K. Y. Hu1, J. J. Blank1, Y. He2, T. J. Ridolfi1, K. A. Ludwig1, C. Y. Peterson1 1Medical College Of Wisconsin,Division Of Colorectal Surgery, Department Of Surgery,Milwaukee, WI, USA 2Medical College Of Wisconsin,Division Of Biostatistics,Milwaukee, WI, USA
Introduction:
Unplanned readmissions have negative consequences for hospitals and patients. Preoperative patient factors have been shown to be most predictive of readmission in surgical patients, with improved prediction after inclusion of postoperative variables such as laboratory values. In patients admitted to medicine services, vital sign instability on discharge has been associated with increased readmission and mortality. We hypothesized that certain abnormal laboratory values and vital signs at time of discharge may be predictive of readmission in surgical patients and attempted to identify patients at increased risk for readmission.
Methods:
This was a single-institution retrospective review of patients discharged from surgical inpatient units between 11/1/16 and 11/30/17 after admission for surgery. Patients were stratified into those with unplanned 30-day readmissions from their index admission, and those who were not readmitted. The last filled vital signs, most recent laboratory values (white blood cell count (WBC), hemoglobin, glucose, blood urea nitrogen (BUN), and albumin), number of bowel movements, ASA score, and insurance status were analyzed. Patients with planned readmissions were excluded. The primary outcome was 30-day readmission.
Results:
Of 2607 surgical admissions, 243 were readmitted within 30 days (9.1%). Readmitted patients were more likely to have an increased length of stay during their index admission (12.01 vs 6.55 days, p<0.01). In unadjusted univariate analysis, heart rate (HR) >99 (p=0.03, positive predictive value (PPV) 11.6%), BUN >23mg/dL (p<0.01, PPV 19.4%), albumin <3.8 g/dL (p<0.01, PPV 18.4%) and presence of any abnormal lab value (p<0.01, PPV 13.6%) were associated with readmission. In risk-adjusted multivariate logistic regression, variables associated with readmission were ASA of 4-5 (OR 3.31, 95% CI 1.87-5.84, p<0.01), abnormal HR (OR 1.46, 95% CI 1.07-1.98, p=0.02), and BUN >23mg/dL (OR 1.57, 95% CI 1.05-2.34, p=0.03).
Conclusion:
HR >99, BUN >23mg/dL, albumin <3.8g/dL, and presence of any abnormal lab were associated with readmission, but with poor sensitivity and weak predictive value, limiting their clinical utility. With risk-adjustment, high ASA (4-5), HR >99, and BUN >23mg/dL were associated with readmission; however, ASA and BUN may be reflective of unmodifiable patient factors and of minimal clinical significance. Although identifying key predictors at time of discharge could aid in patient counseling and optimization of high-risk discharges, our results show that readmission is challenging to predict in surgical patients based on discrete numeric data. Focus should be turned to identifying social factors that contribute to readmission.
