Abstract
Background: Current risk-stratification models insufficiently identify readmission risk.
Setting: Academic medical center in Boston, MA.
Patients: One hundred seventy-seven medicine inpatients.
Methods: We prospectively interviewed clinicians about whether they would be surprised if patients scheduled for discharge were readmitted within 30 days and to identify one patient at the highest risk. Multivariate models examined the impact of clinicians' judgment on readmission.
Results: The 30-day same-hospital readmission rate was 10.7%. The number of hospitalizations (odds ratio [OR], 1.16; 95% confidence interval [CI], 1.04-1.30), emergency department visits (1.10, 1.02-1.19), and discharge medications (1.07, 1.00-1.14) were associated with readmission in bivariate models. The negative-predictive value when clinicians would be surprised about a readmission was high (95%).
Conclusion: Clinicians are better at predicting those not readmitted than those who are.