Back

Design, EHR Integration and Evaluation of Clinical Decision Support Workflows Driven by a Mortality Prediction Model to Promote Goal Concordant Care

Roberts, L. M.; Rossi, L. A.; Zachariah, F. J.

2023-02-15 health informatics
10.1101/2023.02.14.23285926 medRxiv
Show abstract

We present a clinical decision support (CDS) framework to promote goal-concordant care for cancer patients nearing end of life, enabled by a 90-day mortality prediction model. Design, workflow, and deployment for four model driven CDS tools are described alongside utilization patterns and detailed performance analysis of the first of such tools integrated into clinical practice: an alert promoting advance directive completion. The alert achieved a precision of 44.1% (95% CI: 39.6 - 48.8%) and a recall of 34% (95% CI: 30.3 - 37.4%) with 9.8% 90-day mortality prevalence over a period of 8 months. Our analysis shows that both precision and recall of the alert were significantly impacted by the underlying clinical workflows. Based on feedback, clinicians have accepted model-driven CDS as a legitimate means to prioritize patients for advance care planning interventions.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.