A novel framework for validating and applying standardized small area measurement strategies
Published in Population Health Metrics, September 2010
Researchers at IHME have created a new approach for generating estimates of health trends in counties and other small population areas. They used this new small area estimation methodology to estimate the prevalence of diagnosed diabetes in all counties in the United States for 2008. The study, A novel framework for validating and applying standardized small area measurement strategies, was co-authored by Dr. Ali Mokdad, Professor of Global Health at IHME and the former head of the Behavioral Risk Factor Surveillance System (BRFSS) at the US Centers for Disease Control and Prevention.
Previously, data on prevalence of risk factors or diseases have mostly been limited to the state level. This innovative method allows local health officials, government agencies, and researchers to analyze existing data from BRFSS, censuses, and surveys to create estimates at the county level and among other population subsets. Here are some highlights of the paper:
- The state-level data often mask challenges or successes at the county level. This method reveals those anomalies. Colorado, for example, has the lowest rate of diabetes in the nation, with 7.1% of the population diagnosed with the disease. Within the state, though, Pitkin County has one of the lowest rates in the US at 4.5%, while Crowley County has a rate of 14.1%, ranking it near the bottom of all 3,140 counties.
- A large number of counties in the South, in Alaska, in the Dakotas, and along the Mexican border have the highest prevalence of diagnosed diabetes. In contrast, many counties along the Rocky Mountains, including counties in Montana and Wyoming, have the lowest rates of diagnosed diabetes.
- The new method allows local policymakers to compare different health needs and set priorities to use scarce resources efficiently.
- Counties will be able to monitor trends over time to evaluate the success of specific programs for risk factor prevention and control. They also will be able to apply and qualify for state and federal funds that had been inaccessible because of a lack of documentation of county-level trends.
Researchers used data on population, race, the number of fast food restaurants, household income, and the number of medical doctors to devise a new model for estimating risk factors and disease prevalence. They also used those data to create county-level estimates. They validated the model by testing it against different population scenarios. For example, they took counties with large sample sizes, such as King County in Washington, and sampled smaller subsets of that population to mirror counties with smaller sample sizes. Ultimately, they created a model that allowed them to make estimates based on data from 10 people that were of the same quality as estimates based on data from 3,000 people.
Part of IHME's core mission is to create the necessary tools for measuring health trends and tracking health system performance. By developing this new analytical method for making small area estimates, IHME intends to help local health officials and policymakers conduct their own analyses and create policies with the greatest possible impact on improving health.
Recommendations for future work
The researchers recommend these areas for future study:
- Researchers now can use our methodology to rank counties by risk factors and create a score that combines different risk factors such as smoking, obesity, and alcohol use. It also could include interventions such as seat belt use, Pap smear screenings, and influenza immunizations.
- County health officials can use the estimates available for download from IHME's website to create benchmarks to evaluate progress in their counties and to better allocate resources for prevention and control programs.
- In the future, IHME intends to use this approach, combined with other sources of data such as access to medical care, number of parks, and smoking bans, to create a health performance score for every county.
- IHME also intends to adjust the findings about diabetes by taking into account surveys that measure blood glucose levels to provide a total diabetes number. IHME is currently leading a study in Washington state to assess blood glucose levels, among other risk factors.
Citation: Srebotnjak T, Mokdad AH, Murray CJL. A novel framework for validating and applying standardized small area measurement strategies. Population Health Metrics. 2010; 8:26.
Data and Methods
Data for download. Diabetes prevalence rates by age, sex, and county, 2008 (21KB xls)
For additional information, visit our Global Health Data Exchange (GHDx). The GHDx includes data records with information on more than 200 countries.