Skip to content Skip to navigation

Health systems use data to predict behavioral health risks

September 26, 2016
by Julie Miller, Editor in Chief
| Reprints

As the healthcare system shifts payment models to favor comprehensive care, more stakeholders are looking for new tools that offer strategic advantages. One such tool is large-scale data analysis. 

Data analysis, for example, might help to predict a patient’s risk of depression in advance so clinicians can address it early and possibly offer referrals to behavioral health providers long before the condition becomes severe. According to Karen Handmaker, vice president of population health strategies for IBM, such predictive modeling is valuable to medical systems and primary care providers that are increasingly being reimbursed based on comprehensive care.

“For managing populations of patients and individuals, we need data in real time or near real time—not just claims that tend to lag or clinical data that may not get updated between patient visits,” Handmaker says. “To have a 360-degree view of every patient, we need to integrate across the care spectrum, which includes behavioral health, substance use disorders and physical health. The challenge is that the infrastructure to do this heretofore has not existed.”

She says health systems today are beginning to use the IBM Watson Health data processor—the same supercomputer that debuted on TV game show Jeopardy! in 2011—to churn out practical data clinicians can act on. Ideally, patients with unrecognized, untreated behavioral health conditions or those predicted to be at risk will be identified earlier and referred to treatment.

“The big thing the technology will do is tee up those providers in the networks responsible for whole person care,” Handmaker says. “The incentives are now changing to support this level of communication and integration.”

Behavioral inputs

What makes Watson helpful for behavioral health in particular is its ability to process natural language, such as clinical progress notes or inputs that are more nuanced than just numerical test results. It also has the ability to “learn” over time, tweaking its answers as more material chugs through its brain.

“It’s only going to prove its worth the more precise it can get,” Handmaker says.

Insurance giant Anthem and Memorial Sloan-Kettering Cancer Center were the first healthcare outfits to start testing Watson’s data skills in 2012 to examine clinical pathways to treat cancer. That was followed by the official commercial launch of the healthcare vertical in 2015. Handmaker says today IBM is looking for more sources of data inputs for Watson, such as information contained within behavioral health records.