Early identification key to care
Proactively identifying individuals at high risk for chronic conditions is key to providing them with the services and targeted interventions they may need. As a first step, it’s worth asking: How do individuals develop MCCs?
Family history and genetic makeup certainly play a role, but many chronic conditions can also be linked to lifestyle choices and behavior. Tobacco use and lack of physical activity, for example, have been tied to diabetes, cardiovascular disease and cancer.⁴
Lifestyle choices and behaviors are considered social determinants of health (SDOH). Also included in SDOH — and linked to a person’s overall health — are socioeconomic factors like education level, employment and housing stability, and environmental factors like access to healthy food, access to public parks or walkways and air quality.
Risks associated with these conditions can compound over time. For example, an individual may be more predisposed to asthma and type II diabetes if they live in an area without a nearby grocery store, poor air quality and few public green spaces to exercise in.⁵
To find people on the path to MCCs, we can bring together currently available data to predict who is likely to develop more acute forms of disease.
Artificial intelligence (AI) is one way to make these important predictions. First, data scientists train AI models to detect patients with a specific chronic condition, like diabetes or heart disease. AI models do this by analyzing historical data and finding patterns between electronic health record (EHR) data, health plan claims and other relevant datasets like patient-generated health data (PGHD) or SDOH. These models incorporate natural language processing (NLP) and machine learning (ML) techniques to extract and ingest relevant information.
Once AI models prove their ability to accurately identify people with a chronic condition, they can be fed new information to make predictions about other individuals at risk of developing chronic disease months or even years into the future. They do this by looking for common traits and data signals among people who have been diagnosed in the past.
Adding SDOH data to disease progression models can give an even clearer picture of population and individual-level risk for developing MCCs. When models are trained to derive information from data points that influence health — like food insecurity and poor environmental conditions — predictions become even more accurate.
There is a virtuous circle to these efforts. As models churn through more data and learn, their precision and accuracy increase. Then, when a patient is flagged as at-risk for a condition and is later actually diagnosed, that finding feeds back into the models, creating stronger predictive capabilities that benefit future identification of patients.