Holistic Data Strengthens Predictive Analytics in Healthcare

From diabetes to depression, predictive analytics can help identify a course of action for many health conditions. This growing discipline already has some success stories, which has enabled more proactive outreach and early interventions. Even with the breadth of physical health data available, adding more data about social drivers of health could release the untapped potential of predictive analytics.
 

What is Predictive Analytics?

Predictive analytics in healthcare helps support more informed decisions. It uses artificial intelligence to offer insights to providers or health plans, so they can suggest targeted actions that will help a person improve their health. When using near real-time data, personalized actions can be recommended as soon as the data becomes available.

Common data sources that give context about a person include electronic health records (EHRs), lab results, disease-specific patient registries, genetic tests, and health plan claims.
 

Incorporating Social Drivers

Even with all these data sources, there is evidence that adding more data — about social drivers of health (SDoH) specifically — improves predictive models. In a review of studies predicting risk of cardiovascular disease, researchers found that including SDoH data increased the model’s accuracy. This is logical, given that 80% of health is driven by what happens outside the doctor’s office.

So, where does SDoH data come from and how can it be incorporated into predictive analytics on a large scale?

Many doctors and other healthcare providers ask patients about health-related social needs, but standardizing and sharing that data has proven difficult.

Doctors might enter this information in patient notes or record results from their own screening tools in an EHR. Often, this information is not handled consistently and in a way that is easy for AI to analyze.

One solution to better incorporating SDoH data in predictive analytics is to use natural language processing to transform information from freeform notes into usable data.

“Another is to expand and standardize the collection and consumption of social data at scale, including self-attested social screeners, EHR reported social data and other sources,” said Bobby Samuel, staff vice president for Carelon Digital Platforms, a division of Elevance Health. “All of this is to have better context of the consumer and their needs.”
 

Integrating More Data Sources

In addition to improving SDoH data in EHRs, there is opportunity to incorporate data from outside the medical field into health databases.

Predictive capabilities could be strengthened with data about people’s mental health, stress and energy levels, local weather, and driving habits. Community-based data and environmental sensors could also be rich sources of data. Like all data, these would need to be handled in a safe, secure, and ethical manner that respects people’s right to privacy.

A more comprehensive set of inputs about a person’s whole health would improve predictive capabilities over time. It also would detect changes in individuals’ health-related social needs, allowing predictive analytics to alert a person or their doctor to a potential health change before they even visit a medical clinic.

“You can imagine a scenario where a person with prediabetes begins experiencing food insecurity or barriers to transportation. They are so busy managing life that they miss checkups and have difficulties managing their health. Currently, nobody would know if their condition progressed to Type 2 diabetes,” Samuel said. “But what if community, healthcare, and government data were integrated? With the person’s consent, AI could tell their health plan that they have applied for assistance buying food. Their health plan could assess the situation with the person and recommend customized solutions, including transportation, meals, and/or home-based solutions and services to drive healthier outcomes.”

To make the most of predictive analytics, the healthcare system will need to find ways to consistently capture social drivers of health data and use it to improve health outcomes.