New report underscores the importance of documentation. Can AI reduce the burden on clinicians?

Crico benchmarking documentation cover

For the Record: The Effect of Documentation on Defensibility and Patient Safety

While the increasing burden of paperwork and record-keeping is often cited as a contributor to clinician burnout, a new report from Candello, a division of CRICO, underscores the need for timely and accurate documentation in patient care. 

Based on more than 65,000 closed medical professional liability cases in a national database, the report finds that 20% include “documentation failures,” such as undocumented findings and unclear clinical notes. Half of the cases with documentation failures resulted in high-severity injury or death, compared with 41% of cases free of documentation failure. 

The report comes at a time of heightened interest in the process of clinical documentation. The American Medical Informatics Association (AMIA) and the National Library of Medicine convened a symposium in early 2021 to explore how to reduce the burden of documentation on clinicians. Focused explicitly on patient care delivery and clinician-patient communication, the symposium issued a report with 82 action items to help providers, vendors and advocacy groups develop solutions. And in 2022, AMIA charged a task force with reducing the burden of documentation on health professionals to 25% of the 2022 level by 2027. 

The Candello report examines the role of documentation by service and specialty. For example, in surgery, accurate documentation of informed consent, incidental clinical findings, and surgical item counts proves critical. In medical subspecialties, diagnosis, care transitions and medication plans are points of vulnerability that require careful documentation.

The report offers ways to improve safety and lessen the risk of liability claims with high-quality documentation, largely through fine-tuning accepted practices such as closed-loop communication with patients needing follow-up care, accurate clinical notes and medical record audits. Advancements in technology and artificial intelligence (AI) promise to improve many of these processes, but they come with challenges related to data quality, potential inaccuracies and more.

System-based improvements, such as the use of electronic records and AI tools, must be carefully checked and verified by clinicians. Generative AI, or genAI, offers advanced tools that create text and other original content but must still be checked for errors.

The need for careful review may counteract AI’s benefits

Ironically, as automation improves, humans tasked with checking for errors become less reliable. During a Grand Rounds on genAI, Bob Wachter, M.D., Chair of the Department of Medicine at the University of California, San Francisco observed, “It's a problem when the systems are correct 90% of the time, and a human has to watch and catch those times when the system is not right.” 

open quote

It's a problem when the systems are correct 90% of the time, and a human has to watch and catch those times when the system is not right.

Bob Wachter, M.D., Chair of the Department of Medicine, University of California San Francisco
close quote

The tendency to over-trust systems that are usually right — known as “automation complacency” — has been studied in aviation and other industries and is referenced in a 2024 report from the IHI Lucian Leape Institute.

The Institute’s report acknowledges that while genAI can lighten workloads and cognitive demands for busy clinicians — developing patient history summaries and drafting responses to patient emails — the need for careful review counteracts the benefit. The report calls the need for human oversight “a major challenge to patient safety,” citing automation complacency, “deskilling” due to overreliance on AI tools, and the danger that it will not lower time stressors for clinicians in the long run. Organizations may be tempted to reclaim the time saved through automation by adding new demands to clinicians’ schedules.

A JAMA viewpoint co-authored by Dr. Wachter suggests features that could be built into AI tools to help alleviate these problems and “promote clinician vigilance” but also points out that “it is perilous to assume that clinician vigilance is an acceptable safeguard against AI faults.”

In a 2023 TED talk, Eric Topol, M.D., offers a more optimistic view of genAI for clinical documentation. He sees a future for clinicians that includes “keyboard liberation,” a time when technology quickly and accurately handles record-keeping, giving providers more time to connect with patients directly without a computer in the way. 

speech-bubble

We want to hear from you!

Email us your feedback and comments: PatientSafetyBeat@BetsyLehmanCenterMA.gov