Abstract 12818: Identifying Drivers of Hospital Readmission Through Analysis of Discrepant Predicted and Observed Readmission Status

Abstract

Introduction: The Hospital Readmissions Reduction Program (HRRP) lowers Medicare payments to hospitals with excess readmissions for certain conditions. We analyzed FY2020 HRRP data for Brigham and Women’s Hospital (July 2015 to June 2018). We conducted an analysis to identify patients with a discrepant expected vs observed readmission status for AMI, HF and CABG, followed by a chart review to explain this discrepancy. Methods: We calculated the risk of readmission for each patient, which was a summed value of the weights associated with all recorded comorbidities in the CMS data. A negative risk score indicated a patient was unlikely to be readmitted and a positive score indicated the opposite. We then performed a chart review focused on patients who had a high risk of readmission but were not readmitted, and those who had a low risk of readmission but were readmitted. Results: For AMI, 18% (108/596) of patients were readmitted within 30 days, CABG 14% (50/357) of patients, and HF 27% (367/1382) of patients. For AMI, risk of readmission scores ranged from 2.75 to -0.095. 5/596 patients had a negative score, and none were readmitted. For CABG, scores ranged from 2.64 to -0.31, 58/357 people had a negative score, and 6 were readmitted. For HF, scores ranged from 1.94 to 0.055. There were no negative scores, but of the lowest 20/1382 scores, 2 were readmitted. We then performed a chart review of 37 patients whose readmission status was discordant with their risk score, and examined why this occurred. For patients who were low risk but were readmitted across all three conditions, 30% (range 10% - 54%) did not have a follow up appointment scheduled before discharge, 11% (range 0% - 29%) did not have an advanced care plan and 95% (range 86% - 100%) had not had a SIC. Patients who were high risk but not readmitted were evaluated but did not have any notable characteristics. We did not find any evidence of under-coding of risk comorbidities that would have led to falsely low risk scores. Conclusions: Patients with more comorbidities were more likely to get readmitted, and we found these risk factors to be accurately recorded. Clinical care insights from this project include the need for more SICs and palliative care consults and a more targeted effort to ensure patients have appropriate follow-up.