Composite Length of Stay, An Outcome Measure of Postoperative and Readmission Length of Stays in Pancreatoduodenectomy

Purpose: Postoperative length of stay (PLOS) and readmission rate are pancreatoduodenectomy (PD) outcome measures, which are reported individually but may be interrelated. The purpose of this study was to evaluate how well a composite length of stay measure (CLOS) that included PLOS and readmission length of stay describes outcomes. To do so, we evaluated how well CLOS correlated to postoperative complications absolutely and compared to PLOS. Methods: A total of 668 PDs performed between 2011 and 2018 were evaluated. CLOS was calculated from PLOS and readmission length of stay. Complication severity was judged by the Modified Accordion Grading System (MAGS). Multinomial logistical regression models (MLRM) were used to investigate the relationship between either PLOS or CLOS and complications. Multilevel and pairwise area under curves (AUC) using SAS macro %MultAUC were provided for both models. Results: A total of 432 of 668 patients (65%) developed complications. One hundred seventy-seven patients (27%) were readmitted. Mean PLOS was 10.2 days (7.1 SD) and mean CLOS was 12.3 days (10.1 SD). PLOS and CLOS both were correlated linearly to MAGS grade. Spearman correlation coefficient for CLOS vs. MAGS of 0.68 was higher than that of 0.49 for PLOS vs. MAGS. Multilevel AUC from MLRM using PLOS was 0.66, but multilevel AUC from MLRM using CLOS was 0.71. Discussion: CLOS provides an accurate estimate of hospital day utilization per patient for PD, reflecting not only the basal hospital recovery time for PD but the added time needed because of readmissions due to complications. It is tightly correlated to number and severity of postoperative complications.