Algorithm looks to past to predict future health conditions

Tools

With predictive modeling becoming de rigueur in healthcare, this was bound to happen. Just as Netflix or Amazon can predict the movies or books you'd like based on your past purchases, a new algorithm can predict future ailments based on the patient's history--along with data from others who have similar conditions, according to a University of Washington announcement.

The statistical model, to be highlighted in an upcoming issue of the journal Annals of Applied Statistics, shares information across patients who have similar health problems. Tyler McCormick, assistant professor of statistics and sociology at the University of Washington, says the algorithm is designed to deal with a paucity of data. It improves the prediction as opposed to when "a few people have a few things wrong with them," as Popular Science put it.

"This provides physicians with insights on what might be coming next for a patient, based on experiences of other patients. It also gives a predication that is interpretable by patients," McCormick said.

He and colleagues from MIT and Columbia University drew data from a multiyear clinical drug trial involving tens of thousands of patients aged 40 and older. Other demographic data, such as gender and ethnicity, as well as patients' histories of medical complaints and prescription medications were included.

Of 1,800 medical conditions included in the data set, 1,400 of them occurred fewer than 10 times--hence the need to statistically model those conditions. They used a technique grounded in Bayesian methods, called the Hierarchical Association Rule Model (HARM) that creates rules, such as if a patient has already had dyspepsia and epigastric pain, for instance, heartburn might be next, McCormick writes.

Healthcare organizations are embracing predictive modeling in a big way, though as McCormick points out, it's generally not yet used at the patient level. The Department of Health and Human Services is using mathematical modeling to analyze effects of specific healthcare interventions, as FiercehealthIT has reported.  And hospitals are using it to reduce readmissions, according to FierceHealthcare.

McCormick wants to make the model available to doctors and patients. "We hope that this model will provide a more patient-centered approach to medical care and to improve patient experiences," he said.

To learn more:
- read the University of Washington announcement
- here's McCormick's paper
- check out the Popular Science article

Related Articles:
HHS moves to mathematical modeling for research, intervention evaluation
Hospitals cut readmissions with predictive modeling
Algorithm compiles misleading readmissions information
New app helps docs predict post-surgical pain
Decision support, predictive modeling may speed clinical research