Background: Many clinical decision support (CDS) tools that leverage big data analytics attain reasonable predictive power yet are underutilized in practice. Physician user experience (UX) testing is a crucial but little-studied element of CDS success. We explored ED provider UX of behavioral health predictive CDS.
Methods: Semi-structured interviews, based on best practices for cognitive interviewing and UX testing, were conducted to thematic saturation with a purposive sample of EM attendings and residents at urban academic EDs in a single state. After asking about predictive analytics in general, interface design ergonomics were assessed using interactive CDS prototypes predicting hypothetical patients’ risk of opioid overdose and ED recidivism. Interviews were audio-recorded and transcribed verbatim. Transcripts were coded separately by two team members using an iteratively refined coding scheme until concordance was reached. Codes were entered into NVivo 12 software for pattern examination and interpreted using framework and thematic analysis techniques.
Results: Eleven ED physicians diverse in gender, experience and age group (ranging from residents to departmental leadership) were enrolled. Most were familiar with or had personally interacted with CDS technology. Concerns about predictive behavioral health CDS included trustworthiness, loss of physician agency, legal liability, devaluing the "art of medicine" and reimbursement. Potential barriers to use of behavioral health CDS included ED physician time limitations, lack of effective treatment options and difficulty facilitating linkages to outpatient services. Participants expressed discordant data visualization preferences in terms of notification design (e.g. location, color, content). Nearly all providers perceived CDS as potentially helpful for opioid overdose prevention; roughly half perceived utility for reducing ED recidivism. For both conditions, alerts that automated documentation, synthesized established risk factors and contained actionable recommendations were collectively preferred. Preferences did not vary according to age and experience level.
Conclusions: To increase utility and acceptability, behavioral health predictive CDS should incorporate common UX concerns and preferences. Future research will test real-time use of these programs.