News & Commentary

EMR Data in Maine's HIE Mined to Predict Care Costs


Researchers were able to accurately predict health care utilization in Maine for more than a million patients for six months using only demographic and electronic medical records (EMRs) data from the previous year. (Claims data were not included, but more on that later.) As such, the researchers made progress toward two major goals of health care: making EMRs fulfill their promise and changing the delivery system from volume-based to value-based.

Make that three goals: The EMR data were extracted from the state’s health information exchange. The HIE effort is a federal program that started in 2010 with much fanfare but stumbled out of the gate thanks, in part, to wary physicians and unprepared state health care bureaucracies.

This study, published in the Journal of Medical Internet Research, ranks the Maine program as one of HIE’s successes and a sign perhaps that the nearly $600 million investment by HHS’s Office of the National Coordinator for Health Information Technology might still pay off.

It all comes back to value, though. As Managed Care’s August cover package reported, the new value-based equation looks like this:

Value = (Quality + Outcomes)/Cost

To get there from here, researchers who hailed from the United States and China, used clinical data from 2012 to create a predictive model for health care utilization during the subsequent six months.

The model worked for both individuals and populations of 1,000 patients.

“To our knowledge, this study is the first to predict future health care resource utilization using only EMR data at the patient level across an entire state,” the study says.

Researchers grouped the patients into three main risk levels: high, intermediate, and low. Average costs for patients in each level were $385.76 (low), $1,124.33 (intermediate), and $4,276.88 (high).

That finding demonstrates that the model can forecast which patients will cost how much over the next six months, say the authors, some of whom are cofounders of HBI Solutions Inc., a predictive analytics company. Such forecasts may be crucial as more providers enter into value- and risk-based contracts.

The model focused on chronic diseases, especially kidney disease, diabetes, and heart problems, which, as predicted, accounted for the highest utilization in the next six months.

“In other words,” the study states, “our model can give not only patient-oriented forecasts, but also disease-oriented forecasts of future resource utilization.”

Using only EMR data means not having to integrate claims data from multiple payers. In addition, as the authors note, the EMR data are timelier than claims data, which are typically 60 to 90 days old by the time a provider receives the information.

Not that claims data and the health insurance plans that generate them are going to sit on the sidelines. The authors checked their findings against claims data.

Also, they concluded that, “future studies will focus on integrating payer claims data with the HIE data to get a more accurate and timely prediction of projected future resource utilization.”