In 2021, better forecasting can help health organizations meet the challenges of the continuing COVID-19 pandemic while improving the health of their communities and of individuals. PwC’s Health Research Institute spoke with PwC principals Sundar Subramanian and Paul D’Alessandro about the importance of improving health organizations’ predictive modeling capabilities.
In HRI’s fall 2020 survey of health executives, nearly 75% of respondents said their organizations would invest more in predictive modeling in 2021. What has the pandemic shown health leaders about the importance of better forecasting?
Forecasting has been based on taking historical data and projecting forward. The pandemic has shown that the construct completely breaks down when you encounter something so different from the historical experience, and it tests your resilience. CFOs were in the dark because their healthcare cost modeling couldn’t account for the new set of conditions. When an event like this happens, consumers behave differently and the usage of services is markedly different. COVID illustrated the difference between truly predictive thinking and historic-projection thinking, and it was a painful lesson.
In predictive modeling, three ingredients make it real: having real computing power, understanding the underlying science and, most importantly, having the data on which to build these models. Every CEO has seen report upon report of unprocessed data, and they are saying that it’s time to do something with it. The pandemic amplified the presence of unprocessed data and the lack of effort to do enough with it.
HRI: How can better forecasting help health organizations with vaccination efforts?
Paul D’Alessandro: Pharmaceutical executives are confident that eventually we’ll solve the early logistical issues of distribution. What scares them is that a portion of the population will remain reticent to engage in the vaccine. It’s a consumer confidence issue. We’ll have to have leading-edge adopters and influencers giving confidence to other folks. It will require interesting modeling to understand the right buttons to push to gain consumer confidence.
Sundar Subramanian: We face the daunting task of figuring out how to scale up this physical health intervention to something like 3 million shots a day. Everybody has a role to play: retail clinics, payers and insurance companies, manufacturers, hospitals, distributors. There needs to be consumer education, capacity planning and management of the vaccine flow. All that needs to be optimized, so the community needs to come together with data and thoughtful planning.
HRI: How important is it to have leaders who can convene the various actors to address the pandemic or even to address social determinants of health?
Sundar Subramanian: Eighty percent of a person’s health is determined by their ZIP code—what education they had, the employment outlook, and all the other factors affected by where they live. That’s why we see so much racial disparity and inequity. To really change healthcare, the whole ecosystem has to work together: digital health players, community workers, state or local government, area health plans and provider systems, and the local physicians. If you come together, share information and act on members’ behalf, you improve outcomes pretty significantly for the member, and reduce costs and create value for everybody.
Paul D’Alessandro: Look at the reaction to the pandemic at the federal level. The government threw the vaccines out to the states and told them to optimize however they could. That state construct broke down. Everyone began to realize that somebody needed to coordinate this across multiple entities so that people in one state in the second priority group weren’t getting the vaccine before another state had finished with the first priority group. In general, there have to be incentives in place; otherwise, companies need to locally or regionally optimize.
HRI: When trying to improve forecasting abilities, should health organizations expand their data sources beyond clinical data?
Paul D’Alessandro: Companies have been reticent to go beyond the usual suspects to fill those gaps in their data. Healthcare companies are now turning to alternative sources that can be very informative for social determinants of health purposes. Machine learning and other new techniques have allowed us to glue all this data together so the unusual suspects can be brought to bear like never before.
Sundar Subramanian: Companies exist in an ecosystem, so they need to have macro-level data that gets at the intersection of health, consumer behavior and economics. The pandemic showed these are very intertwined. All three data sets are very important to understand, and if you can model over them, you can get to the company-specific data sets needed for financial, medical, clinical and actuarial context.
HRI: How can health organizations improve individual health once they have data to give them better context and understanding of individual patients’ lives?
Sundar Subramanian: Modeling for the future means showing which pathway has better results so clinicians can present that information to consumers and consumers will be more likely to take better actions and stick with them.
Paul D’Alessandro: A good analogy is the anti-lock braking system in cars. It’s a model that takes action to optimize our brakes. We need to stop bad actions and start good actions, what we call the path of least resistance. It’s identifying the path of least resistance through predictive models on an individual basis, just like we’ve done in mechanical systems for years.
HRI: How can health organizations use real-time data insights to improve community health?
Paul D’Alessandro: In the past when we talked about population health management, it was top-down. Now we’re able to look at populations from the bottom up because we have the ability to do things at scale like never before. When in retrospect organizations see that community investments were wasted, people say they wish they could have known. Predictive models allow us to take away that 20/20 hindsight and go into things with more clarity.
Sundar Subramanian: Data is absolutely critical to avoid wasting valuable resources. Modeling can help with things like identifying food or transportation deserts so community health investments can be targeted where the intervention will have an impact. Prioritizing your intervention based on the impact on the healthy life years of that community can help you be much more focused on solving problems. Some problems take decades to solve, so you better have that kind of data and science to predict the impact before you invest.
HRI: In our survey, 73% of healthcare executives said they were starting to collaborate or had plans to collaborate with other providers and payers as a result of the pandemic. Why is it important to convene a broader response to health challenges?
Sundar Subramanian: In healthcare, it really does take a village. It’s understanding the science, engaging with consumers, bringing in clinical interventions, understanding gaps in care, and aligning incentives across all players. There are so many examples of how, when people work collaboratively, everyone wins. Yet modeling has not been set up to do that. It’s unfortunate that it takes a pandemic to get to 73% agreeing on the need to collaborate. The next question is: What are we going to do about it?
Paul D’Alessandro: In modeling, we use the terms local and global optimization. When you’re hiking up a mountain, get to the top and then see another huge one up ahead, that’s the world of local optimization. We’ve been living in that world because the incentives are often aligned to local optimization, and there’s a feeling that even if I do everything I possibly can, I’m going to end up with local optimization. Our resiliency work has been based on the notion that we all live in an ecosystem and that unless you begin to join forces and respect all the feedback loops that are in play, you’re always going to be climbing the shortest hill.