The widespread use of EMRs in hospitals gives them unprecedented access to vast amounts of data. However, the application of "big data" techniques to predict patient outcomes has not materialized in most hospitals, according to the Harvard Business Review.
Raw data alone is not valuable when it comes to improving care delivery; rather, it is the insights derived from such data that give hospitals a leg up. Furthermore, prospective analysis — in addition to just retrospective analysis — empowers clinicians to act earlier and preemptively, ultimately improving patient outcomes.
To leverage the full benefits of patient data, hospitals must identify how to integrate predictive analytics into clinical practice. Consider the following steps from the Harvard Business Review.
1. Determine what clinical decision the algorithm will inform. Wherever there is data, there is a potential algorithm, so it is important to be specific about which clinical decision a particular algorithm will inform, according to the report. For instance, there are many algorithms that can be used to predict a patient's risk of being readmitted to the hospital. However, just knowing the percentage risk of readmission doesn't tell clinicians what signs to look out for. Instead, an algorithm should answer specific questions, such as, "Should this patient be discharged to a rehabilitation facility before going home?" or "Does this patient need a home visit in the next two days?"
2. Ensure the accuracy of the data. Algorithms will only produce reliable results if they are based on reliable data. According to the report, algorithms that use large pools of data to predict risk have higher accuracy and greater potential clinical applications.
3. Focus on high-impact areas. Clinicians often end up over-treating or under-treating patients when they are uncertain of a clinical decision, according to the report. Predictive analytics can help inform clinicians to steer high-cost interventions to the patients who are actually at the most risk of needing them. For instance, while less than 0.05 percent of newborns have infections confirmed by blood cultures, 11 percent of them receive antibiotics. To reduce over-treatment, researchers at Kaiser Permanente of Northern California devised an algorithm to accurately predict the risk of severe neonatal infection to better determine which babies need antibiotics, ultimately saving 250,000 newborns from receiving unnecessary antibiotics each year, according to the report.
4. Don't force analytics into the workflow. Physicians see hundreds of numbers each day. If data analytics is not integrated effectively into the workflow, it could be interpreted as just another number and ignored. For best results, the incorporation of analytics should be done gradually with an emphasis on education — not forced upon physicians.