Overdose deaths from opioid misuse have reached an all-time high during the COVID-19 pandemic.Substance misuse or unhealthy substance use is a common problem identified in hospitalised patients associated with poor health outcomes, but it is not prioritised, and frequently unaddressed during routine care, because current approaches for screening with structured diagnostic interviews require additional staffing and effort during clinical care. Important details about substance use are captured in the clinical notes of the electronic health record (EHR) but the data are difficult to mine and analyse.
Authors trained a convolutional neural network to screen and identify alcohol misuse, opioid misuse, and non-opioid drug misuse with high accuracy using notes collected during clinical care.Substance misuse is a complex condition that frequently occurs as polysubstance misuse. The model offers the advantage of modelling different types of substance use jointly rather than using individual models.
The study has several limitations, Chicago has long encountered an opioid epidemic with heroin use in middle-aged non-Hispanic Black people, so the model was trained with a greater proportion of these individuals than encountered within other health systems. The rates of substance use in the study were low and might represent under-reporting. Moreover, authors show an adequate criterion and face validity of the classifier, but its effect on programme fidelity and health outcomes remains unknown.【MORE】