Reversing a five-year risk-score slide — and beating the recovery target the defensible way

The situation

I was brought in to a regional Medicare Advantage plan staring at a significant Medicare shortfall, with risk adjustment expected to recover half of it — and no one able to explain why the plan’s risk scores had fallen so far. Leadership handed me two suspects: a submissions problem and a deletes problem. They also mentioned, almost in passing, that they’d let their provider-facing staff go a few years earlier. That last detail turned out to be the whole story.

What was actually wrong

The easy answers didn’t survive contact. On my fifth day, the team was set to present a $30M submissions opportunity. The night before, I asked them to confirm one thing — that the encounters they were counting were actually risk-adjustable. By morning, the $30M was $0.5M. A single unasked question had been propping up a number sixty times too large, and that gap — between what the data appeared to say and what it actually meant — was the same gap that had let the whole program drift.

The real cause sat upstream. With no one engaging providers for years, the providers heard the message loud and clear: this plan doesn’t care. Clinicians put their documentation attention where it was being asked for — the other payers actively engaging them, not us — and ours quietly decayed for want of anyone paying attention.

What I did

I rebuilt the relationship with providers first. We gave them prioritized member lists so their time went to the members who mattered, improved the incentives so the effort was worth it, and put integrated, concurrent risk adjustment tools in front of them so gaps surfaced in the EHR at the point of care — where a treating physician can actually document them. The sentinel effect came back, and performance followed.

Provider performance wouldn’t have moved the number if the data couldn’t get out the door, so I fixed the pipes — made the submission pipeline reliable end to end and put the vendors back under real oversight — so the accurate documentation actually reached CMS. And to make it last, I translated operational performance into financial models my Finance and Actuary partners could act on, so hitting or missing a target became a visible line on the plan’s P&L rather than an operational abstraction.

The outcome

Average RAF recovered from 0.85 to 0.95 in 18 months — 160% of the recovery goal I’d been brought in to hit. Every point of it was either documented by a provider treating the patient or verified by an auditor, and everything we couldn’t stand behind, we let go. For the first time, Finance could see risk adjustment as the P&L lever it actually is.

What this means for your plan

The lesson here is specific. Finding the truth behind an inaccurate RAF takes a detailed assessment, data you can trust, and the willingness to confront bad news instead of the convenient answer. Solving it takes more than a tidier process — it takes a real will to change, hands-on provider and vendor performance management, and support for the work across the enterprise. Where those exist, the score recovers, and it holds up.