L. M. Moradi1, C. Nguyen1, L. Wood1, A. Liwo1, D. Chu1, G. Kennedy1, M. Morris1 1University Of Alabama at Birmingham,Department Of Surgery,Birmingham, Alabama, USA
Introduction: Surgical quality improvement programs measure postoperative outcomes and benchmark these outcomes against best practices. However, it is important to validate findings before ranking hospitals or making systemic changes. One measure frequently targeted is length of hospital stay and predicted length of stay (pLOS) can be calculated. The National Surgical Quality Improvement Program (NSQIP) tool calculates pLOS by entering patient specific data into its model derived from thousands of patients’ data collected by trained nurse abstractors. Vizient quality improvement program calculates pLOS using diagnosis related group (DRG) billing and data codes. Therefore, we hypothesize that NSQIP will be more accurate in predicting length of stay compared to Vizient.
Methods: We queried our institutional NSQIP database for all patients undergoing pancreatectomies, hepatectomies, and esophagectomies between Jan 2015 and Feb 2019. pLOS was calculated using the NSQIP Risk Calculator and Vizient’s reported data. Paired t-tests were used to compare pLOS to each patient’s actual surgical length of stay (sLOS). Kruskal Wallis tests were used for continuous variables and Wilcoxon tests were used for categorical variables. Patients were then stratified by procedure type and occurrence of NSQIP measured, pre-discharge complication. All tests were conducted at an alpha level of 0.05.
Results: Of 1,015 patients, 553 underwent pancreatectomy, 340 hepatectomy, and 121 esophagectomy. The median age of our cohort was 62 years and 48% were male. Overall median pLOS was significantly different between NSQIP and Vizient (7 vs. 6.1 days, p<0.03). When all patients were stratified by procedure type, NSQIP significantly overestimated LOS for all cases and Vizient overestimated LOS for pancreatectomies and hepatectomies (Table 1). In patients with pre-discharge complications, both NSQIP and Vizient significantly overestimated LOS for all procedure types. In patients with no pre-discharge complications, NSQIP accurately predicted LOS in patients having hepatectomies and esophagectomies, but not pancreatectomies, while Vizient accurately predicted LOS in pancreatectomies and esophagectomies but not hepatectomies.
Conclusion: NSQIP and Vizient overestimate postoperative LOS in patients undergoing major surgery. More work is needed to understand the differences in both LOS calculators and to confirm these findings on a larger sample size.
