Can Targeted Provider Outreach Change Claim Behavior?
by Yubin Park, Co-Founder / CTO and Evan Brociner, Data Product Lead
TL;DR: We looked back at 72 provider outreach letters covering 50 providers. The claims showed real corrections after the letters went out. Early evidence also points to a possible effect on future billing—a promising signal we can test as more data develops.
Finding an unusual billing pattern is one thing. Getting someone to do something about it is another.
That is the idea behind targeted provider outreach. A client reviews a specific set of claims, sends the provider a letter explaining the concern, and gives them a chance to check the work. Sometimes the issue is documentation. Sometimes it is a coding habit or a coverage rule that was misunderstood. Not every strange claim needs to begin with an accusation.
We have seen providers correct claims after receiving these letters. The obvious next question was whether we could measure that effect more carefully.
Sent Letters Versus Letters That Stayed Drafts
The study looked at 72 valid letters sent to 50 providers. It compared them with letters that had been drafted through the same general process but never sent. That gave us a practical control group: similar concerns had been identified, but only one group actually received the outreach.
We looked at two things. First, did providers go back and revise the earlier claims named in the outreach? Second, did payment for related services change afterward?
Those sound similar, but they are not. A cancelled claim is something we can see. Future payment asks us to estimate what might have happened without the letter. That second question requires much more caution.
The Claim Corrections Were Real
The first look found $45,722 in cancellation-backed reversals. At the latest observation point, $13,134 remained reversed.
Why did the number move? Because claims keep moving. A cancellation can later be adjusted or resubmitted. Calling the first number a permanent recovery would be easy, but it would not be honest. Initial reversals and retained reversals tell different parts of the story.
We also built progressively closer control groups using specialty, baseline revenue, and overlap in BETOS service categories. In the main Tier 2 comparison, 11 providers had useful controls. They showed $18,150 in observed reversals against $6,346 in estimated background reversal. The difference was $11,804.
This was not a randomized trial, so we cannot say every reversed dollar was caused by a letter. We can say the contacted providers made measurable corrections beyond the background level estimated from the matched controls.
Future-Payment Results Show Early Promise
The later-payment analysis produced a large estimate, but the strongest comparison only included two providers and one of them drove most of the result. That is not enough to make a broad savings claim.
Maybe the letter changed the provider's billing. Maybe the patient mix changed, a high-cost episode ended, or the latest claims were not fully developed yet. It could also be ordinary regression to the mean after the unusual activity that caused the provider to be reviewed in the first place.
So we are treating this finding as a reason to keep watching, not as money already saved. Observed corrections, retained reversals, background activity, and estimated future effects should not be rolled into one impressive-looking number.
What Good Outreach Looks Like
The letter needs to be specific enough that the provider knows exactly what to review. It should connect the claims to the relevant evidence without jumping straight to an accusation. And a human should stay in the loop. Falcon can surface the pattern and help draft the outreach, but the client reviews and sends it.
The paper trail matters too. The claims, the approved letter, the send date, the response, and the later transactions need to stay connected. Otherwise, it becomes almost impossible to tell whether the outreach worked.
The takeaway is clear: targeted, human-reviewed outreach was followed by measurable claim corrections beyond expected background activity. The early future-payment signal gives us a promising next question to test as more data develops.