An electronic health record metadata-mining approach to identifying patient-level interprofessional clinician teams in the intensive care unit.

Abstract

Advances in health informatics rapidly expanded use of big-data analytics and electronic health records (EHR) by clinical researchers seeking to optimize interprofessional ICU team care. This study developed and validated a program for extracting interprofessional teams assigned to each patient each shift from EHR event logs. A retrospective analysis of EHR event logs for mechanically-ventilated patients 18 and older from 5 ICUs in an academic medical center during 1/1/2018-12/31/2019. We defined interprofessional teams as all medical providers (physicians, physician assistants, and nurse practitioners), registered nurses, and respiratory therapists assigned to each patient each shift. We created an EHR event logs-mining program that extracts clinicians who interact with each patient's medical record each shift. The algorithm was validated using the Message Understanding Conference (MUC-6) method against manual chart review of a random sample of 200 patient-shifts from each ICU by two independent reviewers. Our sample included 4559 ICU encounters and 72846 patient-shifts. Our program extracted 3288 medical providers, 2702 registered nurses, and 219 respiratory therapists linked to these encounters. Eighty-three percent of patient-shift teams included medical providers, 99.3% included registered nurses, and 74.1% included respiratory therapists; 63.4% of shift-level teams included clinicians from all three professions. The program demonstrated 95.9% precision, 96.2% recall, and high face validity. Our EHR event logs-mining program has high precision, recall, and validity for identifying patient-levelshift interprofessional teams in ICUs. Algorithmic and artificial intelligence approaches have a strong potential for informing research to optimize patient team assignments and improve ICU care and outcomes.