College of Health and Human Services
George Mason University
George Mason University Mason
George Mason University

Using Big Data To Estimate Prognosis Of Patients With Multiple Diseases

October 14, 2016

This figure compares the performance of the MM Index to the Charlson and Elixhauser variants.

Prognostic data is important to health care providers as it can be used by patients and clinicians to set treatment priorities, evaluate the effectiveness of various treatment options, and be used to plan for end of life decisions, if necessary. The challenge is gaining an accurate prognosis that is tailored to each individual patient and their medical history.

In a new study, Farrokh Alemi, professor of health administration and policy, along with Cari Levy of Denver Veterans Affairs and Raya E. Kheirbek of the Washington DC VA Medical Center, provide an overview of the Multi-Morbidity (MM) Index, review the accuracy of already published data, and develop procedures for implementing the MM Index in electronic health record systems. The study is published in eGEMs.

The MM Index, developed by Alemi and colleagues, can be used to predict the prognosis of patients using their diagnostic history.

“Traditionally 30 to 160 diagnostic categories were used to predict prognosis. The MM Index scores each diagnosis separately, a model with more than 5,000 predictors. It is not surprising that it is more accurate than existing approaches,” Alemi said. “More predictors mean more accuracy even in cross-validation in different data sets.”

The authors compared the accuracy of the MM Index to physiological markers and other diagnosis-based indices, such as the Charlson and Elixhauser indices. The MM Index analysis was examined on massive data from the Veterans Affairs data warehouse and from the Healthcare Cost and Utilization Project of the Agency for Health Care Research and Quality. They concluded that the MM Index was several-fold more accurate than existing physiological or diagnostic-based approaches.

“The MM Index described in the paper can be used to conduct more accurate policy analysis and program evaluation, taking into account patient's history of illness in comparing treated and untreated groups,” Alemi said. “This is impactful as it is one of the first times in which ‘big data’ has moved the needle on what can be done in predicting prognosis.”

The study was supported by appropriation #3620160 from the VA Office of Geriatrics and Extended Care to Cari Levy at Denver Veterans Affairs, in addition to resources from the District of Columbia Veterans Affairs Medical Center.