How Predictive Analytics Can Identify Rising Risk for COVID-19

By Emmet O'Gara
May. 12, 2020

Share this post:

Some areas of the United States believe that they have reached the peak of COVID-19 infections, while others continue to experience rising numbers of cases. Nationwide, cities and states are contemplating what life looks like after the peak and how social distancing guidelines can eventually be loosened in safe ways.

Predictive analytics can play an important role in these strategies. As Harvard Business Review noted in a recent article, companies have used predictive analytics for years to make consumer credit decisions, as well as to make product pricing decisions. In today’s world, predictive analytics can also be used to identify individuals who are at higher risk for contracting COVID-19.

Healthcare systems can use these insights to inform members and patients about how to proceed safely in the coming days and weeks, as our country adjusts to a new normal. For instance, individuals in high risk groups may be advised to continue social distancing and other precautions.

 Adopting Predictive Analytics in the Age of COVID-19

Implementing a predictive analytics program may seem daunting to some healthcare organizations, but it doesn’t have to be that way. HMS’ Elli, for example, is a risk intelligence, risk stratification, and analytics platform that combines both clinical and non-clinical data about members and patients. Elli can identify specific population cohorts and provides prescriptive insights to care management teams.

Let’s consider the types of data that would be used to create a predictive analytics model for identifying individuals who are more vulnerable to severe complications resulting from COVID-19:

  • Diagnosis Codes. On February 20, 2020, the Centers for Disease Control and Prevention (CDC) announced supplemental coding guidance for healthcare encounters related to COVID-19. This included codes related to pneumonia, acute bronchitis, lower respiratory infection, and acute respiratory distress syndrome (ARDS), as well as possible and actual exposure to COVID-19.

 In addition, on March 18, 2020, CDC announced a new diagnosis code that will be used to identify COVID-19 cases effective April 1, 2020. With these diagnosis codes, it’s possible to quantify COVID-19 utilization measures, such as admit rates, emergency department rates, ICU rates, and mortality rates.

  • Risk Factors for Severe Illness Due to COVID-19. According to information from CDC, there are several health conditions that can contribute to more severe cases of COVID-19. These include chronic lung disease, moderate to severe asthma, serious heart conditions (CAD/CHF), severe obesity (body mass index of 40 or higher), diabetes, chronic kidney disease that requires dialysis, and liver disease. People who are immunocompromised are also at higher risk. This may be due to cancer treatment, smoking, bone marrow or organ transplantation, immune deficiencies, poorly controlled HIV or AIDS, and prolonged use of corticosteroids or other immune weakening medications.
  • Demographic Information. Data related to age, gender, and ZIP code are all helpful for predictive analytics. Information related to social determinants of health (SDoH) based on census tract level data is also valuable.

Based on these types of data, Elli can build a predictive model to determine which members or patients are at greater risk from COVID-19. This can answer a variety of important questions, such as:

  • Which populations with specific risk factors are likely to experience higher rates of ER or ICU admissions?
  • Based on SDoH information, which individuals may face barriers to care during the pandemic? For instance, it may be helpful to ask to what degree in the past month, members or patients have had concerns about life necessities, such as having a place to live, having enough to eat, or feeling like they are safe?
  • Are individuals with COVID-19 experiencing a worsening of other chronic conditions?
  • Do environmental factors, such as the weather or air quality, have an effect on infection rates? If so, can higher risk populations be warned proactively?

Taking Insights About COVID-19 Risks to the Next Level

Using predictive analytics to identify people who are at higher risk from COVID-19 is just the first step. In the weeks and months ahead, healthcare organizations must make these insights actionable by communicating with vulnerable populations. The key is to focus on proactive care by identifying and then helping individuals before they end up in high-cost settings. By focusing on rising risk populations and proactively urging them to take action now, healthcare organizations can prevent people from becoming high risk.

HMS’ Eliza platform takes information from Elli and transforms it into customized engagement campaigns that healthcare organizations can easily deploy to members and patients. These outreach campaigns deliver information, but they also can be used to collect data from individuals. Here are some ideas for communicating with high-risk populations:

  • Design outreach campaigns as questionnaires which determine whether members or patients should visit providers for COVID-19-related assessment and testing. For instance, healthcare organizations may want to update their health risk assessments to include COVID-19-related questions that gauge exposure and/or risks. Additional assessment questions might include: “Have you stayed in a hospital recently?” “During the past four weeks, how often have you had shortness of breath?” or “How would you rate your overall health in the past month?”
  • Use responses to outreach campaigns as input to predictive models. An unfortunate by-product of COVID-19 is a decrease in healthcare visits to monitor chronic medical problems. Responses to outreach campaigns can evaluate the impact that deferred appointments may be having on members’ and patients’ daily lives. For instance, it may be helpful to ask questions like: “During a typical day, does your health limit you a lot, a little, or not at all doing moderate activities like vacuuming or playing golf?” Or “During the past four weeks, how much of the time that you had any problems with your work or other regular activities or accomplished less than you would like as a result of emotional problems, such as feeling depressed or anxious?” Based on member and patient responses, healthcare organizations can evaluate rising risk and initiate the appropriate proactive measures.
  • Conduct regular followups related to behavioral health. In light of social isolation, loneliness, and the stresses of life in general, members diagnosed with anxiety, panic disorders, or related mental health conditions may need increased medical services. During phone, email, or text follow-ups, healthcare organizations may want to inquire about medication adherence and identify potential barriers such as forgetfulness or failure to refill prescriptions. Other helpful questions may include: “Compared to one year ago, how would you rate your overall emotional health today?” or “Over the past 2 weeks, have you experienced a significant change in the amount you normally sleep, either trouble getting to sleep or sleeping too much?” Answers to these types of questions can assess an individual’s current state and prompt real-time transfer to health coaches for those at higher risk.

Conclusion

With herd immunity to COVID-19 far away, a reasonable alternative is to develop evidence-based recommendations for loosening stay-at-home orders. Given the heterogeneity of our nation, in terms of health conditions, age, social determinants of health, and more, a one-size-fits-all solution is impossible. Risk forecasts based on predictive analytics models are a promising way to develop more personalized plans for the next phase of life with COVID-19. To get a head start, many healthcare organizations are leveraging population health management solutions.

Your Details

Please enter your details to access this resource.