The Influence of Allowable Refill Gaps on Detecting Long-Term Opioid Therapy: An Analysis of Population-Based Administrative Dispensing Data Among Patients with Knee Arthritis Awaiting Total Knee Arthroplasty.

Abstract

It is challenging to detect long-term opioid therapy (LTOT) using administrative data, as refill gaps can disrupt opioid utilization episodes. Previous studies have used various methods to define LTOT and allowable refill gaps with little supporting evidence.To describe the effect of allowable refill gaps on detecting LTOT among a cohort of patients with arthritis awaiting total knee arthroplasty (TKA) using 3 different methods.A retrospective analysis of multicenter population-based data between January 1, 2012, and December 31, 2016, identified patients prescribed opioids before TKA in Alberta, Canada. We described 3 methods to detect LTOT based on a (1) fixed number of days between prescriptions; (2) fraction of the preceding prescription length; and (3) combination method that selected whichever refill gap was greatest. We then compared the number of patients classified as long-term opioid users by varying the number of days between prescriptions from 1-90 days (fixed method) or 0.04-3.2 times the duration (fraction method) for each method and refill gap.Of the 14,252 patients included in our cohort, 4,393 patients (31%) had an opioid prescription within 180 days before TKA. Detection of LTOT varied from 4.4% to 14.6% (fixed method), 4.2% to 13.2% (fraction method), and 4.5% to 15.1% (mixed method) as refill gaps varied from minimum to maximum. As refills gaps increased, the dose and duration of opioids in the utilization episode decreased for all 3 methods.The allowable refill gap between opioid prescriptions can influence the estimated rate of LTOT when using administrative pharmaceutical dispensing data. Definitional parameters should be carefully considered when using administrative data to define consistent opioid use.This work was supported by the Department of Surgery's Clinical Research Grant at the University of Alberta (RES0039945). The authors have no potential conflicts of interest.