Unadjusted Prevalence Rates: Why they Still Matter for Older Adults' Disability Rates

We thank Martin and colleagues for their thoughtful and important comments. We very much appreciate their supplementary age-standardized analysis of the public use American Community Survey (ACS) data. Their findings give support to our suggestion (1) that the rise in prevalence of activities of daily living (ADL) disabilities between 2000 and 2005 among those aged 65 and older was largely driven by the disproportionate growth in the oldest age category. Their findings also have important ramifications in the projection of the burden of disabilities in the American population aged 65 and older. Projections suggest that the percentage of senior citizens who are older than age 85 will increase from 12.2% in the year 2000 to 24.1% in the year 2050 (2). We deliberately chose to report the crude (unadjusted) trend in prevalence for those aged 65 and older rather than age-adjusted figures for two reasons. Most importantly, the government is currently mandated to provide assistance in many programs (eg, Medicare) to those aged 65 and older. Therefore, it is the unadjusted prevalence rate of disability that is a measure of the burden of care and is of immediate salience to policy makers. Secondly, our report is in keeping with a National Institute of Aging funded technical working group that chose to compare unadjusted estimates of disability across five national data sets (3). This group justified its decision to use crude rates because “conclusions about standardized rates are sensitive to the choice of age distribution” and “stratification by age, gender and other subgroups of interest leads to an unmanageable number of comparisons.” (3, p. 421). Martin and colleagues also raised the concern that the change in interview mode between 2000 and 2005 made our estimates of trends in disability unreliable. We did a sensitivity analysis of the relative change and find it unlikely that the change in mode would substantially alter our results. Unfortunately, we did not have access to information on interview mode for the population aged 65 and older. However, relying on the information Dr. Martin and colleagues reported for the population aged 16–64, there was a 6% change in interview mode in the period of the study from 58% via mail in 2000 to 52% via mail in 2005. Martin and colleagues also reported that in 2003, the prevalence of ADL limitations for computer-assisted interviewees was 1.9% versus 1.8% for the mail responders. Thus, the difference in ADL limitation rates between responders in each mode was 0.1%. The relative difference between individuals in the two mode groups was 0.1/1.8 = 5.6%. With a 6% change in interview mode between 2000 and 2005, the estimated change in ADL limitations due to mode change was very modest (.06 × .056 =.0033 = 1/3 of 1%.). As reported in our footnote of Table 1a, the relative increase in ADL limitations during the period was 9.1%, a figure 30 times bigger than the estimated increase due to mode change. In addition, for the population aged 16–64, the percent reporting an ADL disability within each mode of response appeared to be increasing between 2000 and 2003. Stern and Brualt (4) reported that for mail respondents, the prevalence of ADL limitations increased from 1.7% in 2000 to 1.8% in 2003. For computer-assisted interview (CAI) respondents, the prevalence of ADL limitations increased from 1.8% in 2000 to 1.9% in 2003. This indicates a relative increase of 5.9% for mail respondents (0.1/1.7) and 5.6% for CAI respondents (0.1/1.8). These percentages are similar to the 5.4% estimated relative increase in ADL limitations for the population aged 65 and older during a 3-year period using our regression equation from Table 1a ((3 y × 0.16)/8.84). Martin and colleagues have highlighted two important issues. Researchers wishing to investigate age-adjusted trends in disability and to explore more closely the impact of mode change would benefit from use of the embargoed ACS data. We believe that the ACS provides an important tool for assessing trends in the crude prevalence rates of late-life disabilities. As of 2006, the ACS has included institutionalized older adults in the survey that will make future estimates of the trends in total burden of disability in the older US population more precise.