Opioid use trajectories after thoracic surgery among veterans in the United States.

ABSTRACT

BACKGROUND: Opioid use has increased to epidemic levels over the past decade within the United States, particularly among vulnerable populations. This retrospective study aimed to evaluate rates of prolonged opioid use in the Veteran population after thoracic surgery and identify specific risk clusters. METHODS: Veterans Administration data on patients who underwent thoracic surgery between January 1, 2006 and September 30, 2015 included preoperative opioid use information for stratification of patients to preoperative chronic opioid use (PCOU; nPCOU  = 16,612) versus patients without preoperative chronic opioid use (WPCOU; nWPCOU  = 2,328). A Poisson regression model and prior literature were used to identify variables for use in a Latent Class Analysis (LCA) model for each stratum. Three-cluster models were selected, and identified as 'low-', 'intermediate-' and 'high-' risk groups. RESULTS: Cluster interpretations included (a) Low risk no psychiatric diagnoses, preoperative medication use, or preoperative chronic pain, (b) Intermediate risk no psychiatric diagnoses, but had preoperative medication use and some preoperative chronic pain and (c) High risk psychiatric diagnoses, preoperative medication use and preoperative chronic pain. For the PCOU stratum, rates of prolonged opioid use 1 year after surgery were as follows 46.3%, 61.9% and 66.0%. For the WPCOU stratum, the observed rates were 4.7%, 8.3% and 9.2%. CONCLUSIONS: Prolonged opioid use trajectories obviously differ by PCOU status, as well as preoperative psychosocial diagnoses, medication use and chronic pain. This is a first step in population-level research to curb the rate of prolonged opioid use in Veterans following thoracic surgery. SIGNIFICANCE: This article presents population-level chronic opioid use trajectories after thoracic surgery, using latent class structures. Demographics, preoperative psychological diagnoses, medication usage and chronic pain variables were utilized to identify population-level clusters. The cluster identified as highest risk had preoperative chronic opioid use, psychological diagnoses, other medication prescriptions and chronic pain.