AI scribes add at least $167 per month to clinician income: 5-site study

In the three or so years since it burst into healthcare, ambient AI scribe technology has run away with the win in the market-uptake race. 

More than 80% of physicians already use the tools, which racked up $600 million in sales last year alone, and analysts project the U.S. medical scribe market will approach $3 billion by 2033.

All of which raises a pressing question: What has been the return on investment in all those software packages? 

The answer: modest reductions in EHR time and documentation time, accompanied by a small but meaningful increase in weekly visit volume. 

Those are the findings of researchers who looked into AI scribe ROI across five major academic medical centers—Mass General Brigham, Emory Healthcare, UC–San Francisco, UC–Davis and Yale New Haven Health. 

The team specifically found: 

• AI scribe adopters spent 13 minutes fewer using the EHR in total and 16 minutes fewer on documentation per eight hours of scheduled patient care. 
   
• Specific groups—including female clinicians, advanced practice clinicians, resident physicians and primary care and medical specialists—experienced “greater desirable changes” associated with AI scribe adoption. 
   
• Clinicians who used an AI scribe 50% or more of the time experienced twice the reduction in total EHR time and three times the reduction in documentation time.

Further, AI scribe adoption was associated with a 1.7% increase in weekly visit volume. 

This translated to a “conservatively estimated” additional $167 in monthly evaluation/management visit revenue per clinician. 

Lead author Lisa Rotenstein, MD, MBA, and co-authors share these and more details this month in JAMA.

Multisite, multivendor study sample 

For the study, Rotenstein and colleagues examined the work of more than 8,500 clinicians across the five institutions. Of these, around 1,800 were AI scribe adopters. 

The scribe technologies used across the adopter sample represented numerous, and randomly dispersed, vendor products. 

Most of the adopters, 74%, were attending physicians. Around 18% were advanced practice providers, 8% resident physicians. 

Meanwhile more than 60% were medical specialists, 24% were primary care providers, and 13% were surgical specialists. 

In their discussion, the authors comment that their multisite findings buttress evidence documented in previous single-site studies, which similarly found AI scribes reduce EHR time burdens with varying levels of success. 

“Given their derivation from pooled data from five health systems, these estimates are less likely to be biased by site-specific implementation details and represent generalizable results regarding the relative time savings associated with clinicians adopting AI scribes in real-world settings,” Rotenstein and co-authors comment. 

Qualifying the quantified monetary ROI 

Since none of the five organizations in the study required clinicians to book additional patients to qualify for AI scribe use, the $167 monthly revenue boost “may have derived from increased visits booked in available or non-templated time or changes in the composition of level-4 and level-5 evaluation/management visits coded facilitated by better documentation, as suggested by prior single-site evidence,” the authors point out. 

For that reason, the estimated financial ROI “could be interpreted as a conservative lower bound of the financial benefits of AI scribes and could be compared with site-specific implementation costs.” 

Rotenstein and team call for future studies to assess the longevity and reproducibility of their present observations, “as well as specific workflows and supports that can enhance the benefits of this technology.”

Ambient AI: Essential component of contemporary care? 

Commenting on the work for JAMA, Kaiser Permanente researchers Aaron Tierney, PhD, and colleagues note the Rotenstein et al. study highlights an imminent challenge. 

“While the community interested in quantifying the impact of health AI tools has become adept at measuring inputs and proximal outputs suitable for assessing time savings and financial impact (minutes, clicks, codes), it remains limited in measuring the effects of new technologies on the core priorities of the quintuple aim—patient experience, population outcomes and health equity,” Tierney and co-authors write. 

“If ambient AI is becoming a default component of healthcare delivery,” Tierney et al. add, “evaluation objectives must evolve accordingly.”

The study is posted here, the commentary here.