EHR Data Predicts Diabetes, CV Disease

Recently, a survey of health care leaders showed they were bullish about the idea that data contained in electronic health records (EHRs) will eventually prove useful in improving quality and reducing the costs of care.

Soon thereafter, Boston-based scientists proved the survey respondents to be prescient by releasing a report which showed that EHR data could identify people at risk for domestic abuse 2 years before physicians actually established the diagnosis.

Now, another Boston-based group has provided additional support for the prediction. In this new study, scientists showed that EHR data could identify patients who were at risk for developing diabetes and coronary heart disease, and who were thus candidates for early intervention programs designed to prevent or delay the onset of both conditions.

Marie-France Hivert and colleagues from Massachusetts General Hospital examined EHR data for 122,715 patients that was derived from 12 primary care practices which used the same EHR. All clinical and utilization data were searchable using a specially-designed "Research Patient Data Repository" that is proprietary to the Mass General, its parent company, Partners Healthcare and its clinical researchers.

The scientists excluded patients with established diabetes or coronary heart disease, and then developed and validated a predictive tool based on 5 criteria that are normally used to diagnose metabolic syndrome, a precursor to diabetes. They relied on data that had been entered into the EHR during 2003-2004.

The criteria were central obesity, high triglycerides, low HDL (good) cholesterol, impaired glucose tolerance, and elevated blood pressure.

The scientists used some clever work-arounds to deal with missing data, and stratified the remaining patients into 3 groups: a low risk group that had none of the 5 criteria, a moderate risk group that had 1 or 2 of them in any combination, and a high risk group that had any or 3 or more of the criteria.

The scientists then used age- and sex-adjusted regression models to compare outcomes in the 3 groups during the ensuing 3-year period (2005-2007).

It turned out that only 1.5% of the low risk patients developed diabetes during follow-up. For moderate and high risk patients, the incidence of diabetes rose to 4.0% and 11.0% respectively, a highly significant trend.

Similarly, the incidence of coronary heart disease rose from 3.2% to 5.3%, to 6.4% respectively for the low, moderate and high risk groups.

The EHR-based model also predicted resource utilization and overall health care costs. For example, the average number of admissions per patient rose from 0.10 to 0.23 and 0.36 in the low, moderate and high risk groups, respectively. The average annual number of outpatient clinic visits rose from 4.3 to 7.2 to 11.0, and the total annual health care costs per patient rose from $3,804 to $4,665 to $5,271.

The scientists concluded that risk factor clustering of EHR data can be used to identify primary care patients at risk for developing diabetes and coronary heart disease and consequently, higher resource utilization.

They were particularly pleased that they could develop a predictive model despite modest problems with missing data that are endemic to data bases created by practitioners during the course of routine patient care.

They were optimistic that lifestyle interventions typically used for patients with classic metabolic syndrome would work equally well in the populations they identified.

This study succeeded because the 12 participating PCP practices used EHRs that automatically contributed data to a single, albeit proprietary data repository. Since all patient data entered by Practice Fusion's EHR users is stored centrally, Practice Fusion can support similar research projects with ease.

And Practice Fusion data bases contain information on patients in all 50 states, not just those finicky New Englanders. Practice Fusion welcomes the opportunity to collaborate with clinical investigators as they seek new breakthroughs in the quality of care.

Glenn Laffel MD, PhD
Sr. Vice President, Clinical Affairs, Practice Fusion