Models Getting a Little Long in the Tooth

Much of what we know–or think we know–about clinical effectiveness is derived from large-scale data aggregation and its subsequent analysis. But the traditional data models used for aggregation–taking massive sets of claims data and throwing in some lab data–just isn’t cutting it anymore, according to David Stumpf, M.D. "That produces large and powerful data sets, but they’re missing critical information," Stumpf said during a presentation at Payer Symposium held at the HIMSS 2010 Conference & Exhibition in Atlanta. "Using just claims data is like trying to figure out how good your meal was by looking at your credit card bill."


Much of what we know--or think we know--about clinical effectiveness is derived from large-scale data aggregation and its subsequent analysis. But the traditional data models used for aggregation--taking massive sets of claims data and throwing in some lab data--just isn't cutting it anymore, according to David Stumpf, M.D. "That produces large and powerful data sets, but they're missing critical information," Stumpf said during a presentation at Payer Symposium held at the HIMSS 2010 Conference & Exhibition in Atlanta. "Using just claims data is like trying to figure out how good your meal was by looking at your credit card bill."

Stumpf is one the leaders of DARTNet, a federated network of electronic health records data from multiple provider organizations representing thousands of clinicians and practices and approximately 3.5 million patients. The network compiles data from 14 different EHR systems used by its member organizations. DARTNet (www.dartnet.info/) is seeking new member practices and organizations to expand the network. Stumpf is senior vice president, clinical data strategy, for the enterprise services group at UnitedHealth Group.

The goal of DARTNET, funded by the Agency for Healthcare Quality and Research and other sources, is to break new ground in comparative effectiveness. Stumpf discussed the progress made by DARTNet and its plans to take research "to the next level" by combining EHR data with claims data to get a complete picture of what's being done to patients, and the true clinical and financial impact of those treatments.

Data from each member practice's EHR is captured, de-identified, coded, standardized and stored in a clinical data repository which resides at each individual practice. The repository also connects to other data sources such as billing, lab, hospital, and prescription databases.

Data from the repository is transferred to a second database called the Gateway, which makes de-identified data from DARTNet practices available through a secure Web portal, to be used for research studies and quality improvement activities. But the provider members themselves have to "flip a switch" to send their data past their own firewalls into the aggregated database, Stumpf said.  "That was by design--we could have created a network that automatically pulled data, but we didn't think we'd have too many people sign up for that."

The Gateway makes de-identified data from DARTNet practices available via a secure Web portal, to be used for research studies and quality improvement activities. DARTNet has used that data to create decision support tools for member organizations to use at the point of care.

Creating an infrastructure to create a standardized set of information from myriad EHR products has been an eye-opening experience, Stumpf said. "To give an example of what we face, in a five-physician practice we work with, we found 136 variations in the location and entry method of cancer screening data. We have a long way to go before we can pull standard information from this data milieu."

The idea of adding payers into the DARTNet mix has physician members "extremely concerned," Stumpf said. But the network is exploring different data sharing models to alleviate those concerns and make sure there's some quid pro quo for both groups. "The truth is that I have a data set you don't have, and vice versa. But we have to find a way to make this works for both sides because without both data sets we'll never drive the value up of comparative effectiveness research."

--Greg Gillespie