Estimates of county-level disease prevalence have a variety of applications. Such estimation is often done via model-based small-area estimation using survey data. However, for conditions with low prevalence (i.e., rare diseases or newly diagnosed diseases), counties with a high fraction of zero counts in surveys are common. They are often more common than the model used would lead one to expect; such zeros are called ‘excess zeros’. The excess zeros can be structural (there are no cases to find) or sampling (there are cases, but none were selected for sampling). These issues are often addressed by combining multiple years of data. However, this approach can obscure trends in annual estimates and prevent estimates from being timely. Using single-year survey data, we proposed a Bayesian weighted Binomial Zero-inflated (BBZ) model to estimate county-level rare diseases prevalence. The BBZ model accounts for excess zero counts, the sampling weights and uses a power prior. We evaluated BBZ with American Community Survey results and simulated data. We showed that BBZ yielded less bias and smaller variance than estimates based on the binomial distribution, a common approach to this problem. Since BBZ uses only a single year of survey data, BBZ produces more timely county-level incidence estimates. These timely estimates help pinpoint the special areas of county-level needs and help medical researchers and public health practitioners promptly evaluate rare diseases trends and associations with other health conditions.

INTRODUCTION:Demonstrating the validity of a public health simulation model helps to establish confidence in the accuracy and usefulness of a model's results. In this study we evaluated the validity of the Prevention Impacts Simulation Model (PRISM), a system dynamics model that simulates health, mortality, and economic outcomes for the US population. PRISM primarily simulates outcomes related to cardiovascular disease but also includes outcomes related to other chronic diseases that share risk factors. PRISM is openly available through a web application.METHODS:We applied the model validation framework developed independently by the International Society of Pharmacoeconomics and Outcomes Research and the Society for Medical Decision Making modeling task force to validate PRISM. This framework included model review by external experts and quantitative data comparison by the study team.RESULTS:External expert review determined that PRISM is based on up-to-date science. One-way sensitivity analysis showed that no parameter affected results by more than 5%. Comparison with other published models, such as ModelHealth, showed that PRISM produces lower estimates of effects and cost savings. Comparison with surveillance data showed that projected model trends in risk factors and outcomes align closely with secular trends. Four measures did not align with surveillance data, and those were recalibrated.CONCLUSION:PRISM is a useful tool to simulate the potential effects and costs of public health interventions. Results of this validation should help assure health policy leaders that PRISM can help support community health program planning and evaluation efforts.

The National Immunization Survey (NIS) is the United States’ primary tool for assessing immunization coverage among 19- to 35-monthold children. Although annual estimates from the NIS are quite precise at the national level, US State-level estimates have much larger sampling error than national-level estimates. We combined two independent unbiased estimates of US State-level coverages within a given year to obtain new estimates which are more precise than previously published estimates. We first calculated a model-based estimate for each State for 2001 using multiple years of NIS data. Next, we combined each model-based estimate with the corresponding, previously reported NIS estimate for 2001. Our resulting estimates of State-level immunization coverage had smaller standard errors than the previously published estimates. To make similar improvements in precision by increasing sample size would, depending on State, require an increase in sample size of 30% – 120%.

Objective: After the discontinuation of School Health Policies and Practices Study (SHPPS) in 2016, the US Centers for Disease Control and Prevention (CDC) began exploring innovative ways to gather school health information using existing surveillance systems. School Health Profiles (Profiles) is a school-based system of surveys that monitors school health policies and practices in states and other jurisdictions. The objective of this study was to assess whether prevalence estimates calculated using nationally representative SHPPS as an established benchmark were similar to estimates using aggregated Profiles data. Method: Nationwide 2014 Profiles estimates were calculated from data across all 50 US states and the District of Columbia and compared to national 2014 SHPPS estimates. Fifty-seven questions were identical between the data sources. Equivalence tests were used to determine similarity between data sources. Results: Overall, the median difference between 2014 SHPPS and 2014 Profiles estimate was one percentage point and distribution-free 95% confidence intervals were (−0.8, 3.1). Of the 57 school health policy and practice indicators examined in this study, 38 (66.7%) were found to be equivalent. Of these equivalent indicators, the median percentage point difference between data sources was 0.8 (−0.8, 2.5). A nonparametric sign test showed that none of the medians of the estimate differences examined were significantly different from zero. Conclusion: This study demonstrates the expanded utility of state-level data to meet public health surveillance needs. This study found that aggregated, state-level Profiles data can be used to calculate nationwide prevalence estimates that are reasonably consistent with results from a nationally representative survey.

Bayesian hierarchical regression (BHR) is often used in small area estimation (SAE). BHR conditions on the samples. Therefore, when data are from a complex sample survey, neither survey sampling design nor survey weights are used. This can introduce bias and/or cause large variance. Further, if non-informative priors are used, BHR often requires the combination of multiple years of data to produce sample sizes that yield adequate precision; this can result in poor timeliness and can obscure trends. To address bias and variance, we propose a design assisted model-based approach for SAE by integrating adjusted sample weights. To address timeliness, we use historical data to define informative priors (power prior); this allows estimates to be derived from a single year of data. Using American Community Survey data for validation, we applied the proposed method to Behavioral Risk Factor Surveillance System data. We estimated the prevalence of disability for all U.S. counties. We show that our method can produce estimates that are both more timely than those arising from widely-used alternatives and are closer to ACS' direct estimates, particularly for low-data counties. Our method can be generalized to estimate the county-level prevalence of other health related measurements.

People with type 2 diabetes mellitus experience an increased prevalence of non-alcoholic fatty liver disease (NAFLD) compared with the general population and often with worse outcomes. As part of the ABCD Liraglutide Nationwide Audit Programme, we obtained and analysed data from 2009 to 2018 to assess the impact of liraglutide on alanine aminotransferase (ALT) levels as a marker of liver inflammation (often used in clinical trials as a marker of NAFLD). After excluding those with insufficient or incomplete data, we analysed the results from 1,759 patients treated in the real-world clinical setting. Our results demonstrated an overall significant decrease in median ALT (−1 U/L, 95% CI −1 to −2, p<0.001) compared with baseline, which was more pronounced in patients with elevated ALT based on gender- specific ranges (male: −4 U/L, 95% CI −3 to −6, p<0.001; female: −3 U/L, 95% CI −2 to −4, p<0.001). There was no correlation between weight loss and degree of ALT change (rho=−0.0002, p=0.41). Our data mirror outcomes from large randomised controlled trials and show that the impact of liraglutide on ALT is likely generalisable to real-world practice. Some of our data suggest that there may be a slight increase in ALT in those with normal levels at baseline, although the clinical significance of this is uncertain.

Introduction Burden of disease is often defined by using epidemiologic measures. However, there may be latent aspects of disease burden that are not factored into these types of estimates. This study quantified environmental burden of disease by using population health indicators and exploratory factor analysis at the county level across the United States. Methods Ninety-nine variables drawn from public use data sets from 2010 to 2016 were used to create a multifactor index - the burden index. We applied principal components analysis with promax rotation to allow the factors to correlate. Correlation coefficients for each factor and the outcome of interest, age-adjusted cancer death rate, were calculated. We used both unadjusted and adjusted linear regression techniques. Results The final additive county-level index included 9 factors that explained 68.3% of the variance in the counties and county equivalents. The burden index had a moderate association with the ageadjusted cancer death rates (r = .48 , P <.001), and adjusted linear regression with all 9 factors explained 34% of the variance in the age-adjusted cancer death rate. Results were mapped, and the geographic distribution of both the burden index and age-adjusted cancer mortality were assessed. There are distinct geospatial patterns for both. Conclusions Results from this study show potential areas of need, as well as the importance of including environmental variables in the study of cancer etiology. Future studies can aim to validate these findings by quantifying burden as it relates to overall cancer mortality by using epidemiologic measures, along with other confirmatory statistical methods.

BACKGROUND:Health-related quality of life (HRQOL) is a multi-dimensional concept commonly used to examine the impact of health status on quality of life. HRQOL is often measured by four core questions that asked about general health status and number of unhealthy days in the Behavioral Risk Factor Surveillance System (BRFSS). Use of these measures individually, however, may not provide a cohesive picture of overall HRQOL. To address this concern, this study developed and tested a method for combining these four measures into a summary score.METHODS:Exploratory and confirmatory factor analyses were performed using BRFSS 2013 data to determine potential numerical relationships among the four HRQOL items. We also examined the stability of our proposed one-factor model over time by using BRFSS 2001-2010 and BRFSS 2011-2013 data sets.RESULTS:Both exploratory factor analysis and goodness of fit tests supported the notion that one summary factor could capture overall HRQOL. Confirmatory factor analysis indicated acceptable goodness of fit of this model. The predicted factor score showed good validity with all of the four HRQOL items. In addition, use of the one-factor model showed stability, with no changes being detected from 2001 to 2013.CONCLUSION:Instead of using four individual items to measure HRQOL, it is feasible to study overall HRQOL via factor analysis with one underlying construct. The resulting summary score of HRQOL may be used for health evaluation, subgroup comparison, trend monitoring, and risk factor identification.

Introduction Racial/ethnic disparities have been studied extensively. However, the combined influence of geographic location and economic status on specific health outcomes is less well studied. This study’s objective was to examine 1) the disparity in chronic disease prevalence in the United States by county economic status and metropolitan classification and 2) the social gradient by economic status. The association of hypertension, arthritis, and poor health with county economic status was also explored. Methods We used 2013 Behavioral Risk Factor Surveillance System data. County economic status was categorized by using data on unemployment, poverty, and per capita market income. While controlling for sociodemographics and other covariates, we used multivariable logistic regression to evaluate the relationship between economic status and hypertension, arthritis, and self-rated health. Results Prevalence of hypertension, arthritis, and poor health in the poorest counties was 9%, 13%, and 15% higher, respectively, than in the most affluent counties. After we controlled for covariates, poor counties still had a higher prevalence of the studied conditions. Conclusion We found that residents of poor counties had a higher prevalence of poor health outcomes than affluent counties, even after we controlled for known risk factors. Further, the prevalence of poor health outcomes decreased as county economics improved. Findings suggest that poor counties would benefit from targeted public health interventions, better access to health care services, and improved food and built environments.

BACKGROUND:In recent decades, the United States experienced increasing prevalence and incidence of diabetes, accompanied by large disparities in county-level diabetes prevalence and incidence. However, whether these disparities are widening, narrowing, or staying the same has not been studied. We examined changes in disparity among U.S. counties in diagnosed diabetes prevalence and incidence between 2004 and 2012.METHODS:We used 2004 and 2012 county-level diabetes (type 1 and type 2) prevalence and incidence data, along with demographic, socio-economic, and risk factor data from various sources. To determine whether disparities widened or narrowed over the time period, we used a regression-based β-convergence approach, accounting for spatial autocorrelation. We calculated diabetes prevalence/incidence percentage point (ppt) changes between 2004 and 2012 and modeled these changes as a function of baseline diabetes prevalence/incidence in 2004. Covariates included county-level demographic and, socio-economic data, and known type 2 diabetes risk factors (obesity and leisure-time physical inactivity).RESULTS:For each county-level ppt increase in diabetes prevalence in 2004 there was an annual average increase of 0.02 ppt (p<0.001) in diabetes prevalence between 2004 and 2012, indicating a widening of disparities. However, after accounting for covariates, diabetes prevalence decreased by an annual average of 0.04 ppt (p<0.001). In contrast, changes in diabetes incidence decreased by an average of 0.04 ppt (unadjusted) and 0.09 ppt (adjusted) for each ppt increase in diabetes incidence in 2004, indicating a narrowing of county-level disparities.CONCLUSIONS:County-level disparities in diagnosed diabetes prevalence in the United States widened between 2004 and 2012, while disparities in incidence narrowed. Accounting for demographic and, socio-economic characteristics and risk factors for type 2 diabetes narrowed the disparities, suggesting that these factors are strongly associated with changes in disparities. Public health interventions that target modifiable risk factors, such as obesity and physical inactivity, in high burden counties might further reduce disparities in incidence and, over time, in prevalence.

The residuals of a least squares regression model are defined as the observations minus the modeled values. For least squares regression to produce valid CIs and P values, the residuals must be independent, be normally distributed, and have a constant variance. If these assumptions are not satisfied, estimates can be biased and power can be reduced. However, there are ways to assess these assumptions and steps one can take if the assumptions are violated. Here, we discuss both assessment and appropriate responses to violation of assumptions.

OBJECTIVE High out-of-pocket (OOP) costs can be an obstacle to health care access and treatment compliance. This study investigated trends in high OOP health care burden in people with diabetes. RESEARCH DESIGN AND METHODS Using Medical Expenditure Panel Survey 2001–2011 data, we examined trends in the proportion of people aged 18–64 years with diabetes facing a high OOP burden. We also examined whether the trend differed by insurance status (private insurance, public insurance, or no insurance) or by income level (poor and near poor, low income, middle income, or high income). RESULTS In 2011, 23% of people with diabetes faced high OOP burden. Between 2001–2002 and 2011, the proportion of people facing high OOP burden fell by 5 percentage points (P < 0.01). The proportion of those who were publicly insured decreased by 22 percentage points (P < 0.001) and of those who were uninsured by 12 percentage points (P = 0.01). Among people with diabetes who were poor and near poor and those with low income, the proportion facing high OOP burden decreased by 21 (P < 0.001) and 13 (P = 0.01) percentage points, respectively; no significant change occurred in the proportion with private insurance or middle and high incomes between 2001–2002 and 2011. CONCLUSIONS The past decade has seen a narrowing of insurance coverage and income-related disparities in high OOP burden in people with diabetes; yet, almost one-fourth of all people with diabetes still face a high OOP burden.

OBJECTIVE To assess the cost implications of diabetes prevention, it is important to know the lifetime medical cost of people with diabetes relative to those without. We derived such estimates using data representative of the U.S. national population.RESEARCH DESIGN AND METHODSWe aggregated annual medical expenditures from the age of diabetes diagnosis to death to determine lifetime medical expenditure. Annual medical expenditures were estimated by sex, age at diagnosis, and diabetes duration using data from 2006-2009 Medical Expenditure Panel Surveys, which were linked to data from 2005-2008 National Health Interview Surveys. We combined survival data from published studies with the estimated annual expenditures to calculate lifetime spending. We then compared lifetime spending for people with diabetes with that for those without diabetes. Future spending was discounted at 3% annually.RESULTSThe discounted excess lifetime medical spending for people with diabetes was $124,600 ($211,400 if not discounted), $91,200 ($135,600), $53,800 ($70,200), and $35,900 ($43,900) when diagnosed with diabetes at ages 40, 50, 60, and 65 years, respectively. Younger age at diagnosis and female sex were associated with higher levels of lifetime excess medical spending attributed to diabetes.CONCLUSIONSHaving diabetes is associated with substantially higher lifetime medical expenditures despite being associated with reduced life expectancy. If prevention costs can be kept sufficiently low, diabetes prevention may lead to a reduction in long-term medical costs.

OBJECTIVE We examine barriers to receiving recommended eye care among people aged ≥40 years with diagnosed diabetes. RESEARCH DESIGN AND METHODS We analyzed 2006–2010 Behavioral Risk Factor Surveillance System data from 22 states (n = 27,699). Respondents who had not sought eye care in the preceding 12 months were asked the main reason why. We categorized the reasons as cost/lack of insurance, no need, no eye doctor/travel/appointment, and other (meaning everything else). We used multinomial logistic regression to control for race/ethnicity, education, income, and other selected covariates. RESULTS Among adults with diagnosed diabetes, nonadherence to the recommended annual eye examinations was 23.5%. The most commonly reported reasons for not receiving eye care in the preceding 12 months were “no need” and “cost or lack of insurance” (39.7 and 32.3%, respectively). Other reasons were “no eye doctor,” “no transportation” or “could not get appointment” (6.4%), and “other” (21.5%). After controlling for covariates, adults aged 40–64 years were more likely than those aged ≥65 years (relative risk ratio [RRR] = 2.79; 95% CI 2.01–3.89) and women were more likely than men (RRR = 2.33; 95% CI 1.75–3.14) to report “cost or lack of insurance” as their main reason. However, people aged 40–64 years were less likely than those aged ≥65 years to report “no need” (RRR = 0.51; 95% CI 0.39–0.67) as their main reason. CONCLUSIONS Addressing concerns about “cost or lack of insurance” for adults under 65 years and “no perceived need” among those 65 years and older could help improve eye care service utilization among people with diabetes.

IMPORTANCE:Although the prevalence and incidence of diabetes have increased in the United States in recent decades, no studies have systematically examined long-term, national trends in the prevalence and incidence of diagnosed diabetes.OBJECTIVE:To examine long-term trends in the prevalence and incidence of diagnosed diabetes to determine whether there have been periods of acceleration or deceleration in rates.DESIGN, SETTING, AND PARTICIPANTS:We analyzed 1980-2012 data for 664,969 adults aged 20 to 79 years from the National Health Interview Survey (NHIS) to estimate incidence and prevalence rates for the overall civilian, noninstitutionalized, US population and by demographic subgroups (age group, sex, race/ethnicity, and educational level).MAIN OUTCOMES AND MEASURES:The annual percentage change (APC) in rates of the prevalence and incidence of diagnosed diabetes (type 1 and type 2 combined).RESULTS:The APC for age-adjusted prevalence and incidence of diagnosed diabetes did not change significantly during the 1980s (for prevalence, 0.2% [95% CI, -0.9% to 1.4%], P = .69; for incidence, -0.1% [95% CI, -2.5% to 2.4%], P = .93), but each increased sharply during 1990-2008 (for prevalence, 4.5% [95% CI, 4.1% to 4.9%], P < .001; for incidence, 4.7% [95% CI, 3.8% to 5.6%], P < .001) before leveling off with no significant change during 2008-2012 (for prevalence, 0.6% [95% CI, -1.9% to 3.0%], P = .64; for incidence, -5.4% [95% CI, -11.3% to 0.9%], P = .09). The prevalence per 100 persons was 3.5 (95% CI, 3.2 to 3.9) in 1990, 7.9 (95% CI, 7.4 to 8.3) in 2008, and 8.3 (95% CI, 7.9 to 8.7) in 2012. The incidence per 1000 persons was 3.2 (95% CI, 2.2 to 4.1) in 1990, 8.8 (95% CI, 7.4 to 10.3) in 2008, and 7.1 (95% CI, 6.1 to 8.2) in 2012. Trends in many demographic subpopulations were similar to these overall trends. However, incidence rates among non-Hispanic black and Hispanic adults continued to increase (for interaction, P = .03 for non-Hispanic black adults and P = .01 for Hispanic adults) at rates significantly greater than for non-Hispanic white adults. In addition, the rate of increase in prevalence was higher for adults who had a high school education or less compared with those who had more than a high school education (for interaction, P = .006 for <high school and P < .001 for high school).CONCLUSIONS AND RELEVANCE:Analyses of nationally representative data from 1980 to 2012 suggest a doubling of the incidence and prevalence of diabetes during 1990-2008, and a plateauing between 2008 and 2012. However, there appear to be continued increases in the prevalence or incidence of diabetes among subgroups, including non-Hispanic black and Hispanic subpopulations and those with a high school education or less.

IMPORTANCE:This study provides further evidence from a national sample to generalize the relationship between depression and vision loss to adults across the age spectrum. Better recognition of depression among people reporting reduced ability to perform routine activities of daily living due to vision loss is warranted.OBJECTIVES:To estimate, in a national survey of US adults 20 years of age or older, the prevalence of depression among adults reporting visual function loss and among those with visual acuity impairment. The relationship between depression and vision loss has not been reported in a nationally representative sample of US adults. Previous studies have been limited to specific cohorts and predominantly focused on the older population.DESIGN:The National Health and Nutrition Examination Survey (NHANES) 2005-2008.SETTING:A cross-sectional, nationally representative sample of adults, with prevalence estimates weighted to represent the civilian, noninstitutionalized US population.PARTICIPANTS:A total of 10 480 US adults 20 years of age or older.MAIN OUTCOME MEASURES:Depression, as measured by the 9-item Patient Health Questionnaire depression scale, and vision loss, as measured by visual function using a questionnaire and by visual acuity at examination.RESULTS:In 2005-2008, the estimated crude prevalence of depression (9-item Patient Health Questionnaire score of ≥10) was 11.3% (95% CI, 9.7%-13.2%) among adults with self-reported visual function loss and 4.8% (95% CI, 4.0%-5.7%) among adults without. The estimated prevalence of depression was 10.7% (95% CI, 8.0%-14.3%) among adults with presenting visual acuity impairment (visual acuity worse than 20/40 in the better-seeing eye) compared with 6.8% (95% CI, 5.8%-7.8%) among adults with normal visual acuity. After controlling for age, sex, race/ethnicity, marital status, living alone or not, education, income, employment status, health insurance, body mass index, smoking, binge drinking, general health status, eyesight worry, and major chronic conditions, self-reported visual function loss remained significantly associated with depression (overall odds ratio, 1.9 [95% CI, 1.6-2.3]), whereas the association between presenting visual acuity impairment and depression was no longer statistically significant.CONCLUSIONS AND RELEVANCE:Self-reported visual function loss, rather than loss of visual acuity, is significantly associated with depression. Health professionals should be aware of the risk of depression among persons reporting visual function loss.

Although diabetes is one of the most costly and rapidly increasing serious chronic diseases worldwide, the optimal mix of strategies to reduce diabetes prevalence has not been determined.