Verification Debt: The Lesson Carbon Accounting Should Take From Healthcare

Every executive in the energy transition knows the uncomfortable open question at the centre of it. The commitments are real, the capital is moving, and the reporting frameworks are maturing fast. But underneath sits a dependency that rarely gets discussed at conferences: enormous sums now flow according to numbers that reporting entities largely produce about themselves.

If you want to know how that story ends without waiting fifteen years, American healthcare has just finished running the experiment.

The parallel system

Medicare Advantage covers more than thirty million older Americans through private insurers. The government pays each insurer monthly, and the amount rises with how ill each member’s documented medical conditions show them to be. Reasonable design: caring for sicker people costs more, and without adjustment insurers would avoid them.

The structural feature worth noting is that the paying party depends on documentation generated by the party being paid. Not fraudulently, just structurally. The insurer’s own records, produced by clinicians the insurer works with, determine the insurer’s revenue.

Verification existed, in principle, from the beginning. In practice it was sporadic for well over a decade, and the gap between incentive and verification is where the trouble accumulated.

What accumulated

Insurers built substantial operations to re-read patients’ historical records looking for conditions that could be added, each addition raising the payment. Nothing equivalent was built to find conditions that should be removed.

The results became public this spring. Federal auditors examining three insurance plans found that 81 to 91 percent of the high-risk diagnosis codes they sampled were not adequately supported by the underlying records. A major insurer paid 117.7 million dollars in March to settle federal claims focused specifically on the one-directional design of its review programme. Congressional advisers estimate the accumulated overpayments across the industry in the tens of billions of dollars annually.

Call it verification debt. Like technical debt, it accrues invisibly, it is cheap to ignore in any given quarter, and the repayment terms are set by whoever eventually calls it in.

The correction, and its cost

The repair has been expensive and structural. The federal audit workforce grew from roughly forty reviewers to around two thousand certified coders on a rolling quarterly cycle. Sample error rates are now applied across entire contracts, which converts a modest inspection into a material financial event. Organisations have rebuilt internal processes around continuous self-audit, and the emerging standard for defensible risk adjustment for payers rests on four requirements: every claimed condition tied to a real dated encounter, supported by evidence in the record, with the reasoning traceable and the whole chain reconstructable years later by a hostile reviewer.

Note that none of those four are exotic. They are what any assurance professional would specify at the outset. They cost far more to retrofit under enforcement pressure than they would have cost to build in.

Why this belongs on an energy executive’s desk

The parallels are close enough to be uncomfortable.

Emissions data, offset quality, hydrogen provenance, grid attribute certificates: in each case, substantial value transfers according to measurements and attestations produced by interested parties, with verification regimes that are improving but were not built in from day one. Scope 3 reporting in particular rests heavily on supply chain self-declaration.

Three transferable principles come out of the healthcare case.

Fund verification at the same time as the incentive, not after the first scandal. The gap is the exposure, and it compounds.

Insist on provenance rather than plausibility. The healthcare standard now requires each claimed fact to trace to a specific dated event with identifiable participants and retained evidence. Applied to carbon accounting, that is the difference between a credible attribution and a well-formatted estimate.

Watch the direction of corrections. The most damning finding in healthcare was not any individual error. It was that a decade of “accuracy improvements” had almost never produced a correction that reduced revenue. Any assurance regime should be able to answer, on demand, how many of its restatements went against the reporting entity’s interest. A ratio near zero is a finding in itself, and it requires no proof of intent.

The energy transition has one meaningful advantage over the healthcare programme that just paid for this lesson: it is still early enough to design the verification in. That advantage has an expiry date, and healthcare’s experience suggests it is shorter than it feels.