The term “big data” has been tossed around for years, but what does it actually mean in clinical practice? Amy Compton-Phillips admits that the term has never resonated with her, but that Scott Weingarten’s talk about decision support did: “It’s not the data, it’s the knowledge and then transferring that knowledge into decision support.” She asks Weingarten, the Chief Clinical Transformation Officer at Cedars-Sinai Health System, to share examples of how Cedars-Sinai has leveraged big data to offer decision support that, in turn, has transformed clinical practice.
Cedars-Sinai provides evidence-based, or data-based, information to clinicians at the point of care, answers Weingarten. For example, as Dr. Brown sees Mrs. Jones, who has asthma and depression, the doctor receives a recommendation on what the best course of treatment for Mrs. Jones might be, as the evidence permits. The provider is free to accept or reject that recommendation.
“We’ve been doing this for a number of years now, and we find that if you provide valid information, evidence-based information that’s also consistent with the patient preferences, patient values, and the provider agrees with those recommendations, you find a significant improvement in care,” says Weingarten. Cedars-Sinai has about 800 million clinical decision support instances where they’ve tracked how often providers follow the recommendations, or if they rejected the guidance and why — providers can select a standard reason or write free text that is analyzed via natural language processing. “It gives you the opportunity, through constant feedback in an iterative process, to make the recommendations better and better each time to get improved care,” he says.
But do clinicians actually want all of that information? Compton-Phillips refers to the example of airplane pilots so overwhelmed by information fed into the cockpit that they banded together and said, “Stop the madness. Let’s have a threshold.”
What Cedars-Sinai’s decision support does is a trade-off, says Weingarten. They monitor how often providers follow the recommendations, and they might find some recommendations followed only 1/100 times, or 1/1000. “We believe that’s clutter, and we need to almost ceremoniously remove the clutter, the non–value-added information, and get it out, and leave room for the information that will help patient care. We need to respect the clinician’s time,” he says. The second thing is to follow up with providers on what the benefits are to patient care. “If we’re going through the process of alerting providers in real time, we need to show evidence that this has helped patient care — if not, we shouldn’t be doing it.”
From the NEJM Catalyst event The Future of Care Delivery: Relentless Redesign at Providence St. Joseph Health, January 19, 2017.