sponsoring annual surveys based on self-report data from a reduced set of disability measures;
funding additional survey questions, suitable for estimation of the size of the population eligible for disability benefits, as part of, or supplement to, an ongoing household survey;
longitudinal data collection; forming a partnership with other ongoing surveys;
linking survey information with administrative databases; and
ad hoc special studies.
The NSHA should yield a complex and large set of measures that are used to identify alternative estimates of the number, distribution, and characteristics of the working age population in the United States potentially eligible for benefits under the Social Security disability programs. It is likely that the set of NSHA variables used to compute the “best” estimate of the pool of the eligible population would be too large to be feasible in ongoing monitoring because of the time needed for, and the high cost of, mounting a survey with such measures frequently.
How large a set is needed to attain stable estimates? Sensitivity and specificity criteria often favor different subsets of indicators. In any case, the practical problem for SSA is the issue of how large a data collection budget can be allocated to ongoing measurement of these indicators. One key principle of an ongoing monitoring system for disability is the cost efficiency of measuring a small number of attributes continuously. These could probably be self-report measures that require only a few minutes of interview time for the respondent. Thus, the ongoing measurement will be less expensive to support than the large, comprehensive periodic disability surveys.
What indicators should be measured continuously, and what should be measured less frequently? The set of measures in the periodic surveys defines the population of items from which the smaller set of continuous measures would be identified. Statistical analysis of the “best” sets of variables can be conducted (using item response theory notions or more traditional predictive analysis) with the goal of identifying a smaller set of measures that might be used more routinely to estimate the size of the eligible pool. Conceptually the problem of identifying the best subset of indicators devolves to measuring what portion of true eligibles is identified as eligible by the reduced set of measures (sensitivity) and what portion of actual ineligibles is identified as ineligible by the reduced set (specificity). The success of the ongoing monitoring measures depends on the success of the large periodic surveys.