Evaluating markers for methamphetamine use in electronic health records.

Abstract

Methamphetamine is an addictive stimulant with serious health effects and its use has been increasing in the United States. Electronic health records (EHRs) capture timely information on substance use and are increasingly harnessed for public health surveillance and research. Identifying methamphetamine use in EHRs is challenging because existing metrics, such as diagnosis codes, are non-specific to methamphetamine, and clinicians record methamphetamine use in different ways. Little guidance exists about capturing methamphetamine use in EHRs. We identified eight markers of methamphetamine use in the EHR of an integrated, safety-net hospital system. We evaluated markers based on individual or combined likelihood of accurately capturing methamphetamine use and computational lift to extract from the EHR. We conducted chart reviews to calculate the predicted probability of identifying true methamphetamine use for each group of markers. We reviewed 814 charts across eight groups of methamphetamine markers. Positive predictive values (PPVs) of electronic definitions ranged from 0.21 to 1.00, with four groups obtaining PPVs > 0.70 compared to chart review. Groups of methamphetamine markers varied substantially in lift; the easiest was extracting non-specific diagnosis codes and the most challenging was capturing mentions of methamphetamine use in free-text clinical notes. Results provide multiple options to consider when extracting data on methamphetamine use from EHRs with guidance on computational lift and complexity of different markers. Users can select specific definitions of methamphetamine use depending on the project, technical capacity, time, and EHR system. More accurate definitions of methamphetamine use can inform interventions to curb its use.