Unplanned patient readmissions are undesirable from both a clinical quality and a financial perspective. Identifying patients at risk of readmissions allows targeting of interventions aimed at reducing readmission rates. Several algorithms for predicting readmission exist, but are often inaccurate, too complex to use in the clinical setting or both. One promising model is being validated in additional settings.1 We were interested in understanding how well hospitalist physicians are able to predict which patients will be readmitted. A prior study indicating poor ability to identify patients at risk of readmission was conducted on elderly patients in a university hospital. 2This study assesses the accuracy of hospitalists in correctly identifying readmission risk in the full range of patients admitted to the general medicine service at a large community teaching hospital.
All patients discharged over a 4 month period from a hospitalist service at our institution, a 900+ bed independent academic medical center comprised the study sample. The hospitalist of record for each patient was asked at the time of admission to predict whether or not that patient would experience an unscheduled readmission within 30 days. We identified actual readmissions using our electronic health record.
One thousand three hundred seventy‐two patients were discharged during the study period under the care of the hospitalist service and survived at least 30 days beyond discharge. Of these, 36 (2.6%) were readmitted. Physician prediction sensitivity and specificity were 61% (22/36) and 91% (1216/1336), respectively. Physicians overestimated readmission, predicting readmission for 142 (10.3%) patients, almost 4 times as many as the actual number of readmitted patients. The contingency table for predicted versus actual readmission appears as Table 1.
Although physicians overestimate the number of patients that will be readmitted, they also miss almost 40% of paients who are readmitted within 30 days. Our 2.6% readmission rate during the study period was very low compared to nationally reported rates of 11 to 18 percent. These low rates of readmission may have had an impact on physicians’ ability to predict readmissions. However, our study validates the finding of the prior study that clinician ability to identify patients at risk of readmission lacks accuracy. This suggests that quality interventions aimed at reducing readmission rates may need to be applied broadly to mitigate readmission risk overall.
Potentially avoidable 30‐day hospital readmissions in medical patients: derivation and validation of a prediction model; Donzé J, Aujesky D, Williams D, Schnipper JL.; JAMA Intern Med. 2013 Apr 22;173(8):632‐8.
Inability of Providers to Predict Unplanned Readmissions; Nazima Allaudeen, MD, Schnipper JL, MD, MPH,3 E. John Orav, PhD,4 Wachter RW, MD,2 and Arpana R. Vidyarthi, MD2; J Gen Intern Med. 2011 July; 26(7): 771–776.
To cite this abstract:Schwartz E, Bhamidipati S, Kydd‐Hinderlang M, Sonnad S. Readmission Intuition: Can Physicians Accurately Predict Readmissions at the Time of Initial Admission?. Abstract published at Hospital Medicine 2014, March 24-27, Las Vegas, Nev. Abstract 135. https://www.shmabstracts.com/abstract/readmission-intuition-can-physicians-accurately-predict-readmissions-at-the-time-of-initial-admission/. Accessed December 10, 2018.