AI in Value Based Reimbursement

In the new climate of value-based reimbursement, some pharma companies are starting to adopt pricing strategies based on clinical outcomes and financial value. Value-based (VBC) or outcome-based contracts are designed to tie prices to how a drug performs in the “real world”: the price is higher if the drug delivers a desired clinical and economic outcome, and lower it does not. All five of largest US health insurers have signed at least one VBC. Harvard Pilgrim Health Care (HPHC), Advocate Health Care, Express Scripts and Optum Rx, have entered into multiple VBC. In addition, recently, representative of the America’s Health Insurance Plans (AHIP) asked the FDA to work with CMS to expand outcomes-based payments for drugs. VBC are designed to limit payers’ financial risk and usually expand the manufacturer market access because they contain provisions for the accelerated placement on payers formularies.

A list of current VBC can be found at: http://phrma-docs.phrma.org/files/dmfile/PhRMA_ValueBasedContracts_Q2_2019.pdf

Here are some examples of reported VBC for Drugs:

Merck has signed one of the first VBC in 2009 with Cigna, for two diabetes drugs, Januvia and Janumet. Under this agreement, Merck offered rebates if patients taking any oral antidiabetic drug (including those from other manufacturers) achieved certain medication adherence benchmarks and improved A1C. After the implementation of the agreement, Cigna and Merck reported a significant improvement of medication adherence and A1C. Also, Merck saw an increased sales volume resulting from preferred formulary placement associated with the VBC.

Novartis entered into VBP agreements with Aetna and with Cigna in February 2016 for Entresto, targeting the outcomes reported in one of the clinical trials: a 21% in heart failure hospitalizations, 20% reduction in the risk of death from cardiovascular causes, and a 16% reduction in all causes mortality compared with other treatment for heart failure. Under both agreements, the drug was placed on the preferred formulary, which reduces patient’s out-of-pocket costs.

Express Scripts and CVS have started to price certain oncology drugs based on cost-effectiveness for specific indications.

Amgen and HPHC have a money-back guarantee contract for Repatha: full reimbursement for the cost of Repatha for HPHC members who experience an MI or stroke while using the drug. HPHC has signed about 16 VBC, including for Entresto and Trulicity.

Impact of VBCs on cost sharing and patients outcomes

From 2015 to 2017, cost sharing was 28% percent lower for certain plans with outcomes-based contracts compared to the market average. Avalere found 33% of payers that used outcome-based contracts experienced cost savings and 38% reported improved patient outcomes. VBC can be constructed around cost and/ or on clinical outcomes. From a regulatory perspective, contracts based on medical costs are facilitated by the 21st Century Cures Act that make it easier for a manufacturer to communicate the economic advantages of a certain drug. VBC require that payors and manufacturer concur on metrics, risk stratification and overall methodology to be used to determine the drug’s effectiveness in the real word. An important point to consider is that the drug effectiveness must be evident within a short time frame acceptable by a Health Plan, usually around 12 months. Most VBC based on clinical outcomes use the FDA approved label as a target for determine the desired outcome. Regulatory barriers still exist preventing wider implementation of implementation of VBC by payor and Pharma. Some of the administrative barriers cited are the Best Price Rule (Medicaid); the Stark Law; the Anti-Kick-Back Law and the FDA rules on communications of cost benefits and non-label indications. CMS and the HHS are working on mitigation of risks associated with some of the above regulations. For example, drug performance could be evaluated on a population basis, rather than on an individual basis to address the “Best Price Rule” and CMS and HHS have the authority to grant waivers for the Stark and for the Anti-Kickback Statute laws. Also, the 21st Century Cures facilitates communications of health care economic information (HCEI) “related” to an approved indication (language “directly related” removed from previous rules).

The Role of AI and Machine Learning in support of VB Reimbursement Contracts

At Eversana we utilize a sophisticated proprietary artificial intelligence and machine learning platform to analyze data at a scale unique to each client’s needs to predict patterns and provide insight to help make critical business decisions.

To support VBC, we follows the patients health care journey leveraging various data sets, including administrative records and natural language processing (NLP) applied to clinical notes, physicians behavior and drug prescription patterns.

We leverage clinical, administrative and social determinant of health (SDOH) from various data sets, claims and structured and unstructured text from available clinical notes. The extracted features and available claims data are used as input into mutual information analyses (MI) as first step prior to building ML predictive models. Eventually ML predictive algorythms allow to identify distinguishing characteristics and drivers of health care outcomes for the population of interest compared with a propensity matched control population. When applied to drugs related outcomes, ML predictive analytics allow to identify the unique performance of the drug of interest in comparison with that of other drugs in the same class, focusing on specific clinical outcomes and medical costs. Furthermore, in the context of a complex drug regimen, MI and t ML predictive models allow to better understand the specific impact that a specific drug has on outcomes. Similarly, leveraging activity-based intelligence, neighborhood stress indexes and social media scans, our methodology allows to supplement data on the unmet patients’ social needs. Understanding unmet social needs allows the design of mitigating interventions to address social and economic barriers to adherence, one of the essential key to solve the value equation.

In conclusion, relevant real-world data are used to train machine learning models that allow us to predict, with accuracy, the appropriate patient population for which the drug of interest will provide the best clinical and economic value and the best channels to achieve optimal adherence. Simulation solutions further support cost-effectiveness analyses of the targeted drug when compared with other drugs in the same class and for similar indications.

Predictive analytics can be leveraged in negotiations of outcome-based contracts with payors, and to support formulary placement based on real world data. ML predictive models can be deployed on the payor infrastructure to monitor real-world drug performance and to facilitate efficient data sharing among insurers, providers, and pharma companies.