Self-reported health status is a general measure of health-related quality of life (HRQoL) in a population. Measuring HRQoL helps characterize the burden of disabilities and chronic diseases in a population. Self-reported health status is a widely used measure of people’s health-related quality of life. In addition to measuring how long people live, it is important to also include measures that consider how healthy people are while alive.
The use of self-rated health as a measure to compare health status benefits from its comprehensive, inclusive, and non-specific nature. Furthermore, a meta-analysis of the association between mortality and a single item assessing self-rated health found that people with “poor” self-rated health had a twofold higher mortality risk than persons with “excellent” self-rated health. This analysis concludes that a single measure that takes little time to collect and can be captured routinely is appropriate for measuring health among large populations. A study that investigated the reliability of the HRQoL questions included in BRFSS found high retest reliability for the self-reported health measure.
Poor or Fair Health is a Percentage
Poor or Fair Health measures the percentage of adults in a county who consider themselves to be in poor or fair health.
Poor or Fair Health Estimates are Age-Adjusted
Age is a non-modifiable risk factor, and as age increases, poor health outcomes are more likely. We report an age-adjusted rate in order to fairly compare counties with differing age structures.
The Method for Calculating Poor or Fair Health has Changed
Prior to the 2016 County Health Rankings, the CDC’s BRFSS provided the Rankings with county-level estimates that were constructed from seven years of responses from participants who used a landline phone. Beginning with the 2016 Rankings, the CDC provided single-year modeled county-level estimates that included both landline and cell phone users. These changes were implemented in order to provide users with the most accurate estimates of health in their community as possible.
Poor or Fair Health Estimates are Created Using Statistical Modeling
Statistical modeling is used to obtain more informed and reliable estimates than survey data alone can provide. Modeling generates more stable estimates for places with small numbers of residents or survey responses. Our Poor or Fair Health estimates are produced from one year of survey data and are created using complex statistical modeling. For more technical information on BRFSS modeling please see their methodology.
There are also drawbacks to using modeled data. The smaller the population or sample size of a county, the more the estimates are derived from the model itself and the less they are based on survey responses. Models make assumptions about statistical relationships that may not hold in all cases. Finally, there is no perfect model and each model generally has limitations specific to their methods.
Caution Should be Used When Comparing These Estimates Across State Lines
The model used to create these estimates includes a state-level factor that limits comparability between neighboring counties of adjacent states.
The numerator is the total number of respondents who answered "In general, would you say that in general your health is Excellent/Very good/Good/Fair/Poor?" with fair or poor.
The denominator is the total number of adult respondents in a county.
This measure could be used to measure progress, but only after considering its substantial limitations. Methodological changes in the Behavioral Risk Factor Surveillance System, which are discussed above and were implemented in the 2016 Rankings, make comparisons with estimates prior to that release year difficult. In addition, current estimates are produced using sophisticated modeling techniques which make them difficult to use for tracking progress in small geographic areas.
Modeled estimates have specific drawbacks with regard to their usefulness in tracking progress in communities. Modeled data are not particularly good at incorporating the effects of local conditions, such as health promotion policies or unique population characteristics, into their estimates. Counties trying to measure the effects of programs and policies on the data should use great caution when using modeled estimates. In order to better understand and validate modeled estimates, confirming this data with additional sources of data at the local level is particularly valuable.
Years of Data Used
Behavioral Risk Factor Surveillance System
The Behavioral Risk Factor Surveillance System (BRFSS) is a state-based random digit dial (RDD) telephone survey that is conducted annually in all states, the District of Columbia, and U.S. territories. Data obtained from the BRFSS are representative of each state’s total non-institutionalized population over 18 years of age and have included more than 400,000 annual respondents with landline telephones or cellphones since 2011. Data are weighted using iterative proportional fitting (also called "raking") methods to reflect population distributions. For the County Health Rankings, data from the BRFSS are used to measure various health behaviors and health-related quality of life (HRQoL) indicators. HRQoL measures are age-adjusted to the 2000 U.S. standard population.
Prior to the 2016 County Health Rankings, up to seven survey years of landline only BRFSS data were aggregated to produce county estimates. However, even with multiple years of data, these did not provide reliable estimates for all counties, particularly those with smaller respondent samples. For the 2016 County Health Rankings, the CDC produced 2014 county estimates using single-year 2014 BRFSS data and a multilevel modeling approach based on respondent answers and their age, sex, and race/ethnicity, combined with county-level poverty, as well as county- and state-level contextual effects. To produce estimates for those counties where there were no or limited data, the modeling approach borrowed information from the entire BRFSS sample as well as Census Vintage 2014 population estimates. CDC used a parametric bootstrapping method to produce standard errors and confidence intervals for those point estimates. This estimation methodology was validated for all U.S. counties, including those with no or small (<50 respondents) samples. This same method was used in constructing the 500 cities study, which includes BRFSS data for the 500 largest cities in the U.S. and can be found here: https://www.cdc.gov/500cities.
One limitation of the BRFSS is that all measures are based on self-reported information, which cannot be validated with medical records. Another limitation is that these model-based estimates were created by borrowing information from the entire BRFSS, which may or may not accurately reflect those counties’ local intervention experiences. Additionally, the confidence intervals constructed from these methods appear much smaller than confidence intervals reported for direct survey methods in previous years.
- Zhang X, Holt JB, Lu H, Wheaton AG, Ford ES, Greenlund K, Croft JB. Multilevel regression and poststratification for small-area estimation of population health outcomes: a case study of chronic obstructive pulmonary disease prevalence using the Behavioral Risk Factor Surveillance System. American Journal of Epidemiology 2014;179(8):1025–1033.
- Zhang X, Holt JB, Yun, S, Lu H, Greenlund K, Croft JB. Validation of multilevel regression and poststratification methodology for small area estimation of health outcomes. American Journal of Epidemiology 2015;182(2):127-137.
|Subcounty Area|| |
The 500 Cities project provides city- and census tract-level small area estimates for chronic disease risk factors, health outcomes, and clinical preventive service use, including Poor or Fair Health, for the largest 500 cities in the United States.
For larger counties, you can access county- or MSA-specific data from the CDC at http://www.cdc.gov/brfss/smart/smart_data.htm. However, using this data requires somewhat advanced analytic capabilities. In many states, you can access county level BRFSS estimates, and in some cases you can stratify those estimates by age, gender, education, income, or race.
 Jylhä M. What is self-rated health and why does it predict mortality? Towards a unified conceptual model. Soc Sci Med. 2009;69:307-316.
 DeSalvo K, Bloser N, Reynolds K, He J, Muntner P. Mortality prediction with a single general self-rated health question. J Gen Intern Med. 2006;21:267-275.
 Andresen EM, Catlin TK, Wyrwich KW, Jackson-Thompson J. Retest reliability of surveillance questions on health related quality of life. J Epidemiol Community Health. 2003;57:339-343.
 Zhang X, Holt JB, Lu H, Wheaton AG, Ford ES, Greenlund K, Croft JB. Multilevel regression and poststratification for small-area estimation of population health outcomes: a case study of chronic obstructive pulmonary disease prevalence using the Behavioral Risk Factor Surveillance System. American Journal of Epidemiology 2014;179(8):1025–1033.