Behavioral Risk Factor Survey Module
Technical Notes

Wisconsin Behavioral Risk Factor Survey (BRFS)

The Wisconsin BRFS Program is located in the Office of Health Informatics, Division of Public Health, Wisconsin Department of Health Services. BRFS is managed by the Wisconsin BRFS coordinator/project director in collaboration with the Centers for Disease Control and Prevention (CDC).

Sample Design

BRFS conducts both landline and cell phone-only interviews using separate sampling strategies. Currently, only BRFS landline data are available on WISH. Combined landline and cell phone data will be phased in as multiple years of jointly weighted data become available from CDC. In the interm, it should be kept in mind that some prevalence estimates may be affected by the exclusion of cell phone interview data.

The BRFS landline sample design is a stratified random sample based primarily on geography. At minimum, the sample strata correspond to the five Department of Health Services regions, with additional strata consisting of large counties and oversamples of specific populations such as African Americans. Weighted BRFS landline samples represent the adult population (ages 18+) living in households with a landline telephone. The landline sample excludes people living in institutions and other group quarters.

Within each sample stratum, telephone numbers are randomly generated, and one adult is randomly selected in each eligible household. The Centers for Disease Control and Prevention's (CDC's) BRFS protocol specifies that a new sample is opened and closed each month. Thus, the BRFS consists of 12 separate monthly samples combined to make the annual landline data set.

BRFS cell phone samples are drawn using simple random sampling techniques. Cell samples are not stratified due to the lack of correspondence between cell phone number assignment and sub-state-level geography. BRFS currently conducts cell phone interviews in cell-phone only households but not in households with both landline and cell phone service. Beginning in 2012, BRFS will conduct cell phone interviews with respondents who use a cell phone for 90% or more of received calls.

Recommendations for Analysis and Reporting of Results

Use of confidence intervals around BRFS survey estimates is strongly recommended. The Office of Health Informatics recommends basing all reported estimates on a minimum of 100 cases, or interviews; more than 100 cases is preferable. To increase the number of cases for an estimate using the BRFS Trend Data Module, use more than one year of data.

Note: For additional information about the reliability of estimates produced, it is recommended that you examine the Relative Standard Errors that accompany the results.

Information Specific to the Trends Module

• Analysis of sub-groups (e.g., women of childbearing age) requires consideration of issues concerning adequate sample size and appropriate population estimates.
   
• For results that estimate population numbers and are based on more than one year of data, users should divide the "Estimated Number in the Population" shown in the results by the number of years to resolve the inflated numbers generated by multiple years of summed weights. This will produce an average estimated population number for the time period.
   
  For example, selecting General Health Status for 2003-2005 shows an "Estimated Number in Population" of 7,198,600 who were in excellent/very good health in Wisconsin. Dividing this result by the number of years (3) yields 2,399,533.33, which rounds to 2,399,500. During 2003-2005 there were an estimated 2,399,500 adults in excellent/very good health in Wisconsin.
   
• Weighting - Trends Module
   
  BRFS weighting involves two steps. First, BRFS data sets are weighted to adjust for sample design characteristics and to remove bias due to non-response. Second, weighting incorporates known, population-level demographic characteristics to allow sample data to represent the adult population of Wisconsin.
   
  Weighted BRFS data available on WISH are representative of the non-institutionalized adult population (ages 18+) living in Wisconsin households with landline telephones. [Note: As of 2009, BRFS collects data from cell phone-only respondents as well as landline sample respondents; however, combined cell phone-landline data are not yet available on WISH.]
   
  To summarize, weighting of BRFS landline data does the following:
  
  • Corrects for differences in probability of selection due to disproportionate sampling, and adjusts for non-response and non-coverage.
  • Adjusts for distribution differences by age, race and sex between survey respondents and the Wisconsin adult population.
  • Allows the generalization of findings to the entire Wisconsin adult population living in households with landline telephones, not just to those who respond to the survey.

Information Specific to the All Counties Module

• The All Counties module is designed to provide estimates for three-year time periods, and the time periods cannot be combined. The periods are fixed to assure an adequate number of cases per county for every county.
   
• Weighting in the All Counties module is to individual county adult populations rather than the state adult population. The weighting difference is potentially important for small counties with population demographics that differ notably from the demographics of the state as a whole.

RETURN to previous page

Last Revised:  November 08, 2011