Introduction

Background

Indiana has participated in the Behavioral Risk Factor Surveillance System (BRFSS) since its inception in 1984. The Indiana State Department of Health (ISDH) conducts the BRFSS survey through an annual cooperative agreement with the Centers for Disease Control and Prevention (CDC). All 50 states and the District of Columbia participate in the BRFSS.  This national telephone-based survey monitors modifiable risk factors associated with chronic and communicable diseases by collecting information from adults on their health behaviors and preventive practices. 

Over 400,000 adult interviews are conducted each year for the BRFSS, making it the largest continuously conducted health survey system in the world.  The surveys are conducted on a continuous basis throughout the year to determine the proportion of adults who engage in health behaviors that increase the probability of both positive and negative health outcomes. These data are also used to monitor progress toward Healthy People 2020 objectives.

In 2011, the BRFSS was enhanced by including interviews via cell phone and introducing a new weighting methodology – iterative proportional fitting (“raking”), which replaced post stratification to weight the data.  The inclusion of cellular-only/mostly households will improve survey coverage for certain population groups, (e.g., younger adults, those with lower education levels, or those with lower incomes) and result in prevalence more representative of adults in each state and the nation. 

Over the years, many ISDH public health promotion initiatives have focused not only on infectious diseases and their prevention, but also on chronic disease prevention and how certain behavioral risks are associated with these diseases.  In recent years, Indiana has obtained information on the prevalence of prediabetes, cognitive decline, consumption of sugar-sweetened beverages, and health care access surveyed among many other variables.

Core questions, which are surveyed by all participating states each year, are developed jointly by the CDC and the participating states. Standard modules on specific topics may be added to accommodate the states' individual surveillance needs. States may also add questions designed to gather data on local or regional needs.

Goals and Objectives of the BRFSS

The ISDH, in partnership with a variety of public and private programs, is responsible for planning, implementing, evaluating, and tracking disease and injury prevention programs in Indiana. The ultimate goal of the national BRFSS survey is to provide valid data to assist in these tasks. To help states achieve these goals, the CDC has established the following objectives for all BRFSS programs:

· Determine prevalence of personal health behaviors associated with the leading causes of premature death.

·
Increase public awareness of lifestyles that can significantly influence health and well-being.

·
Monitor behavioral risk factors over time and focus on factors that are not improving.

·
Assess progress in meeting the national health objectives for health promotion and disease prevention.

·
Assess the impact of state legislation on behavioral risks.

·
Share BRFSS data with state and local agencies involved with health-related issues.

Questionnaire Design

The BRFSS questionnaire has three basic parts: 1) a set of core questions used by all participating states and U.S. territories; 2) standard modules on selected topics that may rotate from year to year; and 3) state-added questions, which are developed by individual states and relate to state and local health issues. The core and standard modules are jointly developed by the states and the CDC. Because all participants use the same core questions, data can be used for comparative purposes and trend analyses.

Core health-related topics include: health status, health care access, awareness of selected medical conditions (hypertension, diabetes, and high cholesterol), injury control, tobacco and alcohol use, women's health issues, the use of certain preventive health measures, awareness and attitudes concerning HIV/AIDS, and prevalence of testing for HIV infection.  In addition to questions concerning health-related behaviors, respondents were asked to provide demographic information such as age, sex, race, marital status, household income, employment status, and education level.

Standard modules assess such risk factors as weight control, oral health, participation in leisure time and physical activities, consumption of fruits and vegetables, consumption of alcohol, and the use of smokeless tobacco.

Methodology

Indiana's sample size for 2016 was 11,066 randomly selected Indiana residents aged 18 years or older. A disproportionate stratified random sample design was used to generate the sample of telephone numbers. In this design, information obtained from previous surveys was used to classify 100-number blocks of telephone numbers into strata that were either likely or unlikely to yield residential numbers. Telephone numbers in the likely stratum were sampled at a higher rate than numbers in the unlikely stratum.

A sample was drawn independently by randomly sampling the telephone numbers within each stratum. Then, when an interviewer called a sampled household, one adult living in the household was randomly selected to be the respondent. The completed survey sample was designed to produce a representative sample of all adult Indiana residents.

Interview Process and Quality Assurance Measures

In 2016, the ISDH contracted with Clearwater Research, Inc., for telephone interviewing, data preparation, and data quality measures. A computer-assisted telephone interviewing system (CATI) was used in the interview process. A CDC-developed protocol was followed to process the data. When monthly interviewing was completed, the data were summarized utilizing a computer program provided by CDC. The summary and the data were then submitted to CDC whose staff edited, corrected, compiled, and weighted the state data into an annual file that was provided to the BRFSS Program Coordinator at the ISDH.

CDC mandates quality assurance of the survey data.  The quality of the data collected from respondents was evaluated daily to assure proper completion. Unobtrusive monitoring of the interviews and data entry was conducted to ensure adherence to protocols.  Call-backs were also done randomly, on a spot-check basis, to confirm that interviews had been conducted as indicated.

Response Rates

The objective of the CDC data quality protocol is to provide the most accurate data possible. Because non-responses tend to bias the results of a survey, special attention was given to minimizing the non-response rate. The extent to which completed interviews were obtained from among all phone numbers selected is indicated by several different measures of response rate. A high response rate indicates a lower potential for bias in the data.

CDC summarized the quality of the 2016 BRFSS survey data.  Response rates for the BRFSS are calculated using standards set by the American Association of Public Opinion Research (AAPOR). 

AAPOR Response Rate -

These calculations include assumptions of eligibility among potential respondents/households that are not interviewed.  Changes in the distribution of cell phone numbers by telephone companies and the portability of landline telephone numbers are likely to make it ascertaining which telephone numbers are out-of-sample and which telephone numbers represent “likely households” much more difficult than in past years.  In those instances, the BRFSS uses proportions of unknown households in each of the states to estimate the total number of households from those whose eligibility is undetermined.  This “eligibility factor” appears in calculation of the response rate provided below:

 

Overall Response Rate (%)

Indiana -Landline

41.5

US Median-Landline

47.7

Indiana –Cell Phone

41.4

US Median – Cell Phone

46.3

Indiana - Combined

41.5

US Median - Combined

47.1

                   Source:  2016 Summary Data Quality Report, CDC

Survey Limitations

The BRFSS survey relies on self-reported data and has certain limitations. These limitations should be understood in the interpretation of the data. Many times, respondents have the tendency to underreport some behaviors that may be considered socially unacceptable, unhealthy, or even illegal such as high alcohol consumption, drinking and driving, or not using seat belts. Conversely, respondents may overreport behaviors that are considered desirable (the amount of exercise, low body weight, or regular health screenings). Some information is also affected by the ability of the respondent to recall past behaviors and respond accordingly. The validity of survey results depends on the accuracy of the responses to the survey questions from recalled past behaviors.

The BRFSS survey excludes households without telephones, which may result in a biased survey population due to underrepresentation of certain segments of the population. Additionally, breaking down the data into smaller categories decreases the sample size of the original risk factor categories, thereby decreasing the ability to determine statistically significant differences.

Finally, it should be noted that weighting the data by age, race, and sex distribution was done in order to correct for over- or underrepresentation of all groups. Prevalence based on denominators of less than 50 respondents was considered statistically unreliable.

Weighting of Data

Starting with 2011, two changes in the method of the BRFSS were introduced:

·       The inclusion of cell phone interviews.  In 2011, approximately 30 percent of American homes had only cellular telephones.  In 2016, 50.5% of households had only cellular telephones (National Center for Health Statistics).  The trend towards cell-only households has been especially strong among younger adults and among persons in racial and ethnic minority groups.

·       Introduction of a new “weighting” procedure.  BRFSS responses are weighted to account for differences between respondents and the target population – adults ages 18 years and older not living in institutional settings.  Compared to the “post stratification” weighting method used by BRFSS for more than two decades, the new method, iterative proportional fitting (nicknamed “raking”), improves the accuracy of the BRFSS by allowing the use of more demographic variables and interview type (landline or cell phone) in the weighting.

In 2016, the core questionnaire and optional modules of the Indiana BRFSS survey were asked on both the landline and cell phone versions of the survey.  In the tables, _LLCPWT is noted in the denominator when the questions were asked of both landline and cell phone respondents.    

As a result of the inclusion of cell phones and change in survey methodology, results from the 2011-2016 BRFSS are not directly comparable to prior years.       

National BRFSS Prevalence Summary

The CDC Behavioral Surveillance Branch provides a BRFSS summary prevalence report on selected risk factors and preventive health measures from its survey for each participating state. This summary report consists of tables capturing data provided by each participating state indicating whether individual states are at, above, or below the national median prevalence value.

The summary also provides a tool to determine how states compare with each other and to the Healthy People 2020 objectives related to the risk factors that BRFSS measures.  Summary information includes the name of the participating state, the sample size, the prevalence of the risk behavior, and the confidence interval for the prevalence figure.  An additional table was developed using information from the 2016 BRFSS Prevalence Summary.  It includes information for the 50 states and the District of Columbia for a selected number of risk factors. For each measure shown, a rank of 1 indicates the negative end of the ranking.  For all measures, a low percentage indicates the desired prevalence.

A further comparison of risk factors to national objectives is provided in Appendix A.  These tables compare Indiana risk factor prevalence from BRFSS data with national objectives and national BRFSS prevalence for those behaviors for which Healthy People 2020 have explicitly stated objectives and that were measured in Indiana in 2016.