Key Points Question Are medication monitoring programs within a hospital associated with more accurate identification of patients with opioid use disorder through the use of proxy Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition) ( DSM-5) criteria for opioid use disorder extracted from electronic health records? Findings This cross-sectional study demonstrated that DSM-5 criteria for opioid use disorder can be extracted through review of electronic health records and that patients who are part of a drug monitoring program had a higher mean prevalence of opiod use disorder and a higher mean number of psychiatric comorbidities associated with opioid use disorder. Meaning Proxy measures that rely on multiple sources of data, including prescription drug history and notes in the electronic health record, may help identify patients with opioid use disorder who have not received a diagnosis.

No abstract

BackgroundOpioid use disorder (OUD) is underdiagnosed in health system settings, limiting research on OUD using electronic health records (EHRs). Medical encounter notes can enrich structured EHR data with documented signs and symptoms of OUD and social risks and behaviors. To capture this information at scale, natural language processing (NLP) tools must be developed and evaluated. We developed and applied an annotation schema to deeply characterize OUD and related clinical, behavioral, and environmental factors, and automated the annotation schema using machine learning and deep learning-based approaches.MethodsUsing the MIMIC-III Critical Care Database, we queried hospital discharge summaries of patients with International Classification of Diseases (ICD-9) OUD diagnostic codes. We developed an annotation schema to characterize problematic opioid use, identify individuals with potential OUD, and provide psychosocial context. Two annotators reviewed discharge summaries from 100 patients. We randomly sampled patients with their associated annotated sentences and divided them into training (66 patients; 2,127 annotated sentences) and testing (29 patients; 1,149 annotated sentences) sets. We used the training set to generate features, employing three NLP algorithms/knowledge sources. We trained and tested prediction models for classification with a traditional machine learner (logistic regression) and deep learning approach (Autogluon based on ELECTRA's replaced token detection model). We applied a five-fold cross-validation approach to reduce bias in performance estimates.ResultsThe resulting annotation schema contained 32 classes. We achieved moderate inter-annotator agreement, with F1-scores across all classes increasing from 48 to 66%. Five classes had a sufficient number of annotations for automation; of these, we observed consistently high performance (F1-scores) across training and testing sets for drug screening (training: 91–96; testing: 91–94) and opioid type (training: 86–96; testing: 86–99). Performance dropped from training and to testing sets for other drug use (training: 52–65; testing: 40–48), pain management (training: 72–78; testing: 61–78) and psychiatric (training: 73–80; testing: 72). Autogluon achieved the highest performance.ConclusionThis pilot study demonstrated that rich information regarding problematic opioid use can be manually identified by annotators. However, more training samples and features would improve our ability to reliably identify less common classes from clinical text, including text from outpatient settings.

Background Prescription opioids (POs) are commonly used to treat moderate to severe chronic pain in the health system setting. Although they improve quality of life for many patients, more work is needed to identify both the clinical and genetic factors that put certain individuals at high risk for developing opioid use disorder (OUD) following use of POs for pain relief. With a greater understanding of important risk factors, physicians will be better able to identify patients at highest risk for developing OUD for whom non-opioid alternative therapies and treatments should be considered. Methods We are conducting a prospective observational study that aims to identify the clinical and genetic factors most stongly associated with OUD. The study design leverages an existing biobank that includes whole exome sequencing and array genotyping. The biobank is maintained within an integrated health system, allowing for the large-scale capture and integration of genetic and non-genetic data. Participants are enrolled into the health system biobank via informed consent and then into a second study that focuses on opioid medication use. Data capture includes validated self-report surveys measuring addiction severity, depression, anxiety, and nicotine use, as well as additional clinical, prescription, and brain imaging data extracted from electronic health records. Discussion We will harness this multimodal data capture to establish meaningful patient phenotypes in order to understand the genetic and non-genetic contributions to OUD.