When patient data isn’t shared, predictive analytics suffer

Data silos that prevent the sharing of patients’ medical information between organizations may hamper analytics efforts and short-circuit patient care.

Physicians have a vast array of data in their electronic health record systems to help them understand past treatments of a patient. However, what they often don’t have is information from other providers that have treated the patient, and they don’t know if the patient is following the prescribed treatments of these providers.

Lacking a complete picture of the patient’s overall treatment and health status increases the risk that a physician may be unaware that a patient is headed toward a major adverse healthcare event.

The data silos within the healthcare industry today make it tough to achieve personalized predictive analytics, says Yan Li, assistant professor at Claremont Graduate University‘s Center for Information Systems and Technology in California.

Patients seeing specialists may need to ask their other clinicians to send their records to the specialists, but the doctors may not electronically exchange data. That can create another data silo, because a primary care physician may know the details of a patient’s medical needs, but isn’t getting specialist data that could inform future treatment for the patient.

Use of personalized predictive analytics teamed with machine learning systems, along with better physician data management could help a physician predict an adverse event before it occurs, Li contends.

Also See: How the cloud can break down silos within hospitals

For example, if a specialist prescribed medication, the specialist may not know how often the patient is picking up the medication, and the primary care physician likely won’t know either.

However, personalized predictive analytics can paint a better picture of a patient’s present status and needs, starting with behavioral health data, Li believes. This includes health habits, risk factors and genetic conditions.

She advocates designing an app that can go on a diabetic patient’s phone to chart when a patient eats and what they are eating, and an app to collect continuous glucose levels on a wristband.

Personalized analytics, however, don’t necessarily have to be limited to an individual, Li says. “We also need best practices to aggregate data for patients with common conditions.”