Unfiltered conversations on healthcare revenue cycle, risk adjustment, AI in payer operations, and what's actually moving the needle for U.S. health plans and providers. Hosted by the Lumiknox clinical and technology leadership team.
The v28 transition has quietly changed how chronic conditions map to HCCs. We walk through the 75 new codes, 267 deletions, and the recalibration impact on plan revenue - with a checklist for coding ops leaders.
A frank debrief on our first full agentic prior-auth deployment - the wins, the edge cases that tripped the agent, the human escalation patterns, and the 67% cycle time reduction we did not expect to hit this fast.
Which measures moved the composite rating most? How we sequenced interventions, handled hybrid season surge staffing, and kept physician engagement alive through abstraction cycles. With a guest from the plan's quality team.
CMS's final RADV rule introduced extrapolation for the first time. We break down what it means for audit exposure, documentation standards, and why plans relying on retrospective chart review alone are under-prepared.
Statistical power, confidence intervals, and the three things a 50-chart pilot should measure to give a plan actionable signal. Plus: what we see most often when coding accuracy looks fine on the surface but isn't.
Most utilization models are overfit on historical data and underperform in live environments. We discuss what good feature engineering looks like for healthcare claims data, and where LLMs genuinely help vs. add noise.
A close look at the three concurrent-coding workflows that separate top-decile MA plans from the middle of the pack.
What worked, what broke, and the supervision pattern that finally made unattended denial triage stick across 11 provider organizations.
Tukey outlier deletion is in. Here is how to recalibrate your gap-closure investment portfolio for the new cut-point regime.
Eleven items every payer ops leader should validate before the next sweep. The CMS-HCC v28 transition has changed coder workflow at the line level.
Three categories of edits eat the gap between current and theoretical max first-pass rates. The fix is mostly configuration discipline, not platform replacement.
Why 50 is the smallest sample that is big enough to mean anything - and what its confidence intervals actually tell you about your baseline accuracy.