Transcript BRFSS - North Carolina Public Health Association

An Overview of the North
Carolina Behavioral Risk Factor
Surveillance System (BRFSS)
Presentation to the
State Health Director's Conference
January 23, 2014
State Center for Health Statistics
Division of Pubic Health
NC Dept. of Health & Human Services
Presentation Outline

How the survey is done







Questionnaire development
Sampling
Data Collection
Weighting
Future developments
Demonstrate how to access results
Q&A
State Center for Health Statistics
Division of Pubic Health
NC Dept. of Health & Human Services
Timeline for 2014 BRFSS
2012
External review of proposed new/revised questions
2013
State Coordinators vote on proposed questions
CDC field tests questionnaire
CDC Core & Optional Modules checked and finalized
State-added questions reviewed and finalized
CATI coding developed and checked
Interviewers trained
2014
Conduct interviews January through December
2015
CDC sends final weighted data to NC
SCHS reviews data and produce web tables
BRFSS QUESTIONNAIRE
Questionnaire Organization

Each years survey is divided into three main
parts
CDC
Core
Questions
Mandated by CDC;
must ask all questions
exactly as written
CDC
Optional
Modules
States choose which
modules they will
administer; must ask
all questions exactly
as written
StateAdded
Questions
Developed by each
state’s coordinator
The Core




The CDC Division of Behavioral Surveillance (DBS) works with
other CDC programs to develop questions for the core & optional
modules
Proposed questions undergo two rounds of cognitive testing
State Coordinators have input via
 Participation in State Working Group
 Participation as external reviewer for proposed questions
 Vote on proposed questions taken at Annual Conference
Questions rotate in & out of core on a fixed schedule
 Just adopted a new 5-year plan for the core
CDC Optional Modules



Process for developing optional modules is
similar to the core
CDC programs pay DSB for space on the
BRFSS
So the selection of optional modules
available varies each year
State-Added Questions

May include



Questions from other surveys, including questions
on CDC Optional Modules
Questions written by state staff
Each state has a process for soliciting
proposed questions
Survey time is a very scarce resource

When telephone interviews run past 15 to
20 minutes,




Break-offs increase
Data quality suffers as people say whatever
comes to mind so you’ll leave them alone!
Some cell phone respondents are more
sensitive to length because they pay by
the minute
Pew Research Center successfully does
20 minute cell phone interviews

The 2013 NC BRFSS interview averaged 29
minutes



17 minutes of that is take up by Core questions and an
expanded Optional Module on Health Care Access
“Core creep” – expansion of the length of the core – is a
major issue with state coordinators
We have been trying to cut down on state-added
and CDC Optional Modules to shorten the interview
SAMPLING
Random Digit Dialing (RDD) Surveys



4 digit numbers are generated at random
within each combination of area code and
phone exchange within a geographic area
Resulting numbers are screened to identify
non-working numbers
Results in a random sample of telephone
numbers for a given geographic region
BRFSS uses a Dual-Frame RDD
Sample

Landline Frame


Numbers generated within area code/exchange
combinations for landline telephones
Telco companies share more information
regarding these numbers

Most importantly where the phone is located

Cell Phone Frame


Numbers generated within area code/phone
exchange combinations for cell phones
Cell phone companies share little information
about these numbers

Little information on where owner of phone lives until
they are interviewed
Increase in “Cell Phone Only” (Wireless)
Households



Roughly 38% of adults as of
the end of 2013.
These folks are:
 Younger
 More are minorities
 Less affluent
 Have some significantly
different health
conditions & behaviors
Must sample this population
to avoid biased estimates
DATA COLLECTION
“BRFSS is a process, not a project”
Survey runs throughout calendar year

Each month we receive separate landline and
cell phone samples (~8,000 phone numbers
total)
To complete each month’s “study”





Each landline number is called up to 15 times
Each cell phone number is called up to 8 times
We call seven days a week around 330
days every year
More difficult to reach cell phone
respondents

In the landline sample




Cell




Get one completed interview every 1.4 hours of interview
time
Perform 38 dialings per completed interview
4,600 sample records yield ~900 completed interviews
Get one completed interview for every 1.7 hours
Perform 97 dialings per completed interview
3,200 sample records yield ~100 completed interviews
Bottom Line: We have to call many, many more cell
phone numbers before finding someone who will
complete the interview
WEIGHTING
Why Weight?


We weight survey data to make our sample
better match the population
Some observations “get counted” more than
others
BRFSS Weighting Systems




The old BRFSS system used gender, age,
race/ethnicity, & region within state to adjust the
final results.
The new rake weights adds adjustment by
education level, marital status, renter/owner status &
phone source (landline vs. cell phone)
The new “raked” weights should improve the
representativeness of the sample, particularly
regarding socioeconomic status
The new weights, together with adding cell-phone
interviews, produces different estimates for some
health indicators
Some estimates change a lot:
Percent of Adults Who Are Current Smokers, 2010
30
24.12
25
20
25.26
19.78
Percentage 15
10
5
0
Landline only, old weight
Landline only, raked weight
Landline & cell, raked weight
Sample & Weight Used
Estimate increases by ~ 5.5 percent
Binge Drinking
16
14.28
14
Percentage
12
11
11.16
Landline only, old weight
Landline only, raked weight
10
8
6
4
2
0
Landline & cell, raked weight
Sample & Weigh Used
Estimate increases by ~ 3.3 percent
Some estimates change very little:
Percentage of Adults Diagnosed with Diabetes, 2010
14
11.62
12
Percentage
10
10.74
9.77
8
6
4
2
0
Landline only, old weight
Landline only, raked weight
Landline & cell, raked weight
Sample & Weight Used
Estimate increases by ~ 1.0 percent
Percentage of Adults Getting Recommended Amout of Physical Activity, 2009
50
45
40
35
30
Percentage 25
20
15
10
5
0
46.41
Landline only, old weight
43.98
45.46
Landline only, raked weight
Landline & cell, raked weight
Sample & Weight Used
Estimate increases by ~ 0.9 percent
FUTURE DEVELOPMENTS
Change to Geographic Strata

Beginning in 2014, geographic strata will be
based on Area Health Education Center
regions

Able to stratify both landline and cell phone
samples for these area

We will produce estimates for




State as a whole
Eastern, Piedmont & Western Carolina
AHECS
Counties when possible

At least 500 completed interviews in the sample
HOW TO ACCESS RESULTS


Data for a calendar year are combined and
analyzed
Results for state and sub-regions are posted
to SCHS Web page

Main SCHS Web Page


http://www.schs.state.nc.us/
NC BRFSS Web Page

http://www.schs.state.nc.us/units/stat/brfss/
Contact Information:

James Cassell
Head of Survey Operations & BRFSS Coordinator
State Center for Health Statistics
North Carolina Division of Public Health
2422 Mail Service Center
Raleigh, NC 27699-2422
Voice: 919-855-4485
Fax: 919-715-7899
Email: [email protected]
NC BRFSS Web Page
http://www.schs.state.nc.us/units/stat/brfss/
State Center for Health Statistics
Division of Pubic Health
NC Dept. of Health & Human Services