Understanding the Accuracy of Clinician Provided Estimated Discharge Dates

Discharge planning is a vital tool in managing hospital capacity, which is essential for maintaining hospital throughput for surgical postoperative admissions. Early discharge planning has been effective in reducing length of stay and hospital readmissions. Between 2014 and 2017, Vanderbilt University Medical Center (VUMC) implemented a tool in the electronic health record (EHR) requiring providers to input the patient’s estimated discharge date on each hospital day. We hypothesized discharge estimates would be more accurate, on average, for surgical patients compared to non-surgical patients because treatment plans are known in advance of surgical admissions. We also analyzed the data to identify factors associated with more accurate discharge estimates. In this retrospective observational study, via an analysis of covariance (ANCOVA) approach, we identified factors associated with more accurate discharge estimates for admitted adult patients at VUMC. The primary outcome was the difference between estimated and actual discharge date, and the primary exposure of interest was whether the patient underwent surgery while admitted to the hospital. A total of 304,802 date of discharge estimate entries from 68,587 inpatient encounters met inclusion criteria. After controlling for measured confounding, we found the discharge estimates were more precise as the difference between estimated and actual discharge date narrowed; for each additional day closer to discharge, prediction accuracy improved by .67 days (95% confidence interval [CI], 0.66 to 0.67; p < 0.001), on average. No difference was observed on the primary outcome in patients undergoing surgery compared with non-surgical treatment (0.02 days; 95% CI, 0.00 to 0.03; p = 0.111). Faculty members were found to perform best among all clinicians in predicting estimated discharge date with a 0.24-day better accuracy (95% CI, 0.20 to 0.27; p < 0.001), on average, than other staff. Weekend and holiday, specific clinical teams, staff types, and discharge dispositions were associated with the variability in estimated versus actual discharge date (p < 0.001). Given the widespread variation in current efforts to improve discharge planning and the recommended approach of assigning a discharge date early in the hospital stay, understanding provider estimated discharge dates is an important tool in hospital capacity management. While we did not determine a difference in discharge estimates among surgical and non-surgical patients, we found estimates were more accurate as discharge came nearer and identified notable trends in provider inputs and patient factors. Assessing factors that impact variability in discharge accuracy can allow hospitals to design targeted interventions to improve discharge planning and reduce unnecessary hospital days.