Effects of a computerized preadmission screening program on appropriateness of inpatient categorization

A preadmission screening system was implemented at a 520 bed acute care hospital in Salt Lake City, UT. The system integrated a PC-based expert system with existing mainframe hospital information system (HIS). The screening system assessed the appropriateness of patients for the inpatient setting. The effects of the computerized screening system were measured in a randomized control trial. As a surrogate of inappropriateness, the trial measured the change in the number of patients whose categorization as inpatient was retrospectively reconsidered and changed (reclassified) to meet reimbursement requirements. Medicare inpatient admission for the 20 week period April 12 and August 30, 1991 were reviewed using the computerized system. Patients in the experimental cohort not meeting the inpatient admission criteria were identified. The physicians of these patients were notified that the admission was not appropriate for the inpatient setting. A recommendation was made to admit patients to an alternate level of care (such as outpatient, observation of short stay). The rate of retrospective reclassifications in the experimental and control groups was compared. The preadmission screening system shoed a reduction in the rate of reclassifications per admission in the experimental group (3.6%) as compared to the control group (3.9%), The reduction was not significant (p = 0.43). Physician compliance with the intervention recommendation was high (78.3%). The rate of successful intervention was low (31.8%). The low rate was attributed to reviews completed after admission or after discharge, too late for physician intervention. The capability of the computerized expert system to predict reclassification was moderate (sensitivity = 46.0%, specificity - 96.3%), However, the expert system was not sensitive enough, even with ideal intervention, to eliminate the need for post discharge review for reclassifications. Reclassification was found to reflect, in part, nonclinical review criteria and, therefore, was poor surrogate for inappropriate inpatient admission. In was concluded that computerized preadmission screening in NOT effective in reducing postdischarge changes to patient type (reclassifications). Preadmission screening my be effective in detecting inappropriate inpatient admissions.