For years, success in Medicare Advantage (MA) risk adjustment was measured by one simple question:
Did our risk scores go up?
Higher risk scores meant higher payments, more flexibility, and more room to invest in member programs. As a result, risk analytics, chart reviews, in-home assessments, and suspect models all evolved around capturing as many diagnoses as possible.
The recent Senate report sends a clear message to the entire MA industry:
That way of thinking is no longer safe. This report isn’t just about one company. It’s a warning sign for every health plan that relies on risk adjustment.
The Real Problem: A Disconnect Between Coding and Care
The report highlights a growing gap between:
- How diagnoses are being captured VS
- How those diagnoses are actually used to care for members
Analytics, in-home assessments, and chart reviews aren’t bad tools. CMS expects health plans to identify missing diagnoses. The problem starts when these tools are used mainly to increase revenue, instead of supporting real clinical care.
CMS has already begun pushing back. The V28 model removed more than 2,000 ICD codes, signaling a clear shift away from rewarding coding intensity alone.
Where Risk Analytics Has Gone Wrong
1. Treating RAF as the End Goal Many programs have a singular focus:
“How do we capture every possible HCC?” This led to overuse of subjective diagnoses, heavy reliance on health risk assessments (HRAs) and chart reviews, and weak connection to follow-up care.
2. Diagnoses Without Clinical Continuity
The report highlights diagnoses that appear once…often during assessments…but never show up again in claims, labs, or treatment plans. From a CMS perspective, these look more like documentation artifacts than managed conditions (and who could blame them?).
3. Gray-Zone Conditions Driving Revenue
Conditions with vague thresholds and a history of CMS scrutiny(such as malnutrition, asymptomatic vascular disease, and BMI-based obesity)often drive risk adjustment factor (RAF) growth while increasing audit risk.
4. One-Directional Chart Reviews
Programs that only add diagnoses and rarely remove unsupported ones have drawn Department of Justice (DOJ) and Office of Inspector General (OIG) attention. Failing to correct errors is just as risky as creating them.
What Better Risk Analytics Looks Like
The report isn’t saying risk adjustment is wrong. It’s saying unchecked risk adjustment is dangerous. Stronger programs focus on:
Defensible risk scores tied to real encounters and follow-up care
Diagnosis quality, not just volume
Audit readiness, before audits happen
Provider support, helping clinicians document meaningful conditions correctly
CMS is clearly pushing the industry back to basics: real conditions, real care, real documentation.
How Technology Enables the Shift
Risk analytics has fundamentally changed. It’s no longer about finding more diagnosis codes. It’s about proving diagnoses are real, supported, and clinically tied to care delivery.
Health plans, regulators, and auditors are all asking the same questions:
- Was the condition truly evaluated?
- Was it monitored or treated?
- Is there clear, defensible clinical evidence?
Technology is what enables this shift—from volume-driven coding to evidence-driven risk validation.
Why NLP Matters
The majority of audit-ready evidence does not live in structured fields. It lives in unstructured clinical data, such as:
Provider progress notes
Discharge summaries
Assessments and care plans
Scanned documents and PDFs
Humans can read this information…but not at the scale required for modern risk adjustment, quality programs, or audits.
This is where Natural Language Processing (NLP) becomes essential.
When applied correctly, NLP helps validate:
- Whether a diagnosis is explicitly documented
- Whether it is active versus historical
- Whether there is supporting clinical evidence (assessment, monitoring, or treatment)
- Where in the medical record the evidence exists
NLP does not “create” diagnoses. It extracts and organizes clinical truth already present in the chart.
Without NLP, MA plans miss a crucial step in code validation, exposing them to unnecessary scrutiny.
Where GenAI Fits
Generative AI (GenAI) builds on NLP by improving usability, clarity, and efficiency for human reviewers.
Used responsibly, GenAI adds value by:
- Summarizing long and complex charts
- Explaining clinical context in plain language
- Highlighting missing or incomplete documentation
- Supporting coders, auditors, and clinicians with faster review
Importantly, GenAI should not make final coding decisions or automatically submit diagnoses. Its role is to assist humans, not replace them.
The goal is better decisions, not automated decisions. When NLP and GenAI are combined thoughtfully:
- Reviews become faster and more consistent
- Audit risk is reduced through stronger evidence
- Providers receive clearer feedback
- Risk adjustment becomes more defensible and clinically aligned
The report marks an inflection point for the entire MA industry. Health plans that continue to treat risk adjustment purely as a revenue optimization exercise will find themselves increasingly exposed…not just to audits and penalties, but to a fundamental loss of trust with regulators, providers, and members.
The organizations that will thrive in this new environment are those that reimagine risk analytics as a clinical integrity function, where technology like NLP and GenAI doesn’t just find more codes, but proves the codes submitted reflect genuine, managed conditions.
This shift requires courage as—it may mean lower RAF scores in the short term,—but it’s the only sustainable path forward. The question is no longer whether your risk scores went up.
It’s whether you can defend every point of RAF growth with clear clinical evidence. That’s not just good compliance. It’s the future of Medicare Advantage.
