Insights

In the early 2010s, when I developed Pulse8’s initial targeting algorithms, most of the work focused on the obvious question:

What diagnosis might be missing?

But I remember thinking we needed the inverse too:

What diagnosis is already there that maybe shouldn’t be?

We actually built it. We called it Detector8.

And yes, I still have the original logo to prove it.

We got a little carried away with the “8” in those early product names.

Back then, the market wasn’t exactly asking vendors to help identify diagnoses that could lower a risk score.

I wish 2026 Jonas could go back and pat early-2010s Jonas on the back and tell him his crystal ball was working that day.

Because OIG’s recent Medicare Advantage audits are a pretty good reminder that the inverse matters.

What caught my attention in the HumanaChoice audit wasn’t just the dollar amount. It was how OIG identified many of the diagnoses it chose to review.

An acute stroke diagnosis was there. Where was the corresponding hospital event?

An embolism diagnosis was there. Where was the anticoagulant?

An active cancer diagnosis was there. Where was the treatment pattern you’d expect to see around it?

Those aren’t automatically reasons to delete a diagnosis.

But they are very good reasons to ask:

Does everything else we know about this member make this diagnosis make sense?

Risk adjustment programs have gotten very good at finding evidence that suggests something may be missing from the data.

A strong risk adjustment compliance program needs the opposite muscle too:

finding diagnoses that are present when the surrounding longitudinal data says, “This deserves another look.”

Claims. Encounters. Pharmacy. Prior diagnoses. Site of service. Treatment patterns.

Put those signals together and you can create an exception queue before an auditor ever asks for the chart.

The goal isn’t to build an algorithm that automatically removes a diagnosis because an expected signal isn’t there. Clinical reality is much messier than that.

The goal is knowing which diagnoses deserve to move to the front of the review queue.

We called that idea Detector8 in the early 2010s.

I’m not suggesting anyone bring the name back.

But the idea? Maybe we were just a little early.