Identification of risk for high hospital use: cost comparisons of four strategies and performance across subgroups.

Objectives: To determine the relative costs of four risk-identification strategies and compare their performance in predicting hospital use by different subgroups of older persons based on age, sex, and prior hospital use.
Design: Prospective validation study and cost-comparison analysis.
Setting: Community-based.
Participants: Five thousand one hundred thirty-eight participants of the sixth wave of three sites of the Established Populations for Epidemiologic Studies of the Elderly.
Measurements: Four strategies (prior hospitalization data only, a 10-item self-report screen alone, self-report combined with two laboratory tests, and sequential self-report plus as-needed use of laboratory tests when the self-report screen is inconclusive) and 3-year Medicare Part A hospital cost data.
Results: Assuming that interventions based on screening would yield a total benefit of $1,000 per true-positive case and a cost of $400 for each false-positive case, the sequential strategy was slightly less expensive than the self-report only strategy; both were considerably less expensive than the combined or hospitalization-only strategies. Accuracy as measured by the area under the receiver operating characteristic curve for the sequential strategy was comparable for all subgroups (between 0.62 and 0.70) but was least accurate for those who had high prior use and for those aged 85 and older.
Conclusion: A sequential screening strategy that administers laboratory tests selectively is slightly less expensive than one that uses only self-report items. This strategy is also accurate in both sexes, in those with various degrees of prior use, and in the oldest old.