A Data Scientist Thinks About Population Health Management
(The Truven Health Advanced Analytics team is tasked with building new and differentiating analytic methods. Asked to explain some interesting new analytics that are important for managing populations, the Advanced Analytics team wanted first to explain how they’re thinking about Population Health Management.)
What is Population Health Management (PHM)? Much like the adage about the blind men and the elephant, Population Health Management can mean completely different things to different audiences. Hospital systems, practitioners, government, and private insurers all have different interpretations of what the term means. And, in fact, its implications are very different to each of these players.
For most health systems, PHM represents a complete paradigm shift from their traditional way of doing business. Think of it like this: Imagine you own an auto-repair business. Perhaps you have a single facility, perhaps a chain of facilities. You are generally responsible for fixing a car when it’s damaged, and perhaps also performing routine maintenance on that vehicle. Now imagine you are being told that:
- You are no longer simply responsible for the car when it is in your shop, but you are responsible for the car’s general care and maintenance for its lifetime.
- The insurance company is no longer paying for the specific services you provide, they are paying you based on the overall “health” of the cars that you service. You now need to know what happens to that car outside the walls of your facilities.
- You are no longer simply repairing the car when it needs it, you are being paid to keep the car “healthy” and out of your repair shop.
Imagine how foreign that would seem. You have no information about the drivers of the car other than what you can gather publicly. You have no idea what kind of driving record a person has, what kind of routine maintenance they perform on their car (except that which happens to occur in one of your shops), or what kinds of roads they drive on. In short, you have no knowledge of what kind of risk they bring to the table.
Hospital systems are in this situation. Historically, they have not needed to know much about their patients outside of what occurs within their facilities. They don’t have much information on where their patients are seeking care outside of their facilities, what kind of preventive care they are taking, what their social determinants of health are, nor how risky each patient is in terms of lifestyle and overall health, and they don’t have any input to their patients’ health benefit programs.
Now imagine you are the auto mechanic. Your repair shop owner is now asking you to understand the entire spectrum of a given vehicle you are servicing. Perhaps your specialty is body work, but you have to start thinking about the gas mileage and the health of the exhaust system in every car you see. Similarly, practitioners – particularly those who are not primary care physicians and are not used to thinking about “the whole patient” – struggle with the concept of population health because their focus is typically on one patient and one problem at a time.
Taking the analogy further, imagine you are the auto insurer (payer). You have historically managed payment for all the expenses for a given driver (and adjusted your rates to that driver based on their record/perceived risk). However, in this analogy as a healthcare insurer, your ability to refuse coverage to someone is diminishing, and your ability to assess risk is out of date, given that all drivers seem to be getting progressively worse and consequently more expensive. You are eager to shift some of that payment risk to the auto mechanics who are far more “hands on” with the cars, but there is no framework in which to plan and agree to terms. Plus you are still expected to maintain the risk for random “Acts of Nature” such as trees falling on cars, lightning strikes, and accidents caused by others. You are used to thinking about risk stratification and management at the group level, less so at the individual level.
Finally, imagine that you are the civil engineer responsible for designing the infrastructure on which the cars travel. You design roads to accommodate certain volumes, speeds, and types of vehicles, and support laws to enforce speed limits and construction zones. (Besides being the largest healthcare payer, this is the other role government plays in healthcare.) But now you’re being asked to help understand and contribute to improving the overall “health” of the vehicles on your roads, to do this in a way that minimizes the frequency and scope of needed repairs, and to do it all on a reduced budget. Oh, and at the same time, you have to be thinking about how to ensure safe roadways and service for new kinds of cars – self driving, connected, and beyond. . .
So how can Truven Health help? Our job as the analytics specialists is to help provide the information needed to expand the view of patients, and to present the information so that it’s actionable. Providing information on the full spectrum of care, even for something as specific as a surgical patient receiving a joint replacement (as our Bundled Care consultants do), can be invaluable in helping facilities, practitioners and payers understand the downstream implications of the care that is delivered. Helping them understand which patients are at high risk for “collision” (such as our new Risk of Hospitalization models) can lead to timely, cost-effective interventions. Identifying which segments of the population could most benefit from management (such as our forthcoming population classification method) can help focus activities for guiding patients and members towards health and well-being. Bringing valuable analytics to life can only happen if we first understand where our clients are coming from, and second, where they need to go to continue to be successful.
In coming blog posts, I will offer insights into the work of data scientists and into the analytics we are developing to help our clients continue to be successful.
Senior Director, Advanced Analytics