Linkages between survey and administrative data can provide longitudinal information to address some disability data limitations, but longitudinal survey data can also be useful for analyzing many issues that cannot be adequately addressed with administrative data alone. As noted previously, about half of the 40 national surveys we reviewed are longitudinal or have a longitudinal component. Longitudinal data can help researchers to understand the dynamics of disability in a way that is difficult to capture in a cross-sectional survey. Disability is not a static condition; it may grow more or less severe over time, and limitations imposed by a specific condition may change as ones environment changes over time. Longitudinal surveys are generally more expensive to conduct than cross-sectional ones, both because data are collected over a longer period of time and because efforts must be made to locate sample members over time and minimize loss to follow-up. Budgetary pressures threaten the continuation of many federal efforts, including data collection. Creative methods to maintain longitudinal data collection in the face of limited finances are needed.
One option for a new survey or for existing surveys is to collect longitudinal information in a manner similar to that planned for the SIPP. The SIPP is in the process of undergoing a re-design, after which its interviews will be conducted annually instead of three times a year. To accurately account for the entire year, respondents will be given an event history calendar to aid in recall throughout the year. Similar methods could be adopted by cross-sectional surveys. For example, in addition to asking if a person has a disability, cross-sectional surveys would gain from also asking when the onset of disability occurred. Retrospective information about other important events surrounding the onset of disability could also be queried, such as consequences of disability onset for employment, income, health care use, and family well-being. Retrospective information about the availability and use of supports thought to mitigate the consequences might also be collected. Potential problems with collecting retrospective data include recall bias and the possible irrelevance of the information collected if the event of interest (like disability onset) occurred a long time in the past. Respondents might have difficulty recalling experiences, and the experiences themselves may no longer be relevant in the current social and policy context.
Another practical issue that would be associated with studying the dynamics of disability over time is developing measures to capture changes in health and disability status. Unlike factual information such as employment, income, program participation, service use, and other concepts typically measured in existing longitudinal surveys, disability is complex and multidimensional. Currently, little data are available on the validity and reliability of health and disability measures that might be collected via survey over time on the same individuals. Health scales, such as the SF-36, and functional limitation questions with four or five response categories may not be sensitive enough to capture substantive changes in disability status, and so might be of limited use in relating those changes to particular disability-related determinants and outcomes. Improving our ability to efficiently measure and interpret changes in disability status over time might lead to disability questions being added to existing longitudinal surveys.
New or enhanced longitudinal survey efforts could also provide information on other unanswered disability-related questions. Many unanswered questions identified by federal and state agencies focused on transitions. Several surveys monitor children with disabilities over time as they age into adulthood and others follow the elderly population as they retire and experience declines in functioning, but information about the transitions experienced by working-age individuals with disabilities (for example, surrounding disability onset, employment, and disability program participation) is less available. One government respondent noted that it was difficult to answer the question, What happens to persons with disabilities after they leave public assistance programs?
The PSID has the potential to provide some information on these topics, but the sample sizes are too small to be able to analyze the incidence of many types of disability-related transitions. Although information on working-age individuals before and after they participate in government programs is lacking, the act of applying for a program creates an opportunity to identify individuals in transition who might be the target of data collection efforts. A new effort could sample people who apply to a program (for example, SSDI) and who exit from the same programs. These samples might be used to augment the samples of existing surveys in a manner similar to that described previously for Social Security disability program participants in the SIPP.
A more ambitious approach would be to develop a longitudinal disability sample from the SIPP itself. That is, the SIPP could be used to identify respondents with disabilities and particularly those who, during the SIPP period, experience the onset of disability or of a medical condition that puts them at high risk for disability. These respondents could be followed for a longer period (for example, 10 years). Similar to the MCBS, new subjects would be added every year (from the current SIPP) and subjects who had completed all interviews would leave. This approach would only work well if the new SIPP sample sizes are sufficiently large, the panels are fielded on a regular basis, and questions in SIPP that are key to selection of the disability sample are maintained over successive panels. In essence, a re-designed SIPP with a special disability sample could become a longitudinal national disability survey.