ideasIdea
Pay for cancer AI only when it has outcome evidence, then pay properly
Health systems would pay for AI tools that have shown in trials that they help patients, and pay nothing for tools that have not, giving makers a reason to run the trials.
Reimbursement for AI is haphazard: a few tools have billing codes on weak evidence, most have none, so vendors sell on workflow rather than outcomes. The proposal is a payer policy: a temporary payment for AI under coverage with evidence development while a prospective trial runs, converting to durable payment if outcome evidence is positive and ending if not, with payment levels reflecting demonstrated value. Medicare's coverage decisions for a handful of AI devices are a starting point.
Hypothesis
Outcome-conditional reimbursement will increase the number of prospective AI trials started per year and shift the market toward tools with demonstrated benefit.
Rationale
Payment is the strongest signal to developers; where payers required evidence (e.g., for genomic tests via Medicare's MolDX), evidence generation followed.
What would test it
One national payer adopts the policy for two years; count AI trials initiated and tools reaching durable coverage versus the prior period.
Maturity
speculative
Who has to act
payer
Cost to try
Medium ($1M to $50M)
Years to first evidence
3
Bottlenecks it attacks
- AI that is built but not validated or deployed · Thousands of cancer AI models are published; a handful are in clinical use, and fewer have shown they help patients.
- Incentives reward me-too drugs and marginal gains · The system pays the same for a drug that adds two months as for a cure, so companies race to copy rather than to cure.