After after coding: imagining risk adjustment when accuracy breaks the business model 
Abstract
This essay speculates on a future version of risk adjustment in which overcoding is structurally prohibited and undercoding is no longer a meaningful source of underpayment. The premise is not that compliance finally wins, but that measurement improves enough to eliminate diagnosis capture as a primary economic lever. In that world, value based care does not become easier. It becomes more exposed. The essay explores what kinds of regulation could realistically produce this state, what business models would survive it, how capital and care would likely be misallocated during the transition, and why the end state is less morally satisfying but more economically honest than the current system.
Key themes
• Risk adjustment as a measurement problem, not a coding problem
• Regulatory designs that reward symmetry and evidence rather than documentation effort
• Business models that emerge when diagnosis arbitrage disappears
• Capital allocation failures inside VBC once RAF growth flattens
• Why accuracy creates fragility before it creates efficiency
Table of Contents
1. The End of Coding as Strategy
2. When Everyone Is Accurately Sick
3. Separating Discovery From Monetization
4. Turning Risk Scores Into Regulated Systems
5. Redesigning the HCC Instead of Policing It
6. Provider Behavior Without Documentation Leverage
7. What Breaks Inside VBC First
8. Where Founders Should and Should Not Build
9. The Political Economy of Accuracy
10. Why This Is Still Worth Doing
––––––––––––––––––
The End of Coding as Strategy
Risk adjustment has spent the last twenty years pretending it was about fairness when it was really about feasibility. Paying capitated plans without some proxy for disease burden was never going to work, so diagnosis codes stepped in as a stand-in for expected cost. That decision solved one problem and created another. Once diagnoses became a payment primitive, capturing them became strategy.
The system did not drift into this. It behaved exactly as designed. When payment is tied to documented disease states, organizations will invest in finding, documenting, and defending those states. They will build workforces, workflows, analytics, incentives, and software to do it at scale. They will focus on diagnoses with elastic thresholds, ambiguous criteria, or reflexive logic. They will get very good at adding and much worse at removing.
The pathology people now complain about is not abuse in the cartoon sense. It is industrial optimization in a system where documentation is monetized. This matters because banning overcoding without redesigning the substrate does not fix the problem. It just relocates it. The moment one set of diagnoses is excluded or downweighted, attention shifts to the next most profitable surface. The model teaches the market what to care about, and the market listens carefully.
The future imagined here starts when that lesson stops being economically useful. Not because everyone becomes ethical, but because measurement improves enough that diagnosis capture loses its leverage. In that world, coding is no longer a growth strategy. It becomes table stakes at best and liability at worst.
When Everyone Is Accurately Sick
One of the comforting myths in health policy is that better data automatically makes payment fairer. In reality, better data mostly makes mispricing harder to hide. When undercoding is no longer a problem, average disease burden goes up. Not because people suddenly get sicker, but because the system finally notices what was already there.
This creates a distribution problem. Historically underdocumented populations move upward toward their actuarial reality. Populations that benefited from aggressive documentation drift downward. Risk scores converge. Variance collapses. From a modeling perspective, this looks like success. From an operational perspective, it creates confusion.
When everyone is accurately sick, nobody is obviously sicker. Risk stratification becomes less discriminative. Care management programs that relied on diagnosis-derived tiers find half their population clustered in the same band. The instinctive response is to spread resources thinner. That feels safer than making sharper cuts, but it is also how capital gets misallocated quietly.
This is where value based care starts to wobble. Many care models assumed that improved documentation would align revenue with cost. What they discover instead is that revenue aligns with predicted cost, not realized cost. Prediction error becomes the dominant source of variance. The documentation subsidy that quietly covered operational inefficiency disappears. Care models have to stand on their own economics, and many cannot.
Separating Discovery From Monetization
