Introduction: Rising drug costs have sparked interest in tying payments to outcomes to reduce payer risk of adopting low-value interventions. However, performance-linked reimbursement (PLR) plans are poorly understood. We modeled the effects of PLR plans on the distribution of therapy value obtained using sacubitril-valsartan as an illustrative case.
Methods: We modeled PLR plans that issued refunds/discounts based on reductions in cardiovascular mortality or heart failure hospitalization. We used probabilistic sensitivity analyses from a published cost-effectiveness model of sacubitril-valsartan to model real-world uncertainty. We compared the distribution of value obtained (net monetary benefit [NMB]) with PLR plans and conventional payment across 10,000 paired simulations. Price premiums were set so average reimbursement was equal with both schemes. We varied plan structure and treatment/population uncertainty. With successful plans, payment would decrease in simulations with less benefit and increase with more benefit - decreasing the NMB standard deviation. We also calculated incremental NMBs as valuations of the change in NMB variability using a risk-averse utility curve.
Results: A plan that adjusted drug reimbursement based on the ratio of observed to expected reduction of cardiovascular mortality reduced the standard deviation of NMB by 16.6% (Table 1). The incremental NMB - an estimate of per-patient value of implementing the PLR plan - was only $35, 3.3% of the NMB with conventional payment. Cardiovascular mortality plans were more successful than heart failure hospitalization plans. With rare outcomes, refunds based on cohort event rates were larger than refunds for individual adverse events. Plans improved with greater treatment effect uncertainty, larger refunds, and more payer risk-aversion. Their value decreased with more natural history uncertainty. Plans unraveled if the payer asymmetrically selected higher risk patients.
Conclusion: PLR plans require outcomes tightly associated with treatment value and minimally influenced by unrelated uncertainty. Both sides must agree on a predictable population and limit asymmetric selection. For rare outcomes, plans should evaluate population event rates or surrogate measures that consistently reflect the clinical benefit.