At his recent Predictive Analytics World (PAW) keynote address, Eric Siegel posed the question, “How do you know your predictive analytics aren’t BS” (for bad science). The clever play on acronyms implied that many predictive analytics serve up false positives or negatives against self-serving hypotheses. In an industry like retail for example, the consequences of bad science may result in reduced revenues, but is rarely dangerous. If a father fails to buy beer as predicted when sent to the store for diapers, it results in a less crowded fridge. But get it wrong in healthcare and someone could wind up dead.
Moving on, Dr. Martin Kohn’s healthcare keynote warned that predictive analytics is rarely tied to ‘impactability’, the likelihood an individual will respond positively to care management interventions. Resources needed to provide these interventions are scarce and costly meaning healthcare organizations need not only predict which patients need them, but if they’re likely to work at all! Like Siegel pointed out, the stakes are high!
Those of us working in data science know how easy it is to get caught in the weeds without considering impactability. I remember listening to a product manager present the new features of our company’s business intelligence tools when I first entered the analytics industry 19 years ago. I asked how these whiz bang new features would help users make better decisions and lead to IRACIS (Increased Revenue, Avoided Cost, Improved Service). The response was ‘great question’, but there was no answer.
Perhaps part of the answer lies less with more data science than with the intersection between data science, people, and process. In healthcare, with the dizzying pace of medical research and the importance of personalizing care, optimizing impactability is paramount. Our ability to positively impact health outcomes requires human intervention and attention. The technology is simply an enabler that supports clinicians and their patients in managing care. The solution is precision health analytics, a solution that helps care teams predict and prevent disease rather than simply treat it after the fact. This helps HCO’s improve health and reduce costs at the same time.
As an example, our client, St. Joseph Healthcare spoke to this during their PAW talk, “Real-Time Clinical Driven Predictive Analytics for Care Management”. Dr. William Wood, VP Medical Affairs and Jessica Taylor, RN, Director Care Management, described the three-legged stool required for effective and efficient care management. The legs being a collaborative team of care managers across the care continuum, established care management workflows and processes, and predictive analytics powered by real-time clinical data. And so while we can now more easily predict the patients most at risk, we must also have the infrastructure needed in place to act. And furthermore, we must begin to better understand the potential impact of those actions.
Jessica described how HBI’s precision health analytics tool enables her to identify in real-time the right patients, at the right time, with the right interventions. Leveraging data science and HBI’s Spotlight Data Solution helped her and her team focus on impactablity. They knew which patients would benefit from interventions based on science and could then apply their knowledge and experience to personalize patient care. Everyone sensed the enthusiasm and confidence Jessica projected as she described St. Joseph’s significant results in reducing ED visits and readmissions and improving health.
Video: Learn more about how St. Joseph Healthcare, a community health system in Bangor, Maine has integrated our tools into their daily workflow.