Validating Value Data: AI and Outcomes

October 8, 2019

Pharmaceutical Executive

Volume 39, Issue 10

Realizing the full potential of enabling technologies such as AI and machine learning in measuring treatment value and outcomes.

The application of new technologies such as AI and machine learning in establishing a value proposition is still a relatively new concept in pharma. As such, “it is likely to take some time before AI is able to demonstrate it improves objectivity and reduces uncertainties,” say Katya Svoboda, Taneesha Chawla, and Tanvi Ahuja, authors of the ICON report, The Impact of Artificial Intelligence on Outcomes Based Contracting. The authors point out that AI can be very helpful in identifying appropriate populations, outcomes and metrics, and predicting costs; in removing some uncertainty about how effective a therapy is in specific populations; and in understanding where a product can be placed in a treatment algorithm for a disease to achieve better outcomes. They also note, however, that use of AI in value-based and outcomes-based contracts, for example, is still susceptible to a “lack of trust” between payers and manufacturers. “This mistrust can lead to disagreement between both parties on contract design as well as execution,” they told Pharm Exec.

While recent research carried out by Healthcare Research & Analytics (HRA) on key players in the payer space revealed that half of the payers surveyed said they felt they had enough data to successfully develop an outcomes-based program with a pharma company, about one third felt they did not. And almost half of the payers surveyed insisted that their own data are the “most essential” in such programs, while less than 10% of respondents said data from a pharma company would be the most essential in making an outcomes-based contract a success.1

The question, then, of who “controls” the data around value and outcomes-and how pharma can further the reach of its own data-continues to loom over this area. “One of the key factors associated with value-based models is going to be defining the threshold metric that enables reimbursement,” Alan Louie told Pharm Exec’s Tech Panel. “Providers want that number to be very high so that everybody who gets treated responds fully. Pharma, on the other hand, wants a lower threshold, to give them the broadest pool of patients meeting measurable improvement thresholds. There is going to be a very strong discovery component to it, in terms of identifying the appropriate threshold supporting biomarkers, etc.” He adds, “From a reimbursement standpoint, there’s a lot of technology that will automate the determination of approval to reimburse and, presumably, there will be models built from the pharma side that allow them to maximize their value.”

Customers demanding patient-level value is one of “the confluence of factors [along with the ‘explosion’ of data, analytics, and digital capabilities] forcing the reinvention of pharma’s product development and market access models,” write the Boston Consulting Group’s Sam Marwaha, Michael Ruhl, and Paul Shorkey.2

To thrive in this era of value-based healthcare and personalization, they explain, “companies can’t just make incremental changes within their existing operating model.” Instead, they need an “end-to-end redesign,” which includes “identifying and prioritizing the evidence valued by payers early in the process, shifting the burden of evidence-generation from randomized controlled trials to new approaches that apply predictive analytics and real-world evidence… and engaging customers more proactively throughout the life cycle of a product.” The authors recommend that companies establish a “war room” to promote the culture of data science and create “cross-functional value teams that collaborate with decision-makers to ensure end-to-end optimization of decisions.”

In some cases, the technology itself could help to smooth the data concerns. As far as value- and outcomes-based contracts are concerned, Svoboda, Chawla, and Ahuja observe that “it is expected that AI will reduce the discrepancies seen and improve trust. Using a mutually agreed-upon AI algorithm would help improve data accuracy and remove biases, potentially making this process more efficient and accepted.” Of course, proceeding in this way requires constructive collaboration between the parties. The authors told Pharm Exec, “If a contract is formulated pre-launch, before real-world utilization data are available, a manufacturer’s clinical trial data need to be used.” It then “makes sense to reassess the contract using the payer’s own membership data, after the real-world utilization and outcomes data are available, and adjust the contract specifics if needed.”

However, realizing the full potential of applying enabling technologies such as AI in measuring value and outcomes seems to require a wider culture change. As Marwaha, Ruhl, and Shorkey note in their report, “The real hurdle to capturing value stems from a reluctance to make changes to core business processes, cultural resistance to using new data in decision making, and a disinclination to stop doing things the old way.”

 

Julian Upton is Pharm Exec’s Online and European Editor. He can be reached at jupton@mmhgroup.com

 

References

1. Carl Schneider and Roma Maksymowych, “Before Striking Outcome-Based Pharma Contracts, Payers Need Data,” Inside Digital Health, April 23, 2018.

2. Sam Marwaha, Michael Ruhl, and Paul Shorkey, “Doubling Pharma Value with Data Science,” bcg.com, Feb. 9, 2018.

 

 

 

download issueDownload Issue : Pharmaceutical Executive-10-01-2019