A newly-published study suggests that physicians may not be performing all of the services they document. While the study tracked medical residents rather than physicians who had completed their training, researchers seem to have concluded that their problems with adequate documentation might point to problems affecting all physicians.
To conduct the study, which appeared in JAMA Network Open, a group of 12 observers tracked 9 licensed emergency medicine residents during 180 encounters. These observers were tasked with determining the extent to which the residents documented their review of systems and physician examination taking place during the encounters.
The observations took place at emergency departments in 2 academic medical centers between 2016 and 2018. After observing the encounters, the monitors then reviewed EHR data to determine how accurately the ED residents documented their activities.
Upon reviewing both their audio recordings of the encounters, they found some meaningful differences between the interactions they had captured and what was showing up in the physician documentation.
For example, with regards to the review of systems, physicians documented a median of 14 systems, but the records could only confirm a median of 5 systems. All told, just 744 of 1,961 documented ROS systems (38.5%) were confirmed.
Meanwhile, for physical exams, resident physicians documented a median of 8 verifiable systems, while observers reported that they addressed only 5.5 systems. Overall, only 760 of 1,429 documented PE systems (53.2%) were confirmed by concurrent observation.
Not surprisingly, the researchers concluded that these inconsistencies between observed behavior and EHR-based documentation suggest that we should probably do more studies to find out how widespread this phenomenon may be.
It isn’t necessarily the case that these gaps exist across the entire medical record. As the paper notes, while physicians frequently dictate the history of present illness or medical decision-making know, they often use auto-populated text for ROS and PE sections.
Still, as they point out, payers could help resolve this problem by removing financial incentives to generate lengthy documentation. And of course, if physicians had a bit more time they could work more effectively with auto-populated text.
In any event, in the foreseeable future digital scribes and AI may very well take over much of the patient note generation process. (For an example of how this might work, check out our write up from last year on related work being done by Google AI research team Google Brain.)
Until then, however, this study offers a glimpse at a problem we should all take seriously. There just shouldn’t be major gaps between documentation and reality in EHRs.