Healthcare Analytics Sought For Population Health Management
Hospitals have keen interest in analytics for population health management, but tech is still in early stages, according to new survey.
The lack of good structured data in electronic health records came up in both of the two panel discussions in a webinar convened to respond to the survey results.
Frank Clark, VP for information technology and CIO at the Medical University of South Carolina said that most of the electronic data that MUSC has collected for the past 20 years is unstructured. He expressed hope that a natural-language processing (NLP) vendor that his organization is working with can extract the key data elements from that free text so his organization can start to identify patients who are at risk of becoming sick.
Curt Sellke, VP of analytics for the Indiana Health Information Exchange (IHIE), which connects over 90 of the 120 hospitals in Indiana, said IHIE's clinical data repository contains plenty of transcribed reports. He noted that IHIE's academic partner, the Regenstrief Institute, is working on NLP, but added, "We're at the starting point on that."
In the second panel discussion, Yvonne Hughes, CEO of Coastal Carolinas Health Alliance, another HIE, said that accountable care organizations (ACOs) depend mainly on claims data for a broad picture of patient care. It would be better if they could use a community record that combined clinical data from many EHRs along with claims and other data, she said.
Jonathan Niloff, MD, chief medical officer of MedVentive, a unit of McKesson, suggested that setting such a goal might be counterproductive because EHRs are a flawed source of data and not interconnected. For retrospective analysis of population health management, he said, claims data from payers is the best source. But for functions such as risk stratification, decision support and care management, near-real-time data is needed. A combination of billing data, lab data and Surescripts prescription information, although not perfect, could do the trick, he said.
Pamela McNutt, senior VP and CIO of Methodist Health System in Dallas, expressed dissatisfaction with claims-based analytic tools developed by payers. Eventually, she said, analytics for tracking patients and sending out reminders for care management will be combined with cost and resource utilization tools on a single platform. But it's not there yet.
"Most ACOs are in an immature state," she said. "We don't even know the questions we need to be asking." Later, asked how she would define success in the analytics area, she said, "For those of us who are taking on risk, success is still about providing higher-quality care at lower cost. Products that can show ROI on those two fronts are what everyone wants."
Hughes said success will come from supplying providers with "clean, normalized data" from disparate systems so they can make decisions that will improve population health.
Niloff agreed with both of the other panelists. However, he added, good data must be imbedded in workflows and leveraged through automation to drive improvements in cost and quality.