Medical Record Maestro: Reimagining CDI in the Age of AI and the Longitudinal Record

Download the episode here.

We need a “human in the loop” in mid-revenue cycle work, experts say. 

What they fail to address is the more interesting question:  

Precisely where? 

Back-end AI fact-checker? Front-end query authorizer? Or, maybe something like my current OTR guest Penny Jefferson envisions: Medical record maestro. 

The end-to-end connectedness of the medical record, evolving API standards, increased use of prior authorization, episode-based reimbursement models, and review of all of this by AI and other tools makes documentation cohesion more important than ever.  

CDI is changing with the times, but often not fast enough. It must change, or risk stagnation and possible extinction. The ready availability of vast amounts of information in the EHR and AI-enabled reviews and audits means a slip in a diagnosis can result in a big DRG downgrade or denial of stay altogether. 

But an AI powered enabled CDI maestro conducting the show can make a record strong, front to back. 

Penny has been writing like a fiend about many of these and other related topics for LinkedIn and the likes of ICD-10 Monitor–which is what led her back to the hot seat of the Off the Record studio. Listen in as we discuss: 

• A day in the life of: Penny’s evolving role at UCDavis 

• Her take on today’s CDI work, where is it coming up short and/or at risk of being automated away? 

• UR function/medical necessity and how CDI can assist case management without igniting a turf war. 

• “Longitudinal episode integrity”: Episode-based bundled payments (TEAM, etc.) and how they shift chart review beyond discrete diagnosis review to breadth 

• APIs and FHIR standards: Issues with fields that don’t match throughout the medical record leading to denials, bringing it home with an encephalopathy example 

• Where does CDI need to upskill to meet these new demands? Does the “traditional” chart review/query role need to change … or do we need an entirely new role to evolve to meet these demands? And what about productivity metrics? 

• Addressing the elephant in the room. Where is the human in the loop? For example: should a human review every AI query before it goes out? Spend more time on high dollar, 5+ day stays? Tiebreak a co-equal dx? 

• Updates on admit type. Listeners might recall our previous show on this topic, with  loose NUBC definitions leading to urgent vs. elective categorization to minimize impact on quality metrics. 

• Other fun stuff you only get on #OTR