FAQs (Data Related) Data Analysis
- Is it acceptable to pool years of data with identical BRFSS arthritis questions to obtain a sufficient sample size for population subgroups? Are we really double counting when we do this?
We recommend you either pool data from:
- From 1996 to 2001, when one set of questions was used
- From 2002 to 2009, when a different set of questions was used. Pooling data from these two separate periods is not legitimate because the underlying survey questions were different (see #2 below).
- From 2011 and possibly afterwards. Pooling these data with data from a previous period is not legitimate because major changes in BRFSS methods occurred in 2011. These changes included added surveys of cell phone only users as well as a change in survey processing procedures. Understanding how these changes affect years subsequent to 2011 is ongoing.
Combining two or more years of data within a period can generate more precise prevalence estimates. If you divide the final weight by the number of years you pooled, there will not be double or triple counting.
Note: Starting in 2011 the BRFSS changed the order of arthritis questions by placing the case definition question in a Core Chronic Health Conditions Module and leaving the remaining 4 questions in the separate Core Arthritis Burden Module. The effect of changing the order of the questions on participant responses is not clear.
Estimates of trends can run from 1996–2001 or from 2002-2009 or possibly from 2011 forward but not across these periods. It would be impossible to know if any observed change across these periods is real or due to the changes cited in #1 above.
Regarding trends, it is reasonable to expect an increase in arthritis prevalence over time because the population is aging and we know there are higher rates of arthritis among older persons. Further information is available on: arthritis case definition changes, an overview of BRFSS arthritis question changes and BRFSS questions listed by year.
Prevalence is the number of all people with arthritis in a given population at a given time. The number of adults age 18 years or older with arthritis in Georgia in 2013 is an example of a prevalence estimate. This is the most common estimate of arthritis occurrence in a population.
Incidence is the number of new persons whose arthritis begins during a given time period in a given population. This is a rarely used estimate of arthritis occurrence because the date of onset of new cases is difficult to determine.
Note: Incidence cannot be calculated from NHIS or BRFSS data.
Estimating the incidence of arthritis, that is, the number of new cases of arthritis in a defined population over a defined time (usually a year), is very difficult. It requires knowing the disease status of everyone in the defined population at the start of the defined time period (to remove those with existing cases of arthritis from the estimates) and then counting every new case that occurs until the end of the time period.
These challenges have meant that incidence studies have been done in fairly small populations in geographically small areas over relatively short time periods and, thus, may not be not generalizable to the U.S. population or even a state population. Consequently, we have no national estimate of arthritis incidence.