Derivation and Validation of a Model to Predict 30-Day Readmission in Surgical Patients Discharged to Skilled Nursing Facility

Abstract

ObjectivesTo identify factors associated with 30-day all-cause readmission rates in surgical patients discharged to skilled nursing facilities (SNFs), and derive and validate a risk score. DesignRetrospective cohort. Setting and participantsPatients admitted to 1 tertiary hospital's surgical services between January 1, 2011, and December 31, 2014 and subsequently discharged to 110 SNFs within a 25-mile radius of the hospital. The first 2 years were used for the derivation set and the last 2 for validation. MethodsData were collected on 30-day all cause readmissions, patient demographics, procedure and surgical service, comorbidities, laboratory tests, and prior health care utilization. Multivariate regression was used to identify risk factors for readmission. ResultsDuring the study period, 2405 surgical patients were discharged to 110 SNFs, and 519 (21.6%) of these patients experienced readmission within 30 days. In a multivariable regression model, hospital length of stay [odds ratio (OR) per day: 1.03, 95% confidence interval (CI) 1.02-1.04], number of hospitalizations in past year (OR 1.24 per hospitalization, 95% CI 1.18-1.31), nonelective surgery (OR 1.33, 95% CI 1.18-1.65), low-risk service (orthopedic/spine service) (OR 0.32, 95% CI 0.25-0.42), and intermediate-risk service (cardiothoracic surgery/urology/gynecology/ear, nose, throat) (OR 0.69, 95% CI 0.53-0.88) were associated with all-cause readmissions. The model had a C index of 0.71 in the validation set. Using the following risk score [0.8 × (hospital length of stay) + 7 × (number of hospitalizations in past year) +10 for nonelective surgery, +36 for high-risk surgery, and +20 for intermediate-risk surgery], a score of >40 identified patients at high risk of 30-day readmission (35.8% vs 12.6%, P < .001). Conclusions/ImplicationsAmong surgical patients discharged to an SNF, a simple risk score with 4 parameters can accurately predict the risk of 30-day readmission.