100%
of sepsis appeals now start with the agent — every case since rollout
~1hr
of chart review per appeal, by hand, for a three-FTE appeals team
8wks
from kickoff to go-live
The Challenge
Payers deny sepsis using their own definition, usually Sepsis-3, and downgrade the DRG. Every appeal means re-reading the chart, re-scoring the patient and rebutting the denial letter point by point.
The team appeals 95% of denials, and each letter takes about an hour of manual chart review to write — for a three-person appeals team handling every sepsis case.
We point out their errors in their arguments… I think that's more powerful, that's more strategic, and I think it's gonna increase our success rate.
Denials team
On the agent's appeals
The Solution
Bunkerhill's Long Horizon Agent now drafts every sepsis appeal. For each denial, the agent pulls new cases from the Epic work queue with the payer's letter, reads the letter — even scanned images — and lists every argument the payer made. It then rescores the chart, recalculating SOFA and SIRS from the record with every value cited, and drafts the appeal in the team's own template, answering each argument. A specialist edits in chat and finalizes every letter.
On one real denial, a Medicare Advantage plan applied Sepsis-3, claimed no acute SOFA increase of 2 or more points, removed sepsis (A41.9) as principal diagnosis and downgraded DRG 871 to 070. The agent's chart rescore found an acute SOFA increase of +7 points, every value cited to the record, and documented SIRS 4 of 4 — directly contradicting the denial's central claim. The drafted rebuttal answered each of the payer's arguments and waited for specialist review.
The Results
Eight weeks after kickoff, the rollout was complete: every sepsis appeal now runs through the agent, and first drafts generate automatically. The agent reads the denial, rebuilds the clinical case and writes the rebuttal — the team signs it.