Prospective evaluation of a screen for complex discharge planning in hospitalized adults

To test the predictive ability of the Probability of Repeated Admission (PRA) screen for nonroutine discharge planning (requiring new referrals for formal services).

Prospective cohort.

Two hospitals within a large Midwestern, tertiary care, referral-based system.

Nine hundred ninety-one hospitalized adults identified using a systematic sampling strategy.

The PRA screen was administered, and use of nonroutine discharge planning resources and nonroutine discharge disposition were determined using observation and open record review. Prolonged length of stay was determined by comparing the actual length of stay with the combined average length of stay for diagnosis-related groupings.

Significant differences in PRA scores existed in two of three endpoints, but the differences were small, and the ranges of scores overlapped almost completely. Using, logistic regression, items predicting use of nonroutine discharge-planning resources were self-rated health, caregiver availability, age, and sex (chi-square (chi2) = 105.7, df = 9, P <.001), accounting for 17.9% of the variability and area under receiver operating characteristic curve (AUC) of 0.74. Self-rated health and sex predicted prolonged length of stay (chi2 = 15.3, df = 5, P =.009), but only explained 2.2% of the variability, with an AUC of 0.58. The predictors of nonroutine discharge disposition were self-rated health, caregiver availability, age, sex, and diabetes mellitus (chi2 = 125.8, df = 11, P <.001), accounting for 23.0% of the variability, with an AUC of 0.79.

The clinical utility of using the PRA as a screen for early identification of persons who use nonroutine discharge planning is limited, although certain individual items may be useful.