Development and validation of an electronic health records-based opioid use disorder algorithm by expert clinical adjudication

Abstract

Background: In the US, over 200 lives are lost from opioid overdoses each day. Accurate and prompt diagnosis of opioid use disorders (OUD) may contribute substantially to prevention of overdose deaths. However, OUD research is limited, the specificity and sensitivity of OUD ICD codes are unknown, and the ICD codes are known to underestimate OUD prevalence. We developed and validated algorithms to identify OUD from EHR data and examine validity of ICD-based definitions for OUD. Methods: Through multiple iterations, we developed EHR-based algorithms to identify OUD. These algorithms and ICD-based OUD definition were validated against a total of 169 independent gold standard EHR chart reviews conducted by an expert adjudication panel of eight pain and addiction medicine clinical experts across four large healthcare systems. The experts relied on clinical judgement and current Diagnostic and Statistical Manual of Mental Disorders-5 criteria for making OUD diagnoses. Results: Of the 169 EHR charts, 81 (48%) were reviewed by more than one expert and exhibited 85% agreement between the reviewing experts. The OUD ICD codes alone had 10% sensitivity and 99% specificity, underscoring the strong potential for OUD underestimation in studies depending on ICD codes alone. In comparison, after four iterations, the algorithms identified OUD with a 23% sensitivity and 98% specificity. Conclusions and Relevance: This is the first study to evaluate the validity of OUD ICD codes and develop validated EHR-based algorithms to address OUD underestimation. This work has the potential to inform future research on early intervention and prevention of OUD.

Association of Opioid Dose Reduction With Opioid Overdose and Opioid Use Disorder Among Patients Receiving High-Dose, Long-term Opioid Therapy in North Carolina

“…Second, we used ICD-9-CM and ICD-10-CM codes to identify OUD diagnoses, which have low sensitivity and typically underestimate OUD prevalence. 48 However, this outcome misclassification is likely nondifferential, thereby biasing results toward the null. Similarly, there could be nondifferential underascertainment of nonfatal opioid overdoses in claims data, especially with use of naloxone in the community.…”

“…However, our use of weighted marginal structural models is an important advance in controlling time-varying confounding without blocking causal mediation pathways, avoiding bias incurred by standard regression models used in prior studies. Second, we used ICD-9-CM and ICD-10-CM codes to identify OUD diagnoses, which have low sensitivity and typically underestimate OUD prevalence . However, this outcome misclassification is likely nondifferential, thereby biasing results toward the null.…”