRFSS) to build a multilevel mixed-effect model with poststratification using US Census data to estimate county-level PSA screening rates for all 3143 US counties adjusted for age, race, ethnicity, and county-level poverty rates. The exposure of interest was average county-level PSA screening rate from 2004 to 2012, defined as the proportion of men aged 40-79 yr who underwent PSA screening within the prior 2 yr. The primary outcomes were county-level age-adjusted incidence of regional/distant prostate cancer during 2015-2019 and age-adjusted prostate cancer mortality during 2016-2020.

KEY FINDINGS AND LIMITATIONS

A total of 416 221 male BRFSS respondents aged 40-79 yr met the inclusion criteria and were used in the multilevel mixed-effect model. The model was poststratified using 63.4 million men aged 40-79 yr from all 3143 counties in the 2010 Decennial Census. County-level estimated PSA screening rates exhibited geographic variability and were pooled at the state level for internal validation with direct BRFSS state-level estimates, showing a strong correlation with Pearson correlation coefficients 0.77-0.90. A 10% higher county-level probability of PSA screening in 2004-2012 was associated with a 14% lower county-level incidence of regional/distant prostate cancer in 2015-2019 (rate ratio 0.86, 95% confidence interval [CI] 0.85-0.87, p < 0.001) and 10% lower county-level prostate cancer mortality in 2016-2020 (rate ratio 0.90, 95% CI 0.89-0.91, p < 0.001).

CONCLUSIONS AND CLINICAL IMPLICATIONS

In this population-based ecological study of all US counties, higher PSA screening rates were associated with a lower incidence of regional/distant prostate cancer and lower prostate cancer mortality at extended follow-up.

PATIENT SUMMARY

US counties with higher rates of prostate-specific antigen (PSA) screening had significantly lower rates of metastatic prostate cancer and prostate cancer mortality in subsequent years. These data may inform shared decision-making regarding PSA screening for prostate cancer.

Using leave-one-out cross validation (LOO) in a multilevel regression and poststratification (MRP) workflow: A cautionary tale

In recent decades, multilevel regression and poststratification (MRP) has surged in popularity for population inference. However, the validity of the estimates can depend on details of the model, and there is currently little research on validation. We explore how leave-one-out cross validation (LOO) can be used to compare Bayesian models for MRP. We investigate two approximate calculations of LOO: Pareto smoothed importance sampling (PSIS-LOO) and a survey-weighted alternative (WTD-PSIS-LOO). Using two simulation designs, we examine how accurately these two criteria recover the correct ordering of model goodness at predicting population and small-area estimands. Focusing first on variable selection, we find that neither PSIS-LOO nor WTD-PSIS-LOO correctly recovers the models' order for an MRP population estimand, although both criteria correctly identify the best and worst model. When considering small-area estimation, the best model differs for different small areas, highlighting the complexity of MRP validation. When considering different priors, the models' order seems slightly better at smaller-area levels. These findings suggest that, while not terrible, PSIS-LOO-based ranking techniques may not be suitable to evaluate MRP as a method. We suggest this is due to the aggregation stage of MRP, where individual-level prediction errors average out. We validate these results by applying to the real world National Health and Nutrition Examination Survey (NHANES) data in the United States. Altogether, these results show that PSIS-LOO-based model validation tools need to be used with caution and might not convey the full story when validating MRP as a method.

Lack of leisure time physical activity and variations in cardiovascular mortality across US communities: a comprehensive county-level analysis (2011-2019)

OBJECTIVES

To investigate the associations between county-level proportions of adults not engaging in leisure-time physical activity (no LTPA) and age-adjusted cardiovascular mortality (AACVM) rates in the overall US population and across demographics.

METHODS

Analysing 2900 US counties from 2011 to 2019, we used the Centers for Disease Control and Prevention (CDC) databases to obtain annual AACVM rates. No LTPA data were sourced from the CDC's Behavioural Risk Factor Surveillance System survey and county-specific rates were calculated using a validated multilevel regression and poststratification modelling approach. Multiple regression models assessed associations with county characteristics such as socioeconomic, environmental, clinical and healthcare access factors. Poisson generalised linear mixed models were employed to calculate incidence rate ratios (IRR) and additional yearly deaths (AYD) per 100 000 persons.

RESULTS

Of 309.9 million residents in 2900 counties in 2011, 7.38 million (2.4%) cardiovascular deaths occurred by 2019. County attributes such as socioeconomic, environmental and clinical factors accounted for up to 65% (adjusted R2=0.65) of variance in no LTPA rates. No LTPA rates associated with higher AACVM across demographics, notably among middle-aged adults (standardised IRR: 1.06; 95% CI (1.04 to 1.07)), particularly women (1.09; 95% CI (1.07 to 1.12)). The highest AYDs were among elderly non-Hispanic black individuals (AYD=68/100 000).

CONCLUSIONS

Our study reveals a robust association between the high prevalence of no LTPA and elevated AACVM rates beyond other social determinants. The most at-risk groups were middle-aged women and elderly non-Hispanic black individuals. Further, county-level characteristics accounted for substantial variance in community LTPA rates. These results emphasise the need for targeted public health measures to boost physical activity, especially in high-risk communities, to reduce AACVM.

Bias-Adjusted Predictions of County-Level Vaccination Coverage from the COVID-19 Trends and Impact Survey

BACKGROUND

The potential for selection bias in nonrepresentative, large-scale, low-cost survey data can limit their utility for population health measurement and public health decision making. We developed an approach to bias adjust county-level COVID-19 vaccination coverage predictions from the large-scale US COVID-19 Trends and Impact Survey.

DESIGN

We developed a multistep regression framework to adjust for selection bias in predicted county-level vaccination coverage plateaus. Our approach included poststratification to the American Community Survey, adjusting for differences in observed covariates, and secondary normalization to an unbiased reference indicator. As a case study, we prospectively applied this framework to predict county-level long-run vaccination coverage among children ages 5 to 11 y. We evaluated our approach against an interim observed measure of 3-mo coverage for children ages 5 to 11 y and used long-term coverage estimates to monitor equity in the pace of vaccination scale up.

RESULTS

Our predictions suggested a low ceiling on long-term national vaccination coverage (46%), detected substantial geographic heterogeneity (ranging from 11% to 91% across counties in the United States), and highlighted widespread disparities in the pace of scale up in the 3 mo following Emergency Use Authorization of COVID-19 vaccination for 5- to 11-y-olds.

LIMITATIONS

We relied on historical relationships between vaccination hesitancy and observed coverage, which may not capture rapid changes in the COVID-19 policy and epidemiologic landscape.

CONCLUSIONS

Our analysis demonstrates an approach to leverage differing strengths of multiple sources of information to produce estimates on the time scale and geographic scale necessary for proactive decision making.

IMPLICATIONS

Designing integrated health measurement systems that combine sources with different advantages across the spectrum of timeliness, spatial resolution, and representativeness can maximize the benefits of data collection relative to costs.

HIGHLIGHTS

The COVID-19 pandemic catalyzed massive survey data collection efforts that prioritized timeliness and sample size over population representativeness.The potential for selection bias in these large-scale, low-cost, nonrepresentative data has led to questions about their utility for population health measurement.We developed a multistep regression framework to bias adjust county-level vaccination coverage predictions from the largest public health survey conducted in the United States to date: the US COVID-19 Trends and Impact Survey.Our study demonstrates the value of leveraging differing strengths of multiple data sources to generate estimates on the time scale and geographic scale necessary for proactive public health decision making.

Neighborhood Environmental Burden and Cardiovascular Health in the US

Importance

Cardiovascular disease is the leading cause of death in the US. However, little is known about the association between cumulative environmental burden and cardiovascular health across US neighborhoods.

Objective

To evaluate the association of neighborhood-level environmental burden with prevalence of cardiovascular risk factors and diseases, overall and by levels of social vulnerability.

Design, Settings, and Participants

This was a national cross-sectional study of 71 659 US Census tracts. Environmental burden (EBI) and social vulnerability indices from the US Centers for Disease Control and Prevention (CDC) and Agency for Toxic Substances and Disease Registry were linked to the 2020 CDC PLACES data set. Data were analyzed from March to October 2023.

Exposures

The EBI, a measure of cumulative environmental burden encompassing 5 domains (air pollution, hazardous or toxic sites, built environment, transportation infrastructure, and water pollution).

Main Outcomes and Measures

Neighborhood-level prevalence of cardiovascular risk factors (hypertension, diabetes, and obesity) and cardiovascular diseases (coronary heart disease and stroke).

Results

Across the US, neighborhoods with the highest environmental burden (top EBI quartile) were more likely than those with the lowest environmental burden (bottom EBI quartile) to be urban (16 626 [92.7%] vs 13 414 [75.4%]), in the Midwest (5191 [28.9%] vs 2782 [15.6%]), have greater median (IQR) social vulnerability scores (0.64 [0.36-0.85] vs 0.42 [0.20-0.65]), and have higher proportions of adults in racial or ethnic minority groups (median [IQR], 34% [12-73] vs 12% [5-30]). After adjustment, neighborhoods with the highest environmental burden had significantly higher rates of cardiovascular risk factors than those with the lowest burden, including hypertension (mean [SD], 32.83% [7.99] vs 32.14% [6.99]; adjusted difference, 0.84%; 95% CI, 0.71-0.98), diabetes (mean [SD], 12.19% [4.33] vs 10.68% [3.27]; adjusted difference, 0.62%; 95% CI, 0.53-0.70), and obesity (mean [SD], 33.57% [7.62] vs 30.86% [6.15]; adjusted difference, 0.77%; 95% CI, 0.60-0.94). Similarly, neighborhoods with the highest environmental burden had significantly higher rates of coronary heart disease (mean [SD], 6.66% [2.15] vs 6.82% [2.41]; adjusted difference, 0.28%; 95% CI, 0.22-0.33) and stroke (mean [SD], 3.65% [1.47] vs 3.31% [1.12]; adjusted difference, 0.19%; 95% CI, 0.15-0.22). Results were consistent after matching highest and lowest environmentally burdened neighborhoods geospatially and based on other covariates. The associations between environmental burden quartiles and cardiovascular risk factors and diseases were most pronounced among socially vulnerable neighborhoods.

Conclusions and Relevance

In this cross-sectional study of US neighborhoods, cumulative environmental burden was associated with higher rates of cardiovascular risk factors and diseases, although absolute differences were small. The strongest associations were observed in socially vulnerable neighborhoods. Whether initiatives that address poor environmental conditions will improve cardiovascular health requires additional prospective investigations.

Embedded multilevel regression and poststratification: Model-based inference with incomplete auxiliary information

Health disparity research often evaluates health outcomes across demographic subgroups. Multilevel regression and poststratification (MRP) is a popular approach for small subgroup estimation as it can stabilize estimates by fitting multilevel models and adjust for selection bias by poststratifying on auxiliary variables, which are population characteristics predictive of the analytic outcome. However, the granularity and quality of the estimates produced by MRP are limited by the availability of the auxiliary variables' joint distribution; data analysts often only have access to the marginal distributions. To overcome this limitation, we embed the estimation of population cell counts needed for poststratification into the MRP workflow: embedded MRP (EMRP). Under EMRP, we generate synthetic populations of the auxiliary variables before implementing MRP. All sources of estimation uncertainty are propagated with a fully Bayesian framework. Through simulation studies, we compare different methods of generating the synthetic populations and demonstrate EMRP's improvements over alternatives on the bias-variance tradeoff to yield valid subpopulation inferences of interest. We apply EMRP to the Longitudinal Survey of Wellbeing and estimate food insecurity prevalence among vulnerable groups in New York City. We find that all EMRP estimators can correct for the bias in classical MRP while maintaining lower standard errors and narrower confidence intervals than directly imputing with the weighted finite population Bayesian bootstrap (WFPBB) and design-based estimates. Performances from the EMRP estimators do not differ substantially from each other, though we would generally recommend using the WFPBB-MRP for its consistently high coverage rates.

Mapping the prevalence of cancer risk factors at the small area level in Australia

BACKGROUND

Cancer is a significant health issue globally and it is well known that cancer risk varies geographically. However in many countries there are no small area-level data on cancer risk factors with high resolution and complete reach, which hinders the development of targeted prevention strategies.

METHODS

Using Australia as a case study, the 2017-2018 National Health Survey was used to generate prevalence estimates for 2221 small areas across Australia for eight cancer risk factor measures covering smoking, alcohol, physical activity, diet and weight. Utilising a recently developed Bayesian two-stage small area estimation methodology, the model incorporated survey-only covariates, spatial smoothing and hierarchical modelling techniques, along with a vast array of small area-level auxiliary data, including census, remoteness, and socioeconomic data. The models borrowed strength from previously published cancer risk estimates provided by the Social Health Atlases of Australia. Estimates were internally and externally validated.

RESULTS

We illustrated that in 2017-2018 health behaviours across Australia exhibited more spatial disparities than previously realised by improving the reach and resolution of formerly published cancer risk factors. The derived estimates revealed higher prevalence of unhealthy behaviours in more remote areas, and areas of lower socioeconomic status; a trend that aligned well with previous work.

CONCLUSIONS

Our study addresses the gaps in small area level cancer risk factor estimates in Australia. The new estimates provide improved spatial resolution and reach and will enable more targeted cancer prevention strategies at the small area level. Furthermore, by including the results in the next release of the Australian Cancer Atlas, which currently provides small area level estimates of cancer incidence and relative survival, this work will help to provide a more comprehensive picture of cancer in Australia by supporting policy makers, researchers, and the general public in understanding the spatial distribution of cancer risk factors. The methodology applied in this work is generalisable to other small area estimation applications and has been shown to perform well when the survey data are sparse.

Geographic Distribution of Social Service Resources for Stroke Survivors in Ohio Varies by Rurality

BACKGROUND

Both social service resources and stroke prevalence vary by geography, and health care resources are scarcer in rural areas. We assessed whether distributions of resources relevant to stroke survivors were clustered around areas of the highest stroke prevalence in Ohio and whether this is varied by rurality using an ecological study design.

METHODS

Census tract (CT)-level self-reported stroke prevalence estimates (Centers for Disease Control and Prevention PLACES-2019 Behavioral Risk Factor Surveillance System) were linked with sociodemographic and rurality data (2019 American Community Survey) and geographic density of resources in Ohio (2020 findhelp data). Resources were grouped into categories: housing, in-home, financial, transportation, education, and therapy. Negative binomial regression models estimated the mean number of resources within 25 miles and 30 minutes of a CT centroid and quartiles of stroke prevalence for each resource group by rurality status (rural, urban, and suburban). Models were sequentially adjusted for total population and CT demographics.

RESULTS

In Ohio, stroke prevalence was 3.9% (0.4%-14.2%). The highest stroke prevalence quartile (versus lowest) was associated with fewer resources within 25 miles overall (resource ratio [RR], 0.57-0.98). The most pronounced disparities were in rural CT; rural CTs with the highest quartile stroke prevalence had fewer housing (RR, 0.49 [95% CI, 0.32-0.75]), in-home (RR, 0.31 [95% CI, 0.20-0.49]), and therapy (RR, 0.23 [95% CI, 0.13-0.43]) resources compared with those with the lowest quartile stroke prevalence (reference: mean, 1.2 housing, 5.1 in-home, and 4.9 therapy resources, respectively). Rural disparities no longer persisted after adjustment for federal poverty limit (rural: housing [RR, 0.69 (95% CI, 0.40-1.20)], in-home [RR, 0.65 (95% CI, 0.34-1.23)], and therapy [RR, 0.66 (95% CI, 0.33-1.32)]).

CONCLUSIONS

Stroke social service resources are inversely distributed relative to stroke prevalence in Ohio, particularly in rural areas. This inverse link in rural Ohio is likely explained by geographic differences in poverty. Stroke-specific resource-related interventions may be needed and should consider the roles of rurality and poverty.

Integrating human activity into food environments can better predict cardiometabolic diseases in the United States

The prevalence of cardiometabolic diseases in the United States is presumably linked to an obesogenic retail food environment that promotes unhealthy dietary habits. Past studies, however, have reported inconsistent findings about the relationship between the two. One underexplored area is how humans interact with food environments and how to integrate human activity into scalable measures. In this paper, we develop the retail food activity index (RFAI) at the census tract level by utilizing Global Positioning System tracking data covering over 94 million aggregated visit records to approximately 359,000 food retailers across the United States over two years. Here we show that the RFAI has significant associations with the prevalence of multiple cardiometabolic diseases. Our study indicates that the RFAI is a promising index with the potential for guiding the development of policies and health interventions aimed at curtailing the burden of cardiometabolic diseases, especially in communities characterized by obesogenic dietary behaviors.

Measuring The Enduring Imprint Of Structural Racism On American Neighborhoods

A long history of discriminatory policies in the United States has created disparities in neighborhood resources that shape ethnoracial health inequities today. To quantify these differences, we organized publicly available data on forty-two variables at the census tract level within nine domains affected by structural racism: built environment, criminal justice, education, employment, housing, income and poverty, social cohesion, transportation, and wealth. Using data from multiple sources at several levels of geography, we developed scores in each domain, as well as a summary score that we call the Structural Racism Effect Index. We examined correlations with life expectancy and other measures of health for this index and other commonly used area-based indices. The Structural Racism Effect Index was more strongly associated with each health outcome than were the other indices. Its domain and summary scores can be used to describe differences in social risk factors, and they provide powerful new tools to guide policies and investments to advance health equity.

Risk Assessment of Passenger Behaviors That Influence Accident Type and Severity in Metro Operation

Background

The unsafe behavior of passengers frequently causes metro operation accidents. This research aims to establish a model for evaluating the risk of unsafe behavior among subway passengers and for assessing the severity of different types of accidents caused by passenger unsafe behavior.

Methods

A risk assessment model that combines the Interaction Matrix (IM) model with a Monte Carlo algorithm was established to quantitatively test the risk of unsafe behavior among passengers. Based on the initial data of 234 cases, the behavioral risks in accidents were simulated, and the resulting risks follow a normal distribution. After analyzing the differences in behavioral risk distribution characteristics, the targeted risk mitigation countermeasures were obtained.

Results

Results showed that there are 12 kinds of unsafe behaviors related to 4 metro operation accident types. Among them, crowded stampede caused by four kinds of passengers' unsafe behavior has the highest risk mean (μ) of 5.14, followed by escalator injury (4.72), pinched by a shielding barrier (4.42) and fall injury (4.14).

Conclusion

The severity of different types of accidents caused by different unsafe behaviors of passengers was obtained, which can provide a basis for targeted risk mitigation strategies and measures.

Validation of a small-area model for estimation of smoking prevalence at a subnational level

INTRODUCTION

Small-area estimation methods are an alternative to direct survey-based estimates in cases where a survey's sample size does not suffice to ensure representativeness. Nevertheless, the information yielded by small-area estimation methods must be validated. The objective of this study was thus to validate a small-area model.

METHODS

The prevalence of smokers, ex-smokers, and never smokers by sex and age group (15-34, 35-54, 55-64, 65-74, ≥75 years) was calculated in two Spanish Autonomous Regions (ARs) by applying a weighted ratio estimator (direct estimator) to data from representative surveys. These estimates were compared against those obtained with a small-area model applied to another survey, specifically the Spanish National Health Survey, which did not guarantee representativeness for these two ARs by sex and age. To evaluate the concordance of the estimates, we calculated the intraclass correlation coefficient (ICC) and the 95% confidence intervals of the differences between estimates. To assess the precision of the estimates, the coefficients of variation were obtained.

RESULTS

In all cases, the ICC was ≥0.87, indicating good concordance between the direct and small-area model estimates. Slightly more than eight in ten 95% confidence intervals for the differences between estimates included zero. In all cases, the coefficient of variation of the small-area model was <30%, indicating a good degree of precision in the estimates.

CONCLUSIONS

The small-area model applied to national survey data yields valid estimates of smoking prevalence by sex and age group at the AR level. These models could thus be applied to a single year's data from a national survey, which does not guarantee regional representativeness, to characterize various risk factors in a population at a subnational level.

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.

County-Level Geographic Disparities in Disabilities Among US Adults, 2018

INTRODUCTION

Local data are increasingly needed for public health practice. County-level data on disabilities can be a valuable complement to existing estimates of disabilities. The objective of this study was to describe the county-level prevalence of disabilities among US adults and identify geographic clusters of counties with a higher or lower prevalence of disabilities.

METHODS

We applied a multilevel logistic regression and poststratification approach to geocoded 2018 Behavioral Risk Factor Surveillance System data, Census 2018 county-level population estimates, and American Community Survey 2014-2018 poverty estimates to generate county-level estimates for 6 functional disabilities and any disability type. We used cluster-outlier spatial statistical methods to identify clustered counties.

RESULTS

Among 3,142 counties, median estimated prevalence was 29.5% for any disability and differed by type: hearing (8.0%), vision (4.9%), cognition (11.5%), mobility (14.9%), self-care (3.7%), and independent living (7.2%). The spatial autocorrelation statistic, Moran's I, was 0.70 for any disability and 0.60 or greater for all 6 types of disability, indicating that disabilities were highly clustered at the county level. We observed similar spatial cluster patterns in all disability types except hearing disability.

CONCLUSION

The results suggest substantial differences in disability prevalence across US counties. These data, heretofore unavailable from a health survey, may help with planning programs at the county level to improve the quality of life for people with disabilities.

Comparison of Lung Cancer Screening Eligibility and Use between Commercial, Medicare, and Medicare Advantage Enrollees

OBJECTIVE

Lung cancer screening does not require patient cost-sharing for insured people in the U.S. Little is known about whether other factors associated with patient selection into different insurance plans affect screening rates. We examined screening rates for enrollees in commercial, Medicare Fee-for-Service (FFS), and Medicare Advantage plans.

METHODS

County-level smoking rates from the 2017 County Health Rankings were used to estimate the number of enrollees eligible for lung cancer screening in two large retrospective claims databases covering: a 5% national sample of Medicare FFS enrollees; and 100% sample of enrollees associated with large commercial and Medicare Advantage carriers. Screening rates were estimated using observed claims, stratified by payer, before aggregation into national estimates by payer and demographics. Chi-square tests were used to examine differences in screening rates between payers.

RESULTS

There were 1,077,142 enrollees estimated to be eligible for screening. The overall estimated screening rate for enrollees by payer was 1.75% for commercial plans, 3.37% for Medicare FFS, and 4.56% for Medicare Advantage plans. Screening rates were estimated to be lowest among females (1.55%-4.02%), those aged 75-77 years (0.63%-2.87%), those residing in rural areas (1.88%-3.56%), and those in the West (1.16%-3.65%). Among Medicare FFS enrollees, screening rates by race/ethnicity were non-Hispanic White (3.71%), non-Hispanic Black (2.17%) and Other (1.68%).

CONCLUSIONS

Considerable variation exists in lung cancer screening between different payers and across patient characteristics. Efforts targeting historically vulnerable populations could present opportunities to increase screening.

Geographic and demographic variation in worry about extreme heat and COVID-19 risk in summer 2020

Extreme heat is a major health hazard that is exacerbated by ongoing human-caused climate change. However, how populations perceive the risks of heat in the context of other hazards like COVID-19, and how perceptions vary geographically, are not well understood. Here we present spatially explicit estimates of worry among the U.S. public about the risks of heat and COVID-19 during the summer of 2020, using nationally representative survey data and a multilevel regression and poststratification (MRP) model. Worry about extreme heat and COVID-19 varies both across states and across demographic groups, in ways that reflect disparities in the impact of each risk. Black or African American and Hispanic or Latino populations, who face greater health impacts from both COVID-19 and extreme heat due to institutional and societal inequalities, also tend to be much more worried about both risks than white, non-Hispanic populations. Worry about heat and COVID-19 were correlated at the individual and population level, and patterns tended to be related to underlying external factors associated with the risk environment. In the face of a changing climate there is an urgent need to address disparities in heat risk and develop responses that ensure the most at-risk populations are protected.

Model-based community health surveillance via multilevel small area estimation using state behavioral risk factor surveillance system (BRFSS): a case study in Connecticut

PURPOSE

The main objective of state behavioral risk factor surveillance system (BRFSS) is to produce reliable state-level estimates of various population health outcomes. A multilevel Regression and Post-stratification (MRP) methodology for small area estimation has been applied to the 500 Cities Project to provide population estimates at both city-level and census tract-level using national BRFSS data. To date, MRP has not been applied to any state BRFSS to produce health data at local geographic areas. In addition, the use of single year BRFSS might produce temporary inconsistency in small area estimates (SAEs). The predicted standard errors (SEs) and confidence intervals (CIs) of SAEs using Monte Carlo simulation could be substantially underestimated or overestimated.

METHODS

By extending the current MRP approach and applying a parametric bootstrapping approach to Connecticut BRFSS (CT BRFSS), we were able to produce SAEs as well as SEs and CIs of SAEs for Connecticut counties and towns. We also applied this model to 5-year CT BRFSS (2011-2015) with an aim to improve the temporary consistency of SAEs.

RESULTS

Both single-year and 5-year estimates with SEs and CIs were generated for six selected population health indicators at town, county and state levels. Model-based SAEs were internally evaluated by comparing to single-year and 5-year direct BRFSS survey (2011-2015). SAEs were also externally validated when external data were available.

CONCLUSIONS

Model-based SAEs are valid and could be used to characterize local geographic variations using single state BRFSS data.

Multilevel Small Area Estimation for County-Level Prevalence of Mammography Use in the United States Using 2018 Data

Background: The U.S. Preventive Services Task Force recommends biennial screening mammography for average-risk women aged 50-74 years. We aim to generate county-level prevalence estimates for mammography use to examine disparities among counties. Materials and Methods: We used data from the 2018 Behavioral Risk Factor Surveillance System (BRFSS) (n = 111,902 women) and linked them to county-level data from the American Community Survey. We defined two outcomes: mammography within the past 2 years (current); and mammography 5 or more years ago or never (rarely or never). We poststratified the data with U.S. Census estimated county population counts, ran Monte Carlo simulations, and generated county-level estimates. We aggregated estimates to state and national levels. We validated internal consistency between our model-based and BRFSS state estimates using Spearman and Pearson correlation coefficients. Results: Nationally, more than three in four women [78.7% (95% confidence interval {CI}: 78.2%-79.2%)] were current with mammography, although with large variations among counties. Also, nationally, about one in nine women [11% (95% CI: 10.8%-11.3%)] rarely or never had a mammogram. County estimates for being current ranged from 60.4% in New Mexico to 86.9% in Hawaii. Rarely or never having a mammogram ranged from 6% in Connecticut to 23.0% in Alaska, and on average, almost one in eight women in all the counties. Internal consistency correlation coefficient tests were ≥0.94. Conclusions: Our analyses identified marked county variations in mammography use across the country among women aged 50-74 years. We generated estimates for all counties, which may be helpful for targeted outreach to increase mammography uptake.

Colorectal Cancer Screening Disparities Among Race: A Zip Code Level Analysis

BACKGROUND

Colorectal cancer (CRC) screening can prevent disease by early identification. Existing disparities in CRC screening have been associated with factors including race, socioeconomic status, insurance, and even geography. Our study takes a deeper look into how social determinants related to zip code tabulation areas affect CRC screenings.

MATERIALS AND METHODS

We conducted a retrospective cross-sectional study of CRC screenings by race at a zip code level, evaluating for impactful social determinant factors such as the social deprivation index (SDI). We used publicly available data from CDC 500 Cities Project (2016-2019), PLACES Project (2020), and the American Community Survey (2019). We conducted multivariate and confirmatory factor analyses among race, income, health insurance, check-up visits, and SDI.

RESULTS

Increasing the tertile of SDI was associated with a higher likelihood of being Black or Hispanic, as well as decreased median household income (P < .01). Lower rates of regular checkup visits were found in the third tertile of SDI (P < .01). The multivariate analysis showed that being Black, Hispanic, lower income, being uninsured, lack of regular check-ups, and increased SDI were related to decreased CRC screening. In the confirmatory factor analysis, we found that SDI and access to insurance were the variables most related to decreased CRC screening.

CONCLUSION

Our results reveal the top 2 factors that impact a locality's CRC screening rates are the social deprivation index and access to health care. This data may help implement interventions targeting social barriers to further promote CRC screenings within disadvantaged communities and decrease overall mortality via early screening.

Re-scaling and small area estimation of behavioral risk survey guided by social vulnerability data

BACKGROUND

Local governments and other public health entities often need population health measures at the county or subcounty level for activities such as resource allocation and targeting public health interventions, among others. Information collected via national surveys alone cannot fill these needs. We propose a novel, two-step method for rescaling health survey data and creating small area estimates (SAEs) of smoking rates using a Behavioral Risk Factor Surveillance System survey administered in 2015 to participants living in Allegheny County, Pennsylvania, USA.

METHODS

The first step consisted of a spatial microsimulation to rescale location of survey respondents from zip codes to tracts based on census population distributions by age, sex, race, and education. The rescaling allowed us, in the second step, to utilize available census tract-specific ancillary data on social vulnerability for small area estimation of local health risk using an area-level version of a logistic linear mixed model. To demonstrate this new two-step algorithm, we estimated the ever-smoking rate for the census tracts of Allegheny County.

RESULTS

The ever-smoking rate was above 70% for two census tracts to the southeast of the city of Pittsburgh. Several tracts in the southern and eastern sections of Pittsburgh also had relatively high (> 65%) ever-smoking rates.

CONCLUSIONS

These SAEs may be used in local public health efforts to target interventions and educational resources aimed at reducing cigarette smoking. Further, our new two-step methodology may be extended to small area estimation for other locations and health outcomes.

Heterogeneity in Disparities in Life Expectancy Across US Metropolitan Areas

BACKGROUND

Life expectancy in the United States has declined since 2014 but characterization of disparities within and across metropolitan areas of the country is lacking.

METHODS

Using census tract-level life expectancy from the 2010 to 2015 US Small-area Life Expectancy Estimates Project, we calculate 10 measures of total and income-based disparities in life expectancy at birth, age 25, and age 65 within and across 377 metropolitan statistical areas (MSAs) of the United States.

RESULTS

We found wide heterogeneity in disparities in life expectancy at birth across MSAs and regions: MSAs in the West show the narrowest disparities (absolute disparity: 8.7 years, relative disparity: 1.1), while MSAs in the South (absolute disparity: 9.1 years, relative disparity: 1.1) and Midwest (absolute disparity: 9.8 years, relative disparity: 1.1) have the widest life expectancy disparities. We also observed greater variability in life expectancy across MSAs for lower income census tracts (coefficient of variation [CoV] 3.7 for first vs. tenth decile of income) than for higher income census tracts (CoV 2.3). Finally, we found that a series of MSA-level variables, including larger MSAs and greater proportion college graduates, predicted wider life expectancy disparities for all age groups.

CONCLUSIONS

Sociodemographic and policy factors likely help explain variation in life expectancy disparities within and across metro areas.

Moderating Effects of Racial Segregation on the Associations of Cardiovascular Outcomes with Walkability in Chicago Metropolitan Area

Cardiovascular diseases (CVDs), as the leading cause of death in the U.S., pose a disproportionate burden to racial/ethnic minorities. Walkability, as a key concept of the built environment, reflecting walking and physical activity, is associated with health behaviors that help to reduce CVDs risk. While the unequal social variation and spatial distribution inequality of the CVDs and the role of walkability in preventing CVDs have been explored, the moderating factors through which walkability affects CVDs have not been quantitatively analyzed. In this paper, the spatial statistical techniques combined with the regression model are conducted to study the distribution of the CVDs' health outcomes and factors influencing their variation in the Chicago metropolitan area. The spatial statistical results for the CVDs' health outcomes reveal that clusters of low-value incidence are concentrated in the suburban rural areas and areas on the north side of the city, while the high-value clusters are concentrated in the west and south sides of the city and areas extending beyond the western and southern city boundaries. The regression results indicate that racial segregation reduced the positive association between health outcomes and walkability, although both racial segregation and walkability factors were positively associated with CVDs' health outcomes.

Contemporary Socioeconomic and Childhood Opportunity Disparities in Congenital Heart Surgery

BACKGROUND

While singular measures of socioeconomic status have been associated with outcomes after surgery for congenital heart disease, the multifaceted pathways through which a child's environment impacts similar outcomes remain incompletely characterized. We sought to evaluate the association between childhood opportunity level and adverse outcomes after congenital heart surgery.

METHODS

Data from patients undergoing congenital cardiac surgery from January 2011 to January 2020 at a quaternary referral center were retrospectively reviewed. Outcomes of interest included predischarge (early) mortality or transplant, postoperative hospital length-of-stay, inpatient cost of hospitalization, postdischarge (late) mortality or transplant, and late unplanned reintervention. The primary predictor was a US census tract-based, nationally-normed composite metric of contemporary child neighborhood opportunity comprising 29 indicators across 3 domains (education, health and environment, and socioeconomic), categorized as very low, low, moderate, high, and very high. Associations between childhood opportunity level and outcomes were evaluated using logistic regression (early mortality), generalized linear (length-of-stay and cost), Cox proportional hazards (late mortality), or competing risk (late reintervention) models, adjusting for baseline patient-related factors, case complexity, and residual lesion severity.

RESULTS

Of 6133 patients meeting entry criteria, the median age was 2.0 years (interquartile range, 3.6 months-8.3 years). There were 124 (2.0%) early deaths or transplants, the median postoperative length-of-stay was 7 days (interquartile range, 5-13 days), and the median inpatient cost was $76 000 (interquartile range, $50 000-130 000). No significant association between childhood opportunity level and early mortality or transplant was observed (P=0.21). On multivariable analysis, children with very low and low opportunity had significantly longer length-of-stay and incurred higher costs compared with those with very high opportunity (all P<0.05). Of 6009 transplant-free survivors of hospital discharge, there were 175 (2.9%) late deaths or transplants, and 1008 (16.8%) reinterventions at up to 10.5 years of follow-up. Patients with very low opportunity had a significantly greater adjusted risk of late death or transplant (hazard ratio, 1.7 [95% CI, 1.1-2.6]; P=0.030) and reintervention (subdistribution hazard ratio, 1.9 [95% CI, 1.5-2.3]; P<0.001), versus those with very high opportunity.

CONCLUSIONS

Childhood opportunity level is independently associated with adverse outcomes after congenital heart surgery. Children from resource-limited settings thus constitute an especially high-risk cohort that warrants closer surveillance and tailored interventions.

Is race-specific neighborhood social cohesion key to reducing racial disparities in late HIV diagnosis: A multiyear ecological study

We examined whether race/ethnic-specific social cohesion is associated with race/ethnic-specific HIV diagnosis rates using Bayesian space-time zero-inflated Poisson multivariable models, across 376 Census tracts. Social cohesion data were from the Southeastern Pennsylvania Household Health Survey, 2008-2015 and late HIV diagnosis data from eHARS system, 2009-2016. Areas where trust in neighbors reported by Black/African Americans was medium (compared to low) had lower rates of late HIV diagnosis among Black/African Americans (Relative Risk (RR)=0.52, 95% credible interval (CrI)= 0.34, 0.80). In contrast, areas where trust in neighbors reported by Black/African Americans were highest had lower late HIV diagnosis rates among Whites (RR=0.35, 95% CrI= 0.16, 0.76). Race/ethnic-specific differences in social cohesion may have implications for designing interventions aimed at modifying area-level social factors to reduce racial disparities in late HIV diagnosis.

Dynamic downscaling and daily nowcasting from influenza surveillance data

Real-time trends from surveillance data are important to assess and develop preparedness for influenza outbreaks. The overwhelming testing demand and limited capacity of testing laboratories for viral positivity render daily confirmed case data inaccurate and delay its availability in preparedness. Using Bayesian dynamic downscaling models, we obtained posterior estimates for daily influenza incidences from weekly estimates of the Centers for Disease Control and Prevention and daily reported constitutional and respiratory complaints during emergency department (ED) visits obtained from the state health departments. Our model provides one-day and seven-day lead forecasts along with 95 % $$ \% $$ prediction intervals. Our hybrid Markov Chain Monte Carlo and Kalman filter algorithms facilitate faster computation and enable us to update our estimates as new data become available. Our method is tested and validated using the State of Michigan data over the years 2009-2013. Reported constitutional and respiratory complaints at the EDs showed strong correlations of 0.81 and 0.68 respectively, with influenza rates. In general, our forecast model can be adapted to track an outbreak with only one respiratory virus as a causative agent.

PLACES: Local Data for Better Health

Local-level data on the health of populations are important to inform and drive effective and efficient actions to improve health, but such data are often expensive to collect and thus rare. Population Level Analysis and Community EStimates (PLACES) (www.cdc.gov/places/), a collaboration between the Centers for Disease Control and Prevention (CDC), the Robert Wood Johnson Foundation, and the CDC Foundation, provides model-based estimates for 29 measures among all counties and most incorporated and census-designated places, census tracts, and ZIP Code tabulation areas across the US. PLACES allows local health departments and others to better understand the burden and geographic distribution of chronic disease-related outcomes in their areas regardless of population size and urban-rural status and assists them in planning public health interventions. Online resources allow users to visually explore health estimates geographically, compare estimates, and download data for further use and exploration. By understanding the PLACES overall approach and using the easy-to-use PLACES applications, practitioners, policy makers, and others can enhance their efforts to improve public health, including informing prevention activities, programs, and policies; identifying priority health risk behaviors for action; prioritizing investments to areas with the biggest gaps or inequities; and establishing key health objectives to achieve community health and health equity.

US county-level estimation for maternal and infant health-related behavior indicators using pregnancy risk assessment monitoring system data, 2016-2018

Background

There is a critical need for maternal and child health data at the local level (for example, county), yet most counties lack sustainable resources or capabilities to collect local-level data. In such case, model-based small area estimation (SAE) could be a feasible approach. SAE for maternal or infant health-related behaviors at small areas has never been conducted or evaluated.

Methods

We applied multilevel regression with post-stratification approach to produce county-level estimates using Pregnancy Risk Assessment Monitoring System (PRAMS) data, 2016–2018 (n = 65,803 from 23 states) for 2 key outcomes, breastfeeding at 8 weeks and infant non-supine sleeping position.

Results

Among the 1,471 counties, the median model estimate of breastfeeding at 8 weeks was 59.8% (ranged from 34.9 to 87.4%), and the median of infant non-supine sleeping position was 16.6% (ranged from 10.3 to 39.0%). Strong correlations were found between model estimates and direct estimates for both indicators at the state level. Model estimates for both indicators were close to direct estimates in magnitude for Philadelphia County, Pennsylvania.

Conclusion

Our findings support this approach being potentially applied to other maternal and infant health and behavioral indicators in PRAMS to facilitate public health decision-making at the local level.

Race, mental health, and evictions filings in Memphis, TN, USA

•

Eviction filing rates are associated with mental distress in Black neighborhoods.

•

Eviction prevention should consider housing market dynamics and racial segregation.

•

Public policy interventions are needed to address the adverse effects of evictions.

Although evictions are a major disruptor of residential stability, their contribution to health disparities is understudied. Both experiencing eviction and the threat of being evicted are associated with adverse physical and mental health outcomes. Communities with higher proportions of Black people have higher rates of eviction filings. Market characteristics alone are insufficient for explaining the clustering of eviction in neighborhoods of color. Memphis is the fastest-growing rental market in the United States, facing an eviction crisis and is rife with persistent racial health disparities. This study explored the relationship between eviction filings, mental health, and neighborhood racial composition in Memphis to inform local policy approaches. We combined health from the City Health Dashboard, 2019 American Community Survey 5-year estimates, and eviction filings from the Shelby County, TN General Sessions Civil Court. Multivariate regression models were used to examine the relationship between health outcomes and eviction filing rates while controlling other relevant neighborhood characteristics. Separate models were run based on neighborhood racial composition. Poor mental health was significantly associated with higher eviction filling rates in majority Black neighborhoods but not in majority white and racially mixed neighborhoods. These findings point to evictions as an important contributor to racial health inequities in Memphis and the importance of race-conscious policy interventions that address the dual crisis of evictions and racial health disparities.

Invited Commentary: Estimation and Bounds Under Data Fusion

In their recent article, Ogburn et al. (Am J Epidemiol. 2021;190(6):1142-1147) raised a cautionary tale for epidemiologic data fusion: Bias may occur if a variable that is completely missing in the primary data set is imputed according to a regression model estimated from an auxiliary data set. However, in some specific settings, a solution may exist. Focusing on a linear outcome regression model with a missing covariate, we show that the bias can be eliminated if the underlying imputation model for the missing covariate is nonlinear in the common variables measured in both data sets. Otherwise, we describe 2 alternative approaches existing in the data fusion literature that could partially resolve this issue: One fits the outcome model by leveraging an additional validation data set containing joint observations of the outcome and the missing covariate, and the other offers informative bounds for the outcome regression coefficients without using validation data. We justify these 3 methods in a linear outcome model and briefly discuss their extension to general settings.

Association of Poor Mental-Health Days With COVID-19 Infection Rates in the U.S

INTRODUCTION

Limited evidence exists about the association between prior prevalence of poor mental health at the area level and subsequent rates of COVID-19 infections. This association was tested using area-level nationwide population data in the U.S.

METHODS

A nationwide study including 2,839 U.S. counties was conducted. Poor mental health was the age-adjusted average number of days within the past 30 days that adults reported poor mental health, including depression, stress, and problems with emotions, from the Behavioral Risk Factor Surveillance System. COVID-19 infection rates were cumulative confirmed cases between January 22 and October 7, 2020 per 100,000 people in the general population. Bayesian spatial mixed-effects regression estimated the relationship between COVID-19 infection and poor mental-health days at the county level in 2019 and change in poor mental health between 2010 and 2019, adjusted for several covariates.

RESULTS

Poor mental-health days in 2019 were positively associated with higher COVID-19 infection rates (RRR=1.059, 95% credible interval=1.003, 1.117). Change in mental health was not significantly associated with COVID-19.

CONCLUSIONS

Prior rates of poor mental health in a county were associated with a higher burden of COVID-19 infection. Interventions that improve well-being and strengthen mental-health systems at the community and other geographic levels are needed to address post-COVID-19 mental health problems.

Multilevel small area estimation for county-level prevalence of colorectal cancer screening test use in the United States using 2018 data

PURPOSE-

National screening estimates mask county-level variations. We aimed to generate county-level colorectal cancer (CRC) screening prevalence estimates for 2018 among adults aged 50-75 years and identify counties with low screening prevalence.

METHODS-

We combined individual-level county data from the 2018 Behavioral Risk Factor Surveillance System (BRFSS) (n = 204,947) with the 2018 American Community Survey county poverty data as a covariate, and the 2018 U.S. Census county population count data to generate county-level prevalence estimates for being current with any CRC screening test, colonoscopy, and home stool blood test. Because BRFSS is a state-based survey, and because some counties did not have samples for analysis, we used correlation coefficients to test internal consistency between model-based and BRFSS state estimates.

RESULTS-

Correlation coefficients tests were ≥0.97. Model-based national prevalence for any test was 69.9% (95% CI, 69.5% -70.4%) suggesting 30% are not current with screening test use. State mean estimates ranged from 62.1% in Alaska and Wyoming to 76.6% in Maine and Massachusetts. County mean estimates ranged from 42.2% in Alaska to 80.0% in Florida and Rhode Island. Most tests were performed with colonoscopy.

CONCLUSIONS-

Estimates across all U.S. counties showed large variations. Estimates may be informative for planning by states and local screening programs.

Estimates of Childhood Overweight and Obesity at the Region, State, and County Levels: A Multilevel Small-Area Estimation Approach

Local-level childhood overweight and obesity data are often used to implement and evaluate community programs, as well as allocate resources to combat overweight and obesity. The most current substate estimates of US childhood obesity use data collected in 2007. Using a spatial multilevel model and the 2016 National Survey of Children's Health, we estimated childhood overweight and obesity prevalence rates at the Census regional division, state, and county levels using small-area estimation with poststratification. A sample of 24,162 children aged 10-17 years was used to estimate a national overweight and obesity rate of 30.7% (95% confidence interval: 27.0%, 34.9%). There was substantial county-to-county variability (range, 7.0% to 80.9%), with 31 out of 3,143 counties having an overweight and obesity rate significantly different from the national rate. Estimates from counties located in the Pacific region had higher uncertainty than other regions, driven by a higher proportion of underrepresented sociodemographic groups. Child-level overweight and obesity was related to race/ethnicity, sex, parental highest education (P < 0.01 for all), county-level walkability (P = 0.03), and urban/rural designation (P = 0.02). Overweight and obesity remains a vital issue for US youth, with substantial area-level variability. The additional uncertainty for underrepresented groups shows surveys need to better target diverse samples.

Generating Subcounty Health Data Products: Methods and Recommendations From a Multistate Pilot Initiative

BACKGROUND

County Health Rankings & Roadmaps (CHR&R) makes data on health determinants and outcomes available at the county level, but health data at subcounty levels are needed. Three pilot projects in California, Missouri, and New York explored multiple approaches for defining measures and producing data at subcounty geographic and demographic levels based on the CHR&R model. This article summarizes the collective technical and implementation considerations from the projects, challenges inherent in analyzing subcounty health data, and lessons learned to inform future subcounty health data projects.

METHODS

The research teams used 12 data sources to produce 40 subcounty measures that replicate or approximate county-level measures from the CHR&R model. Using varying technical methods, the pilot projects followed similar stages: (1) conceptual development of data sources and measures; (2) analysis and presentation of small-area and subpopulation measures for public health, health care, and lay audiences; and (3) positioning the subcounty data initiatives for growth and sustainability. Unique technical considerations, such as degree of data suppression or data stability, arose during the project implementation. A compendium of technical resources, including samples of automated programs for analyzing and reporting subcounty data, was also developed.

RESULTS

The teams summarized the common themes shared by all projects as well as unique technical considerations arising during the project implementation. Furthermore, technical challenges and implementation challenges involved in subcounty data analyses are discussed. Lessons learned and proposed recommendations for prospective analysts of subcounty data are provided on the basis of project experiences, successes, and challenges.

CONCLUSIONS

This multistate pilot project offers 3 successful approaches for creating and disseminating subcounty data products to communities. Subcounty data often are more difficult to obtain than county-level data and require additional considerations such as estimate stability, validating accuracy, and protecting individual confidentiality. We encourage future projects to further refine techniques for addressing these critical considerations.

What is the place for space in epidemiology?

At the heart of spatial epidemiology is the need to describe and understand variation in population health. In this review and introduction to the themed issue on "Spatial Analysis and GIS in Epidemiology," we present theoretical foundations and methodological developments in spatial epidemiology, discuss spatial analytical techniques and their public health applications, and identify novel data sources and applications with the potential to make epidemiology more consequential. Challenges with using georeferenced data are also explored, including dealing with small sample sizes, missingness, generalizability, and geographic scale. Given the increasing availability of spatial data and visualization tools, we have an opportunity to overcome traditionally siloed fields and practice settings to advance knowledge and more appropriately respond to emerging public health crises.

Surveillance of Physical Activity and Sedentary Behavior Among Youth and Adults in the United States: History and Opportunities

BACKGROUND

Surveillance is a core function of public health, and approaches to national surveillance of physical activity and sedentary behavior have evolved over the past 2 decades. The purpose of this paper is to provide an overview of surveillance of physical activity and sedentary behavior in the United States over the past 2 decades, along with related challenges and emerging opportunities.

METHODS

The authors reviewed key national surveillance systems for the assessment of physical activity and sedentary behavior among youth and adults in the United States between 2000 and 2019.

RESULTS

Over the past 20 years, 8 surveillance systems have assessed physical activity, and 5 of those have assessed sedentary behavior. Three of the 8 originated in nonpublic health agencies. Most systems have assessed physical activity and sedentary behavior via surveys. However, survey questions varied over time within and also across systems, resulting in a wide array of available data.

CONCLUSION

The evolving nature of physical activity surveillance in the United States has resulted in both broad challenges (eg, balancing content with survey space; providing data at the national, state, and local level; adapting traditional physical activity measures and survey designs; and addressing variation across surveillance systems) and related opportunities.

Using Multilevel Regression and Poststratification to Estimate Physical Activity Levels from Health Surveys

BACKGROUND

Large-scale health surveys often consider sociodemographic characteristics and several health indicators influencing physical activity that often vary across subpopulations. Data in a survey for some small subpopulations are often not representative of the larger population.

OBJECTIVE

We developed a multilevel regression and poststratification (MRP) model to estimate leisure-time physical activity across Brazilian state capitals and evaluated whether the MRP outperforms single-level regression estimates based on the Brazilian cross-sectional national survey VIGITEL (2018).

METHODS

We used various approaches to compare the MRP and single-level model (complete-pooling) estimates, including cross-validation with various subsample proportions tested.

RESULTS

MRP consistently had predictions closer to the estimation target than single-level regression estimations. The mean absolute errors were smaller for the MRP estimates than single-level regression estimates with smaller sample sizes. MRP presented substantially smaller uncertainty estimates compared to single-level regression estimates. Overall, the MRP was superior to single-level regression estimates, particularly with smaller sample sizes, yielding smaller errors and more accurate estimates.

CONCLUSION

The MRP is a promising strategy to predict subpopulations' physical activity indicators from large surveys. The observations present in this study highlight the need for further research, which could, potentially, incorporate more information in the models to better interpret interactions and types of activities across target populations.

Investigating associations between COVID-19 mortality and population-level health and socioeconomic indicators in the United States: A modeling study

BACKGROUND

With the availability of multiple Coronavirus Disease 2019 (COVID-19) vaccines and the predicted shortages in supply for the near future, it is necessary to allocate vaccines in a manner that minimizes severe outcomes, particularly deaths. To date, vaccination strategies in the United States have focused on individual characteristics such as age and occupation. Here, we assess the utility of population-level health and socioeconomic indicators as additional criteria for geographical allocation of vaccines.

METHODS AND FINDINGS

County-level estimates of 14 indicators associated with COVID-19 mortality were extracted from public data sources. Effect estimates of the individual indicators were calculated with univariate models. Presence of spatial autocorrelation was established using Moran's I statistic. Spatial simultaneous autoregressive (SAR) models that account for spatial autocorrelation in response and predictors were used to assess (i) the proportion of variance in county-level COVID-19 mortality that can explained by identified health/socioeconomic indicators (R2); and (ii) effect estimates of each predictor. Adjusting for case rates, the selected indicators individually explain 24%-29% of the variability in mortality. Prevalence of chronic kidney disease and proportion of population residing in nursing homes have the highest R2. Mortality is estimated to increase by 43 per thousand residents (95% CI: 37-49; p < 0.001) with a 1% increase in the prevalence of chronic kidney disease and by 39 deaths per thousand (95% CI: 34-44; p < 0.001) with 1% increase in population living in nursing homes. SAR models using multiple health/socioeconomic indicators explain 43% of the variability in COVID-19 mortality in US counties, adjusting for case rates. R2 was found to be not sensitive to the choice of SAR model form. Study limitations include the use of mortality rates that are not age standardized, a spatial adjacency matrix that does not capture human flows among counties, and insufficient accounting for interaction among predictors.

CONCLUSIONS

Significant spatial autocorrelation exists in COVID-19 mortality in the US, and population health/socioeconomic indicators account for a considerable variability in county-level mortality. In the context of vaccine rollout in the US and globally, national and subnational estimates of burden of disease could inform optimal geographical allocation of vaccines.

A Warning About Using Predicted Values From Regression Models for Epidemiologic Inquiry

In many settings, researchers may not have direct access to data on 1 or more variables needed for an analysis and instead may use regression-based estimates of those variables. Using such estimates in place of original data, however, introduces complications and can result in uninterpretable analyses. In simulations and observational data, we illustrate the issues that arise when an average treatment effect is estimated from data where the outcome of interest is predicted from an auxiliary model. We show that bias in any direction can result, under both the null and alternative hypotheses.

Constructing Within-City Neighborhood Health Rankings in Philadelphia by Using Data From the 500 Cities Project

Introduction

Profound geographic disparities in health exist in many US cities. Most reporting on these disparities is based on predetermined administrative districts that may not reflect true neighborhoods. We undertook a ranking project to describe health at the neighborhood level and used Philadelphia, Pennsylvania, as our case study.

Methods

To create neighborhood health rankings, we first divided the city into neighborhoods according to groups of contiguous census tracts. Modeling our ranking methods and indicators on the Robert Wood Johnson Foundation County Health Rankings, we gathered census tract–level data from the Centers for Disease Control and Prevention’s 500 Cities Project and local sources and aggregated these data, as needed, to each neighborhood. We assigned composite scores and rankings for both health outcomes and health factors to each neighborhood.

Results

Scores for health outcomes and health factors were highly correlated. We found clusters of neighborhoods with low rankings in Philadelphia’s northern, lower northeastern, western, and southwestern regions. We disseminated information on rankings throughout the city, including through a comprehensive webpage, public communication, and a museum exhibit.

Conclusion

The Philadelphia neighborhood health rankings were designed to be accessible to people unfamiliar with public health, facilitating education on drivers of health in communities. Our methods can be used as a model for other cities to create and communicate data on within-city geographic health disparities.

Quantile Regression Forests to Identify Determinants of Neighborhood Stroke Prevalence in 500 Cities in the USA: Implications for Neighborhoods with High Prevalence

Stroke exerts a massive burden on the US health and economy. Place-based evidence is increasingly recognized as a critical part of stroke management, but identifying the key determinants of neighborhood stroke prevalence and the underlying effect mechanisms is a topic that has been treated sparingly in the literature. We aim to fill in the research gaps with a study focusing on urban health. We develop and apply analytical approaches to address two challenges. First, domain expertise on drivers of neighborhood-level stroke outcomes is limited. Second, commonly used linear regression methods may provide incomplete and biased conclusions. We created a new neighborhood health data set at census tract level by pooling information from multiple sources. We developed and applied a machine learning-based quantile regression method to uncover crucial neighborhood characteristics for neighborhood stroke outcomes among vulnerable neighborhoods burdened with high prevalence of stroke. Neighborhoods with a larger share of non-Hispanic blacks, older adults, or people with insufficient sleep tended to have a higher prevalence of stroke, whereas neighborhoods with a higher socio-economic status in terms of income and education had a lower prevalence of stroke. The effects of five major determinants varied geographically and were significantly stronger among neighborhoods with high prevalence of stroke. Highly flexible machine learning identifies true drivers of neighborhood cardiovascular health outcomes from wide-ranging information in an agnostic and reproducible way. The identified major determinants and the effect mechanisms can provide important avenues for prioritizing and allocating resources to develop optimal community-level interventions for stroke prevention.

Linking Data on Constituent Health with Elected Officials' Opinions: Associations Between Urban Health Disparities and Mayoral Officials' Beliefs About Health Disparities in Their Cities

Policy Points Mayoral officials' opinions about the existence and fairness of health disparities in their city are positively associated with the magnitude of income-based life expectancy disparity in their city. Associations between mayoral officials' opinions about health disparities in their city and the magnitude of life expectancy disparity in their city are not moderated by the social or fiscal ideology of mayoral officials or the ideology of their constituents. Highly visible and publicized information about mortality disparities, such as that related to COVID-19 disparities, has potential to elevate elected officials' perceptions of the severity of health disparities and influence their opinions about the issue.

CONTEXT

A substantive body of research has explored what factors influence elected officials' opinions about health issues. However, no studies have assessed the potential influence of the health of an elected official's constituents. We assessed whether the magnitude of income-based life expectancy disparity within a city was associated with the opinions of that city's mayoral official (i.e., mayor or deputy mayor) about health disparities in their city.

METHODS

The independent variable was the magnitude of income-based life expectancy disparity in US cities. The magnitude was determined by linking 2010-2015 estimates of life expectancy and median household income for 8,434 census tracts in 224 cities. The dependent variables were mayoral officials' opinions from a 2016 survey about the existence and fairness of health disparities in their city (n = 224, response rate 30.3%). Multivariable logistic regression was used to adjust for characteristics of mayoral officials (e.g., ideology) and city characteristics.

FINDINGS

In cities in the highest income-based life expectancy disparity quartile, 50.0% of mayoral officials "strongly agreed" that health disparities existed and 52.7% believed health disparities were "very unfair." In comparison, among mayoral officials in cities in the lowest disparity quartile 33.9% "strongly agreed" that health disparities existed and 22.2% believed the disparities were "very unfair." A 1-year-larger income-based life expectancy disparity in a city was associated with 25% higher odds that the city's mayoral official would "strongly agree" that health disparities existed (odds ratio [OR] = 1.25; P = .04) and twice the odds that the city's mayoral official would believe that such disparities were "very unfair" (OR = 2.24; P <.001).

CONCLUSIONS

Mayoral officials' opinions about health disparities in their jurisdictions are generally aligned with, and potentially influenced by, information about the magnitude of income-based life expectancy disparities among their constituents.