A new article in the Harvard Business Review examines predictive analytics in healthcare and how big data is making an impact.
Electronic health records gather massive amounts of data for many facets of care. However, unless the data is applied appropriately, it won't improve patient care. Here are five key thoughts from the article on optimizing predictive analytics:
1. Be specific about which clinical decisions data will inform when developing an algorithm. Just finding the readmission rate for a procedure isn't helpful; couple the predictive factors in readmission rates with patient education and follow-up support for at-risk populations.
2. Gather as much data as possible for greater accuracy when developing algorithms. The larger data sample will include more potential clinical applications and allow providers to become more specific with patient treatment. In the example discussed, VHA invested in an integrated EHR and data repository that equaled 5 percent of its total health spending, but experienced a return-on-investment more than $3 billion through improved quality.
3. More specifically targeted data can help reduce unnecessary interventions and waste. When uncertainty exists, clinicians will treat more patients with preventative measures. However, if data narrows down the truly at-risk patients, fewer will be exposed to unnecessary procedures. For example, 11 percent of newborns receive antibiotics while less than 0.05 percent have confirmed infections. Kaiser Permanente of Northern California updated protocol to become more specific and found 250,000 annually didn't receive unnecessary antibiotics that would have under the old protocol.
4. Don't overload clinicians with unimportant data; they might start to ignore it. The data that doesn't fit into the clinical workflow doesn't need to be in front of them. The decision-support tool AWARE can help practicing physicians spot the right data for their patients.
5. The sophisticated technology in predictive algorithms won't make a difference if health systems don't apply the algorithms correctly to improve value. Effective application can help health systems reduce spending and improve outcomes because physicians will provide targeted interventions to the patients "who need them most."
"In this next era of value-based care, health systems must critically think about the clinical situations where enhanced analytics can be useful, help providers use them routinely in patient care and develop strategies to evaluate the clinical impact of algorithms," wrote the study authors Ravi B. Parikh, MD; Said Obermeyer, MD; and David Westfall Bates, MD.