R7 Analysis of BRFSS HRQOL Study
Principal Investigator: Willi Horner Johnson, PhD
The Behavioral Risk Factor Surveillance System (BRFSS) is an annual, state-based system of health surveys that generate information about health risk behaviors, clinical preventive practices, and health care access. Several questions on the BRFSS ask about Health Related Quality of Life (HRQOL). This study will examine the performance of the HRQOL questions among people with and without disabilities and how their prediction by other variables may inform development of improved health status measurement questions to assess health and well-being of individuals with disabilities regardless of functional status.
Hypothesis 1: There is adequate internal consistency among items
of the BRFSS/HRQOL to yield a summary score.
Hypothesis 2: The patterns of inter-correlation among items are
similar for persons with disabilities compared to persons without
Hypothesis 3: The concurrent validity of the BRFSS/HRQOL summary
score is similar for persons with and without disabilities.
Background to Hypotheses: Krahn, Drum, Horner-Johnson, and Goldsmith
(2004) used Andersen’s Behavioral Model to examine predictors
of health status (measured as a single question) using 1998 and
2001 national BRFSS data. In univariate analyses, disability was
the largest predictor, accounting for 14% of the variance in health
status. However, when blocks of other predictors were entered first
into a multivariate regression model with disability status entered
last, disability only accounted for 1.73% of the variance in health
status. Much of the variance originally attributed to disability
was accounted for by predisposing variables (gender, age, race/ethnicity),
enabling characteristics (employment, education, income), need
(pain, blood pressure, cholesterol levels), or personal practices
(smoking, body mass index, exercise). Disability and poor health
may be independently caused by similar factors, or pre-existing
disability may result in social disadvantages that lead to poor
health. However, disability in and of itself is not necessarily
a sufficient cause of poor health, as has been presumed by measures
that ascribe a lower health score to people with functional limitations.
In fact, more than half of people with disabilities report excellent,
very good, or good health (Drum et al., 2003). These findings suggest
that a deeper understanding is needed of variables besides disability
that are associated with health and wellness. High intercorrelations
among the standard four HRQOL questions suggest that they might
be summed into a single score. Performance
of the pain question for people with disabilities and its inclusion
in the definition of health and wellness for people with disabilities
(Putnam et al., 2003) suggests it should also be considered for
inclusion. Knowledge about the performance of the questions among
people with and without disabilities, their prediction by other
variables may inform development of improved health status measurement
questions to assess health and well-being of individuals with disabilities
regardless of functional status.
Study Design: Secondary analysis of an aggregated, population-based
data set to examine intercorrelations among items.
Sample: The proposed research study would examine
selected questions from the 2001, 2003, and potentially 2004 BRFSS
sets. The 2001 BRFSS data set aggregated for all states and territories
includes 212,510 respondents of whom 41,380 (19.5% -- unweighted
percentage) are people with self-reported disabilities. The 2003
aggregated data set contains records for 264,684 respondents, including
(22.6% -- unweighted percentage) people with disabilities.
If available in time, the 2004 data set will also be included
in the analyses. Up to three years of data will be employed to
the consistency of relationships among core variables. Additionally,
data from optional modules and rotating core questions that vary
from year to year can be analyzed by drawing on multiple years
Data Collection and Measurement Techniques: The BRFSS is a continuously
conducted random dial telephone survey conducted under the auspices
of the CDC for all states and territories. It contains a series
of core modules with questions relating to a range of topics that
affect health of the population. Optional modules vary from year
to year. In 1996, an optional disability module was introduced.
Two disability identifiers were added to the core in 2001 and were
approved for annual inclusion beginning in 2003. Limitations of
the BRFSS include bias introduced by utilizing a telephone interview
(Hough, 1999), and sampling designs that may vary from state to
state (Perrin & Koshel, 1998). However, the BRFSS is generally
considered superior to the NHIS on area, timeliness, completeness,
accuracy, and reliability (Stroup, Zack, & Wharton, 1994).
Data Analysis Methods: Responses to the four core HRQOL items
will be inter-correlated to determine the relationships of the
questions to each other for people with and without disabilities.
Internal consistency alphas will be calculated separately for people
with and without disabilities to determine whether the four items
can be used as items on a scale that will yield a single summary
score. The individual HRQOL questions, the Healthy Days Index (CDC,
2000), and the summary score will be compared with each other across
the data for each year. Regression analyses will be conducted to
determine the relationship of other variables measured by the BRFSS
(e.g. demographics, health
behaviors, receipt of preventive screenings, pain) to HRQOL scores.
Separate regression analyses will be conducted for people with
and without disabilities to determine whether the pattern of relationship
is similar. In other words, the regression analyses will identify
variables other than disability status that are related to health
status and quality of life, and that should be included in improved
measurement of health status that is not dependent on functional
ability. Analyses will be conducted in SUDAAN to account for the
complex survey sample design of the BRFSS.
Anticipated Findings: The internal consistency of BRFSS HRQOL
questions will be ascertained across a national population-based
sample of people with and without disabilities. These findings
will indicate whether these measures represent a unitary construct
that can serve as the basis for a measure of health status and
will be used in R8. The comparability of correlation patterns among
BRFSS HRQOL items among a sample of people with disabilities relative
to people without disabilities and will help to determine whether
a “universal” measure of health status is feasible.
Finally, the comparability of items predicting to a summary score
of health status for people with disabilities relative to people
without disabilities will be known. This will inform consideration
of additional areas to include in a new measure of health status.
Click here to access information on the previous BRFSS study.