Objective outcome measures in randomized clinical trials of interventions to reduce risk of opioid overdose following discharge from the emergency department: Utility of administrative data linkage

Evaluation of the effectiveness of interventions for opioid use disorders (OUD) are vital to preventing opioid-related deaths.¹ While patient-reported outcomes report health statuses without interpretation from the clinicians or others,² objective outcome assessments in emergency department (ED)-based clinical trials for OUDs address issues of loss to follow-up (LTFU) and reporting bias. Clinical trials and observational cohorts of OUD treatment often fail to achieve optimal retention, with attrition reaching 40%.³ Systematic attrition of particular subgroups may result in spurious associations between intervention and outcome.⁴ As opioid overdose ED visits have increased,⁵ novel, ED-based, peer recovery interventions are being trialed.^6,7

The Navigator Trial is an IRB-approved, randomized clinical trial evaluating the effectiveness of peer recovery support specialists (i.e. “navigators”) compared to standard ED-based licensed clinical social workers for ED-discharged patients at risk of opioid overdose.⁷ Enrollment in a licensed substance use disorder treatment program within 30 days of randomization (first primary outcome) will be assessed using mandated statewide registries of controlled substance prescriptions and licensed behavioural healthcare organisations in Rhode Island. Access to administrative data from health data partners for the current study requires written informed consent from participants, confirmation of ethical oversight, and maintenance of confidentiality. A subsequent ED visit for overdose at any Rhode Island hospital within 18 months of randomization (second primary outcome) will utilize mandatory reporting data from the Rhode Island Department of Health (all EDs in RI),⁸ and electronic health records from the state’s largest health system. In aggregate, these sources should capture >95% of all overdoses presenting to RI EDs. If treatment is accessed or an ED overdose visit occurs outside of Rhode Island, these outcomes would not be captured in the datasets. While this is expected to be minimal, underascertainment of outcomes would also be expected to be non-differential with respect to the treatment group and would bias results toward the null.

In an ad-hoc interim analysis, telephone-based follow-up to capture participant self-reported outcome data (e.g., recent substance use and overdose) resulted in a LTFU rate of 83% LTFU (n=135 of 162). We were unable to detect an association between LTFU and demographic characteristics such as sex at birth, race/ethnicity, reason for ED visit, health insurance coverage, drug treatment history, housing and employment status, and plans to change drug use (Supplementary Table 1), or randomization arm. An association observed between LTFU and younger age [odds ratio per 10-year increase: 0.63 (CI: 0.44, 0.90)] may result in bias,as younger people have demonstrated higher rates of ED visits for opioid overdose and lower OUD treatment uptake.^5,9

Administrative datasets provide objectivity, reduced cost, ambispective data collection, and reduced attrition bias with minimal loss of data.¹⁰ As work progresses to evaluate the efficacy of ED-based OUD navigation interventions in the United States, methods to minimise selection bias will be paramount. Incorporating objective outcome measures can improve the validity of study outcomes in this vulnerable population.