Trajectory Analysis of Health Care Utilization Before and After Major Surgery

Abstract

Objective: To characterize patterns of health care utilization before and after surgery and determine any association with preoperative frailty.

Background: Frail patients experience worse postoperative outcomes and increased costs during the surgical encounter. Evidence is comparatively lacking for the longer-term effects of frailty on postoperative health care utilization.

Methods: Retrospective, longitudinal cohort analysis of adult patients undergoing any elective surgical procedure after preoperative frailty assessment with the Risk Analysis Index from February 2016 to December 2020 at a large integrated health care delivery and financing system. Group-based trajectory modeling of claims data estimated distinct clusters of patients with discrete utilization trajectories. Multivariable regression predicted membership in trajectories of interest using preoperative characteristics, including frailty.

Results: Among 29,067 surgical encounters, 4 distinct utilization trajectories emerged in longitudinal data from the 12 months before and after surgery. All cases exhibited a surge in utilization during the surgical month, after which most patients returned to "low" [25,473 (87.6%)], "medium" [1403 (4.8%)], or "high" [528 (1.8%)] baseline utilization states established before surgery. The fourth trajectory identified 1663 (5.7%) cases where surgery occasioned a transition from "low" utilization before surgery to "high" utilization afterward. Risk Analysis Index score alone did not effectively predict membership in this transition group, but a multivariable model with other preoperative variables was effective ( c = 0.859, max rescaled R2 = 0.264).

Conclusions: Surgery occasions the transition from low to high health care utilization for a substantial subgroup of surgical patients. Multivariable modeling may effectively discriminate this utilization trajectory, suggesting an opportunity to tailor care processes for these patients.