Start Your Search

P. Tan

Author of

  • OS27 - Disrupting Technology into the Next Decade (ID 43)

    • Event: e-Health 2018 Virtual Meeting
    • Type: Oral Session
    • Track: Technical/Interoperability
    • Presentations: 1
    • OS27.06 - Improving Management of Long Stay Patients with Machine Learning Prediction (ID 41)

      P. Tan, Fraser Health Authority; Surrey/CA

      • Abstract
      • Slides

      Purpose/Objectives: A grand challenge that hospitals face is managing facilities and manpower under the uncertainty of how long an admitted patient will stay. This uncertainty can restrict efficient utilization of hospital resources by prohibiting effective scheduling and coordination of early discharge preparations. Long stay patients (?30 days) are particularly impacted, where prolonged hospitalization is frequently associated with significant social, economic, physical and psychological burden. At Fraser Health, the year-to-date average number of long stay patients is 17% higher than the fiscal year-end target. Early identification of long stay patients and effective care planning to reduce long stay can lead to better health outcomes, greater patient and provider satisfaction, reduced risk of adverse incidents and complications, and contribute to overall healthcare system sustainability.

      Methodology/Approach: Health and Business Analytics (HBA) has designed a predictive model called long stay predictor (LSP) that can predict whether an admitted patient will be a long stay patient or a short stay patient with 75% accuracy. Through artificial intelligence trained on over 10,000 historical long stay cases, the model is capable of identifying new long stay patients within 48 hours of admission. Patients flagged by the predictor are notified through daily email, upon which an early escalation of discharge planning is triggered to identify discharge barriers and establish a monitoring process on the patient. Staff can follow-up on individual patients real-time through an in-house web tool. Figure 1. Model’s inputs & outputs Figure 2. Workflow

      Finding/Results: At present, the tool has been deployed across White Rock/South-Surrey, Delta, Burnaby and Chilliwack. A 5-month post-implementation analysis at Peace Arch Hospital reveals a 16% reduction in long stay patients, and 25% reduction in bed days. Operational lessons learned from this project, along with novel insights that reveal why some patients are long stay, will be shared at the conference.

      Conclusion/Implications/Recommendations: Fraser Health has created a tool to predict patients likely to become long stay, and have operationalized this information to help clinicians engage in early care and discharge planning, leading to a significant reduction in bed days. Next steps include expanding adoption across all Fraser Health sites and adapting model to predict readmissions.

      140 Character Summary: Predictive modelling provides timely alerts to improve care planning for long stay patients at Fraser Health, contributing to a 25% reduction in bed days.

      Only Active Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login or select "Add to Cart" and proceed to checkout.