Frequently Asked Questions and Answers on BRFSS Data Analysis and Methods BRFSS Epidemiologists and Staff Division of Behavioral Surveillance PHSIPO, OSELS, CDC Q1: Why Is Data Analysis Important? A1. The goal is to transform data into information, and information into insight. (Carly Fiorina, HP CEO) 5/7/2012 2 Q2: How Reliable Are BRFSS Data? A2. Results are consistent in prevalence estimates of key health indicators of similar caliber between BRFSS and other national surveys. 5/7/2012 3 5/7/2012 4 Q3. Who Uses BRFSS Data? State and Local Health Departments CDC Academic Researchers Health Professionals Nonprofit Organizations Insurance Companies Managed Care Organizations Students The Media The Military 5/7/2012 5 5/7/2012 6 CARRFS e-learning webinar 1
BRFSS by the States BRFSS Annual Report BRFSS by the States BRFSS Newsletter (News Brief) 5/7/2012 7 5/7/2012 8 BRFSS by the States BRFSS, 1985 BRFSS Reports on Special Topics 5/7/2012 9 5/7/2012 10 BRFSS, 1986 BRFSS, 1987 5/7/2012 11 5/7/2012 12 CARRFS e-learning webinar 2
BRFSS, 1988 BRFSS, 1989 5/7/2012 13 5/7/2012 14 BRFSS, 1990 BRFSS, 1991 15% 19% 5/7/2012 15 5/7/2012 16 BRFSS, 1992 BRFSS, 1993 15% 19% 15% 19% 5/7/2012 17 5/7/2012 18 CARRFS e-learning webinar 3
BRFSS, 1994 BRFSS, 1995 15% 19% 15% 19% 5/7/2012 19 5/7/2012 20 BRFSS, 1996 BRFSS, 1997 15% 19% 15% 19% 20% 5/7/2012 21 5/7/2012 22 BRFSS, 1998 BRFSS, 1999 15% 19% 20% 15% 19% 20% 5/7/2012 23 5/7/2012 24 CARRFS e-learning webinar 4
BRFSS, 2000 BRFSS, 2001 15% 19% 20% 15% 19% 20% 24% 25% 5/7/2012 25 5/7/2012 26 BRFSS, 2002 BRFSS, 2003 15% 19% 20% 24% 25% 15% 19% 20% 24% 25% 5/7/2012 27 5/7/2012 28 BRFSS, 2004 BRFSS, 2005 15% 19% 20% 24% 25% 15% 19% 20% 24% 25% 29% 30% 5/7/2012 29 5/7/2012 30 CARRFS e-learning webinar 5
BRFSS, 2006 BRFSS, 2007 15% 19% 20% 24% 25% 29% 30% 15% 19% 20% 24% 25% 29% 30% 5/7/2012 31 5/7/2012 32 BRFSS, 2008 BRFSS, 2009 15% 19% 20% 24% 25% 29% 30% 15% 19% 20% 24% 25% 29% 30% 5/7/2012 33 5/7/2012 34 BRFSS by CDC Healthy People 2010 and 2020 HP 2010 Objectives Tracked by BRFSS Risk Factors Binge drinking Smoking Physical inactivity Use of Preventive Services Immunization Cancer screening Cholesterol screening Health Care Access Health insurance Personal health care provider Health Conditions Diabetes Obesity High blood pressure 5/7/2012 35 5/7/2012 36 CARRFS e-learning webinar 6
BRFSS by CDC Morbidity and Mortality Weekly Report The MMWR series is prepared by CDC. Often called the voice of CDC. In the past 10 years, about 7% of reports published in MMWR used BRFSS data (both from CDC and states). Topics include surveillance summary reports, trends in prevalence, vital signs, and many other analyses. MMWR 5/7/2012 37 5/7/2012 38 Publications From 1987 to 2010: a total of 1,450 papers published using BRFSS data (PubMed: BRFSS or Behavioral Risk Factor Surveillance System on title or abstract). From 1987 to 2010: 1743 papers used or mentioned BRFSS data (PubMed: all fields). 5/7/2012 39 5/7/2012 40 BRFSS by CDC BRFSS Annual Conference Popular Media 5/7/2012 41 5/7/2012 42 CARRFS e-learning webinar 7
Q4: Why Is Sample Weight Needed in Data Analysis? A4. To account for the complex survey design and produce accurate estimates. Q5: Are Sample Weights Readjusted When Combined Many Years of Data? A5. It may be better to re-calculate the sample weights (weight divided by number of years). However, it may not be needed for trend analysis. 5/7/2012 43 5/7/2012 44 Q6: Why is age-adjustment needed even sample weight has been used? A6. Sample weight is used to represent the target population. Age-adjustment is used to control for the possible confounding effect of age. Q7: What standard population is recommended for age-adjustment? A7. 2000 projected U.S. population is recommended (Klein RJ, Schoenborn CA. Statistical Notes, No. 20, 2001). 5/7/2012 45 5/7/2012 46 Q8: Can I make up my own age distribution? A8. Yes. You can collapse age groups according to your data. Q9: What statistical software programs are used in BRFSS data analysis? A9. SUDAAN>SAS SURVEY PROCEDURES, SPSS SURVEY PACKAGE, STATA SURVEY MODULE. 5/7/2012 47 5/7/2012 48 CARRFS e-learning webinar 8
Q10: Do SUDAAN and SAS produce identical estimates? A10. Yes, when there is no single sample unit or PSU in a stage or stratum. Q11: Is direct ageadjustment the same as predictive marginal? A11. PM is equivalent to direct ageadjustment in the saturated model (i.e. include all-way interaction terms). 5/7/2012 49 5/7/2012 50 Q12: How to estimate prevalence ratios? A12. In SUDAAN: Loglink, or rlogistic with adjrr option in the predmarg statement; In SAS: proc gemod (dist=poisson, link=log). Thank You! For further questions, comments, and suggestions: Chaoyang Li (404) 498 0574; cli@cdc.gov 1600 Clifton Road NE, MS E97 Atlanta, GA 30333, USA http://www.cdc.gov/brfss/ 5/7/2012 51 5/7/2012 52 CARRFS e-learning webinar 9