Pharma APMs: A Learning Curve Awaits You, Pharma

They can work, but pharma will need to master the intricacies of adherence metrics, real-world outcomes, and sampling.

Maggie Alston
Bruce Pyenson

Second of two parts

Alternative payment models (APMs)—such as bundled payments, shared-savings arrangements, and capitation—are now common between insurers and health care providers. APMs often include incentives to reduce costs and improve outcomes with the insurer and provider sharing financial responsibility for the patient’s cost of care and health outcomes.

As pharmaceutical manufacturers face increasing public pressure over prices and value, we believe the advantages of pharmaceutical APMs will be increasingly attractive to both pharmaceutical manufacturers and insurers. But pharmaceutical APMs are more nuanced and resource intensive than the traditional fee-for-service contracts that pharmaceutical manufacturers negotiate with payers. There are some valuable lessons for pharmaceutical manufacturers in provider APMs, but whether pharmaceutical APMs succeed or fail will depend on finding solutions to operational and logistical challenges—some of which are unique to the pharmaceutical industry.

Who is responsible for which patients?

Maggie Alston

Bruce Pyenson

Patient attribution is integral to any APM. For provider APMs, attribution means determining which provider is financially responsible for which patients. Attribution is usually based on identifying the provider who has the most “control” over the patient’s care. But patient attribution is tricky because patients use health care in many different ways, so determining attribution can involve complex rules.

In pharmaceutical APMs, patient attribution equates to adherence to a particular medication, because patients cannot be expected to benefit from a product they do not use. But adherence metrics are imperfect, partly because filling a prescription does not mean the patient actually took the drug—it means only they took possession of it. Mechanisms for filling prescriptions like auto-refill, mail order, and 90-day orders are convenient for patients, but they also make determining true medication adherence challenging.

The best way to calculate patient adherence differs by medication. Most prescriptions, such as blood pressure medication, cover a specific number of days, so adherence is usually calculated as the number of days a patient is holding a prescription (medication possession ratio) or the number of days a patient is covered by a prescription (proportion of days covered). For some medications—insulin is a prime example—these calculations don’t work because the medication’s dosing differs by patient and fluctuates over time. For these types of pharmaceuticals, adherence metrics can still be calculated, but metrics for such medications will be subject to more uncertainty. Another problem for pharmaceutical APMs is the inaccurate information that finds its way into claims databases. This is an issue that plagues provider APMs, but it could be an even bigger challenge for pharmaceutical APMs. For example, the days’ supply value that appears on pharmacy claims is integral to adherence metrics, but it often contains inaccuracies. A claim for an auto-injectable pen with exactly four weekly doses, which should always appear as a 28-day supply, can be entered by the dispensing pharmacy as a four-day supply (the number of doses) or even a one-day supply (the number of boxes dispensed). Improving pharmacy benefit data will be an early focus of pharmaceutical APMs.

Gold mines of real-world evidence

The targets used to measure APM performance are critically important because the APM’s shared savings are determined by comparing actual spending to a target. Targets that are too hard to reach could mean the APM will not generate any shared savings payments. On the other hand, targets that are too easy are not sustainable because insurers will not want to keep paying out money based on illusory savings.

Insurers usually set targets like per-member spending using past real-world experience that is often based on experience of the provider or the provider’s peers. For example, CMS’s Oncology Care Model uses trended real-world historical data to develop its targets. Determining outcomes based on real-world performance rather than clinical trial data would be a major change, but a necessary one for pharmaceutical APMs because insurers are going to be most interested in the outcomes for their patient populations.

Pharmaceutical APM targets based on real-world experience can be judged by comparing them to historical data. In contrast, targets based on clinical trial data may not correspond well with outcomes observed in administrative claims—and the dependence on administrative claims could produce unexpected results for the pharmaceutical APM. For example, real-world patients might have more comorbidities than those in a clinical trial. Furthermore, even if clinical trial data indicate that one product is superior to another, differences in population demographics, real-world adherence, and outcome variations could swing a pharmaceutical APM’s results in the opposite direction. In the end, the product that looks best in clinical trial reports might not perform as well under a particular pharmaceutical APM.

On the upside, insurers, health care providers, and other health care stakeholders may be keenly interested in pharmaceutical APMs because they could be a gold mine of data and related insights into real-world pharmaceutical performance.

Sampling populations

Pharmaceutical APMs may require the collection of information beyond the claims data that many provider APMs depend on. Long-established, if imperfect, provider performance metrics, such as readmission rates and total cost of care, can be calculated using medical claims data; however, descriptive clinical information like weight and cancer stage are not directly captured; in most cases, insurers do not receive laboratory results or medical records. Therefore, clinical-event metrics available through administrative claims data will be more practical to implement.

A pharmaceutical APM that measures an outcome found only in laboratory results or in electronic medical records will require new, secure processes to protect, obtain, extract, vet, and analyze these sources. If manual chart audits are required, the administrative burden could be even higher. It is also unlikely that laboratory results or medical records would be available for all of the insurer’s patients.

These data constraints point toward the need for sampling from populations. Sampling is very much a part of modern contractual relationships among insurers, health care providers, regulators, and suppliers. At present, sampling is used by several provider APMs, including CMS’s Pioneer and Next Generation ACOs. Although sampling is not a common component of pharmaceutical business operations, measurements based on samples could be crucial for some pharmaceutical APMs, especially when it is impractical to obtain full population data.

Managing new risks

Managing APM risk requires real-world data analytics to quantify the financial impact of outcomes. Although pharmaceutical manufacturers have expertise in data analytics for clinical trials and outcomes research, successfully managing pharmaceutical APMs requires expertise in different quantitative skills—risk/actuarial analytics. For example, forecasting financial results and accruals are needed because APM results are typically determined soon after the financial reporting period closes despite reporting lags. Pharmaceutical APM forecasts and analytics will not only need to generate information that flows into the company’s audited financial statements, but will also need to consider regulations that are specific to the industry, like Medicaid Best Price, Average Sales Price, and the Anti-Kickback Statute.

As pharmaceutical APMs grow in potential, pharmaceutical manufacturers may choose to adopt more proactive risk management tools, like the actuarial control cycle. The actuarial control cycle promotes proactive risk management through actionable feedback from the four stages of the life cycle of a contract: development, monitoring, evaluation, and renegotiation.

While we don’t expect APMs to dominate insurer-pharmaceutical contracts, it seems inevitable that pharmaceutical APMs will become part of the solution to high health care spending—and to managing the cost of drugs. Some of today’s toe-in-the-water contracts could be foundational. Given a little time and experience, these new pharmaceutical APMs could start to happen fast, thanks to the lessons from provider APMs.

Maggie Alston is a senior health care analytics consultant at Milliman. Bruce Pyenson, FSA, MAAA, is a principal and consulting actuary at the company. The research for this piece was funded by Pharmaceutical Research and Manufacturers of America.