Major policy changes usually arrive with speeches. This one arrived as a version number, took full effect in January, and has redistributed enormous sums of public money while attracting almost no coverage outside trade publications.
It is worth explaining, because it is one of the clearest examples anywhere of governing by formula rather than by legislation.
The formula
The United States covers more than thirty million older citizens through Medicare Advantage, where private insurers deliver government health benefits. The government pays those insurers monthly per member, and the amount is multiplied by a risk score built from the health conditions documented in that member’s records.
The mechanism converting conditions into scores is a payment model that groups thousands of diagnosis codes into weighted categories. Diabetes with complications carries one weight, heart failure another, and the sum produces the multiplier. Change the weights and you change where hundreds of billions of dollars flow, without amending a single statute.
Why it needed rewriting
The previous version had a structural flaw common to any model calibrated on data the model itself influences. Insurers learned which diagnoses moved scores, invested heavily in documenting those diagnoses, and the resulting data then informed the next calibration. The feedback loop ran for years.
The consequences became publicly visible this spring. Federal auditors examining three insurance plans found 81 to 91 percent of certain sampled high-risk diagnosis codes unsupported by the medical records behind them. A major insurer settled federal claims for 117.7 million dollars over how its diagnosis records had been assembled. Congressional advisers estimate the overpayments accumulated across the industry in the tens of billions of dollars annually.
What the new version changed
The rewritten model, fully in effect since January, restructured the condition categories, recalibrated the weights, and stripped scoring value from thousands of diagnosis codes that had become favoured targets of intensive chart review. Analyses of the V28 model impact on risk scores show the effect was deliberately asymmetric: organisations whose scores reflected genuinely complex populations saw modest change, while those whose scores leaned on aggressive documentation of low-specificity codes saw significant reductions.
A second change matters as much. Policy has tightened how diagnoses arriving without a clear link to an actual patient encounter are treated. The direction is unambiguous: documentation produced by review activity alone, disconnected from a real clinical visit, is losing standing in the payment calculation.
Alongside the rewrite came enforcement. The federal audit workforce grew from roughly forty reviewers to approximately two thousand certified coders, running on a rolling quarterly cycle, with sample error rates applied across entire contracts.
The governance question worth asking
For British readers, none of this describes a system we operate. But the method should be familiar, and it deserves more scrutiny than it gets.
The NHS allocates funds to regions using formulas built on population need and recorded data. Regulators across energy, water, and finance set obligations through methodologies rather than primary legislation. In each case, changing a technical parameter can redistribute more money than a well-publicised policy announcement, and it happens with a fraction of the scrutiny, because a version number does not make headlines.
The American case demonstrates both the power and the risk of that approach. The formula was quietly gamed for over a decade because the people affected by it understood it far better than the people watching it. The rewrite was equally quiet, and equally consequential.
Which suggests a simple principle for anyone covering, or overseeing, systems of this kind: follow the methodology documents, not just the ministerial statements. That is where the money actually moves, usually while everyone is looking somewhere else.
