Assessing readmission after general, vascular, and thoracic surgery using ACS-NSQIP

Abstract

Objective: In 2012, Medicare began cutting reimbursement for hospitals with high readmission rates. We sought to define the incidence and risk factors associated with readmission after surgery.

Methods: A total of 230,864 patients discharged after general, upper gastrointestinal (GI), small and large intestine, hepatopancreatobiliary (HPB), vascular, and thoracic surgery were identified using the 2011 American College of Surgeons National Surgical Quality Improvement Program. Readmission rates and patient characteristics were analyzed. A predictive model for readmission was developed among patients with length of stay (LOS) 10 days or fewer and then validated using separate samples.

Results: Median patient age was 56 years; 43% were male, and median American Society of Anesthesiologists (ASA) class was 2 (general surgery: 2; upper GI: 3; small and large intestine: 2; HPB: 3; vascular: 3; thoracic: 3; P < 0.001). The median LOS was 1 day (general surgery: 0; upper GI: 2; small and large intestine: 5; HPB: 6; vascular: 2; thoracic: 4; P < 0.001). Overall 30-day readmission was 7.8% (general surgery: 5.0%; upper GI: 6.9%; small and large intestine: 12.6%; HPB: 15.8%; vascular: 11.9%; thoracic: 11.1%; P < 0.001). Factors strongly associated with readmission included ASA class, albumin less than 3.5, diabetes, inpatient complications, nonelective surgery, discharge to a facility, and the LOS (all P < 0.001). On multivariate analysis, ASA class and the LOS remained most strongly associated with readmission. A simple integer-based score using ASA class and the LOS predicted risk of readmission (area under the receiver operator curve 0.702).

Conclusions: Readmission among patients with the LOS 10 days or fewer occurs at an incidence of at least 5% to 16% across surgical subspecialties. A scoring system on the basis of ASA class and the LOS may help stratify readmission risk to target interventions.

FIGURE 1

Average readmission rate by length of stay for general, vascular, and thoracic surgery. Data fail to approximate the theoretical relationship at increasing length of stay because of immortal person-time bias.

FIGURE 2

Thirty-day readmission by readmission score and subspecialty for the LOS 10 days or fewer. Readmission score = LOS/2 + ASA class, rounded up. Confidence intervals of 95% are plotted.

FIGURE 3

Distribution of readmission score among all and readmitted patients for the LOS 10 days or fewer. Readmission score = LOS/2 + ASA class, rounded up. S&L intestine indicates small and large intestine.