Managing and Predicting Risk: What Health Insurers Need to Know About Patient Data

Published on February 8, 2019

It’s an exciting time to be working with data in health care. With advances in artificial intelligence (AI), machine learning, blockchain and digital health – along with marketing segmentation and personalization – it seems the industry is making advancements every day with how we use patient data. These innovative tactics have the potential to improve population health, further modernize patient care and cut costs but only if they’re executed using the smartest approaches.

Let’s talk about what that means.

As reported recently by Healthcare Finance, insurers and health systems are using artificial intelligence and predictive modeling to manage risk: in one case to “know which members need disease management” and in another to “target outpatients who have a heightened risk of unnecessary hospitalization.” Healthcare Informatics reports here about the use of data on patients with asthma to improve population health and reduce the costs of managing the condition. The article describes the benefit this way: “Delivering excellence in value-based asthma care is predicated on the concept that we can simultaneously improve quality while containing costs. This presupposes that we’re using patient data to design and deliver more successful approaches to care, the results of which can then be measured and replicated.”

Stories like these – you can pull a number of them from the headlines every day – point to the “win-win” (for patients and insurers) from successfully analyzing and leveraging large quantities of data. However, it’s easy for health insurers to get overwhelmed by the process and make some missteps along the way, like chasing solutions that end up being far more complex than needed. Here are two important caveats that can help insurers build on the good work they’re already doing to manage and predict risk without going down paths that lead to dissatisfying results:

  • AI is a very useful tool, just like spreadsheets or EHR systems are tools, but whether or not these tools are useful in practice depends on what data sets are being used by AI to make its recommendations.
  • Claims data is certainly a good start, but clinical and claims data tell you the story of what happened in the past. Eighty percent of health and cost outcomes are driven by social and behavioral factors. Billing and ICD-10 codes do not help insurers identify and solve the upstream drivers of those claims.

From a practical standpoint, it’s critical to avoid wasting time and resources with “data for data’s sake.” What most payers need is not “build it yourself research projects” using raw data or algorithms: they need interactive, easy-to-use solutions that lead to actionable insights. Carrot MarketView, for example, brings together a universe of dozens of social determinants of health (SDoH) data sets and hundreds of validated predicted models into one integrated platform for spotting both future risks and opportunities, and its data visualization tools ensure ease of use for the “non-analyst” health care professional. Want to see it firsthand? Request a demo today!

Related Posts


Local and Individualized SDoH Data Is Critical to Advancing Health Equity


How Health Plans Can Improve the ROI of In-Home Test Kits


Q&A on the Carrot Community Committee (C3)