Risk stratification scoring model for vulnerable populations


At the start of the COVID-19 pandemic, the healthcare community faced many uncertainties. ERs were flooded with COVID patients and providers saw utilization rates for scheduled patients drop dramatically, causing a gap in patient care and decreased revenue.

One of our clients, a large provider group, experienced similar issues and wanted a solution that helped identify the most vulnerable patients in their population and prioritize outreach in case they need care. The client came to EXL Health as a collaborative partner to develop a framework to prioritize their risk-bearing Medicare population.

Human ingenuity

At EXL Health, we leveraged our extensive data sets and advanced analytics knowledge to build a flexible scoring model to risk stratify our client’s entire population. The model predicts and identifies at-risk vulnerable populations, and then prioritizes patients who would benefit from care outreach for either a telehealth consult or an in-person office visit.

When designing the model, EXL Health incorporated prominent medical trends at the time with guidance from our knowledgeable provider client. Together, we developed an algorithm that identified conditions causing additional vulnerability to COVID and its adverse effects using a patient’s medical history, existing conditions, lab and medication utilization, and the length of time from their last physician visit.

To operationalize the output from the algorithm, EXL Health provided our client with prioritization lists stratified into low, medium, and high vulnerability. These were then used to reach out to the patients who would benefit from a scheduled visit or telehealth consult.


Originally used as a pilot in 2020, our client applied it across their population of about 45,000 patients. At the peak of COVID uncertainty, our provider client was able to schedule over 80-85% of their vulnerable patients to have consults or in-person visits, a rate higher than pre-COVID measures.

Our vulnerability model correlated with morality rate of the population in scope:

  • High vulnerability – 7% morality rate
  • Medium vulnerability – 2% mortality rate
  • Low vulnerability – 1% mortality rate

Due to the success of the model, our client implemented the model across their population in 2021, impacting almost 200,000 patients.

Through this prioritization and outreach, our client’s patient population received more timely care and the client was able to recapture risk accurately for their population.