Top 3 Pros and Cons of Generative AI in Medical Record Keeping

Jul 25, 2023 | Blogs


Artificial intelligence (AI) has the potential to change everything about patient care, from customized treatment plans, rapid and more accurate diagnosis, and augmenting workflows. While many of these applications have yet to be realized, generative AI in medical record keeping is already having a tremendous impact on clinicians by automating charting in the electronic health record (EHR).

Estimates suggest that up to 40% of working hours in healthcare can be supported by generative AI.

Healthcare workers have struggled for years with reduced productivity, burnout, and inattentive patient care, largely driven by the burden of medical documentation. Powerful generative AI tools like ambient medical documentation services are rewriting this narrative. By transcribing the conversation between clinicians and patients and extracting relevant details into the medical record, clinicians are getting back in control over their workload. 

What we are seeing today is just the beginning of a revolution in healthcare. With announcements like Med-PaLM 2, Google’s medical Large Language Model (LLM), the potential applications for generative AI seem limitless within healthcare. Yet with any emergent technology, concerns have surfaced around its use, namely the tendency to present false information as factually accurate.

How are health systems using ambient charting to alleviate the biggest issues in the industry? Are concerns around generative AI in medical record keeping warranted? Explore the ways in which generative AI is affecting clinician workloads and patient outcomes.

Top 3 Pros of Generative AI in Medical Record Keeping


   1. Reducing Time on the EHR and Other Administrative Tasks

The rise of the EHR has revolutionized patient care by improving health outcomes and enhancing the patient experience. However, for all the good that the EHR has provided the healthcare industry, it has come at a significant cost: time.

Clinicians now find themselves spending up to 4.5 hours every day on charting alone.

Excess time spent in the EHR leads to lower productivity, fewer patients seen in a day and less revenue. 58% of clinicians believe that the amount of time they spend on charting and paperwork is inappropriate and takes them away from critical patient care. To regain lost hours, clinicians need a solution that allows them to more efficiently complete charting. Ambient medical documentation uses a combination of medical Automatic Speech Technology (ASR) and proprietary Natural Language Processing (NLP) to record a transcript of the clinician-patient conversation and extract relevant information to be reproduced in the medical record. From the perspective of the clinician, the use of generative AI is invisible or facilitated by a Medical Documentation Specialist. However, on the back-end generative AI is significantly improving the quality, accuracy, standardization, and speed of delivery of medical notes,  while giving back up to three hours every work day.

   2. Alleviating Clinician Burnout

Healthcare industry burnout has been a problem for decades, but in the wake of the COVID-19 pandemic, the issue has reached crisis proportions. Overworked and understaffed, nearly 1 in 5 healthcare workers have left the field in the last three years, and if this trend continues, there may be a shortage of up to 124,000 physicians by 2034.

The Mayo Clinic estimates that nearly 63% of clinicians now experience at least one symptom of burnout.

What is driving clinician burnout? According to a study conducted by EHR Intelligence, 57% of experts attribute burnout to documentation tasks such as charting and paperwork. Burnout is driving a staffing shortage within the industry which directly impacts a health system’s bottom line: it is estimated that turnover costs for physicians can equal 2 to 3 times their annual salary In order to address burnout, health systems need to address the source of the problem: charting. Health systems that adopt generative AI in medical record keeping have seen a 40% improvement in work-life balance. Automating these tasks has shown to significantly reduce burnout, help clinicians see more patients, and alleviate employee attrition.

  3. Improving Patient Outcomes

Not only do burnout and staffing shortages negatively affect healthcare workers’ well-being, but these issues can have serious implications for patient health, making enhanced medical documentation all the more critical. Clinicians experiencing burnout are twice as likely to have patient safety incidents such as medication errors. In addition, 9 in 10 nurses believe that the quality of patient care has suffered because of staffing shortages. Using generative AI in medical record keeping can help keep healthcare workers focused on patient needs while resolving urgent staffing issues. Ambient medical documentation also increases attentiveness at the point of care. During a medical exam, clinicians typically spend more time looking at a screen than looking at a patient.

Evidence by the American Medical Association Journal of Ethics supports the fact that clinician attentiveness directly correlates with better patient outcomes.

By removing the obstacle of the computer monitor, clinicians are able to exhibit better verbal and non–verbal communication, resulting in better comprehension of and adherence to treatment plans. Finally, utilizing generative AI in medical documentation produces more accurate charting. While concerns exist around the accuracy of generated responses, tight guardrails and proper quality control measures actually reduce the errors found in the medical record. Produced with greater consistency, AI-enabled charting also produces more consistent results and captures more wRVUs, directly impacting a health system’s bottom line. 

Top 3 Cons of Generative AI in Medical Record Keeping


    1. AI Hallucinations

An ongoing concern among clinicians who are reluctant to adopt generative AI is “hallucinations.” Newer LLMs like GPT-4 that have gained media attention have received criticism for inventing false information and presenting it as a fact. The “black box problem” in generative AI stems from this very issue: how can we trust an LLM to be used accurately when we don’t know or understand from where it derives its answers? In healthcare, a misplaced decimal can spell the difference between life and death for a patient. Generative AI must have tight guardrails in place to ensure the accuracy of a medical note. For AI-enabled medical documentation to work, it cannot fabricate information, and must rely on real numbers and valid information to ensure patient safety. Augmedix’s products use LLMs to extract relevant medical data from a visit’s transcript, but beyond LLM technology, the company’s proprietary NLP models form rigid rules around what information can and cannot be entered into the medical record. Tighter prompts or queries lead to significantly more accurate results, and quality control is built in to help prevent errors. In addition, quality control is monitored throughout the experience. For example, with Augmedix Notes, Medical Documentation Specialists review the results for quality assurance. For all Augmedix products, a note can never be submitted without a clinician’s final review and approval.

     2. Bias Concerns

Bias is an ongoing issue within the healthcare industry, and concern exists around generative AI perpetuating bias even further. The concern isn’t without merit. Discrimination around categories such as race or gender both consciously and unconsciously affect patient care and outcomes in clinical settings. While one would hope that AI would help clinicians be more objective in their approach to care, these disparities are also present in the data collected from medical centers. By using data from a hospital that favors information collected from one group over another, inappropriate care might be given that could result in harmful patient outcomes. Bias in healthcare once again stresses the importance of tight guardrails around generative AI. For AI-enabled medical documentation, these concerns are more limited in scope. Because the output of Augmedix’s charting is controlled by a proprietary NLP, the results aren’t freeform—they are based directly on the conversation between clinicians and patients. Our products have been thoroughly tested to ensure the highest degree of accuracy.

      3. Trust

The World Health Organization (WHO) released its first statement about the potentials and pitfalls of AI in healthcare. While it states that the excitement around healthcare AI is warranted, it cautioned that adopting these technologies too quickly could be dangerous and that thorough testing and oversight is necessary. “While WHO is enthusiastic about the appropriate use of technologies, including LLMs, to support health-care professionals, patients, researchers and scientists, there is concern that caution that would normally be exercised for any new technology is not being exercised consistently with LLMs. This includes widespread adherence to key values of transparency, inclusion, public engagement, expert supervision, and rigorous evaluation.” Augmedix CEO Manny Krakaris echoes these concerns: “While LLMs can be used as a supportive tool, clinicians and patients cannot rely on them as a standalone solution.”  For generative AI in healthcare to be a truly sustainable and effective solution, LLMs need to be combined with ASR, NLP, and structured data models to ensure the results are accurate and relevant. Augmedix is committed to safety and transparency in the development and use of our products. Generative AI in medical documentation remains safe, accurate, and well-tested, yet as these technologies develop and expand, it is critical that companies prioritize safety first. We have recently launched an AI Advisory Council to guide our responsible product development in pursuit of this aim.

The Augmedix Commitment to Progress, Efficiency and Safety

Documentation burden is one of the most pressing concerns facing clinicians today, driving burnout and causing medical professionals to leave the field permanently. By reducing time spent in the EHR, solutions like Augmedix use generative AI to target these issues directly while improving patient health outcomes. Ambient medical documentation is just the beginning of the AI revolution in healthcare. As we look towards the future potential artificial intelligence holds in healthcare, such as through precision medicine, Augmedix understands that caution and care must be taken around these powerful technologies. By placing tight controls around what information can be produced in medical charting, we have created products that safely and effectively enhance clinician workflows.

Interested in learning more? Read our white paper on how clinicians can maximize productivity and efficacy with ambient charting.