Sorry, you need to enable JavaScript to visit this website.

Feedback

Your feedback is important to keep improving our website and offer you a more reliable experience.

Predictive clinical analytics with technologies from Intel, ProKarma, and Cloudera help Sharp Healthcare use electronic medical record data to identify patients at risk of needing an intervention from the rapid response team within the next hour. Brett MacLaren, Vice President of Enterprise Analytics at Sharp HealthCare noted, "With suprising accuracy, we found you really can predict when a patient is heding in the wrong direction, just from anlyzing the EMR data that is available in real time... There is a lot of potential for this type of technology to deliver a clinical and financial return on investment."

The Challenge: Sharp HealthCare utilizes specialized rapid response teams to respond to medical emergencies in the hospital. To maximize their effectiveness, the teams manually review charts to try and predict potential crisis and locate themselves near the patients most at risk to reach them the fastest should a medical emergency arrive. Sharp was interested in seeing if predictive analytic techniques could automate that analysis and identify patients most at risk of medical emergencies, allowing the hospital to proactively position themselves to deliver more effective interventions.

The Solution: Sharp teamed up with Intel and ProKarma, Inc. on an eight week proof of concept project at Sharp Healthcare in Sand Diego, California which analyzed data from the hospital's electronic medical record system to create and train a model that identifies patients most at risk of sudden decline within the hour. The team used a variety of features including blood pressure, temperature, and pulse rate to train the model on a Cloudera cluster powered by the Intel Xeon processor E5 v4 family.

The Results: The model was 80% accurate in predicting the likelihood of a medical emergency within the next hour when compared against historical data. This demonstrates the ability to use predictive analytics trained on previous electronic medical records to drive real-time clinical interventions that enhance resource allocation.

"Healthcare is starting to get beyond traditional transactional decision making, and move towards real-time, predictive, interventional decision making at the point of care," says Sharp's Brett MacLaren. "We're beginning to use analytics not just to understand what happened in the past and make operational decisions, but to predict the future and intervene in real time to influence the clinical outcome."

 

Find us on githb: https://github.com/bartleyintel/model-for-predicting-rapid-response-team-events

 

Get Involved!

 

To get started, follow the links below:

 

  1. Implementation guide: https://github.com/bartleyintel/model-for-predicting-rapid-response-team-events/blob/master/Model-for-predicting-rapid-response-events-implementation-guide-final.pdf
  2. Model: https://github.com/bartleyintel/model-for-predicting-rapid-response-team-events/blob/master/notebooks/modeling/modeling_diff_algorithms.ipynb
  3. Data science notebooks: https://github.com/bartleyintel/model-for-predicting-rapid-response-team-events/tree/master/notebooks
  4. Test data set