Home  |   Subscribe  |   Resources  |   Reprints  |   Writers' Guidelines

Winter 2024 Issue

Automation and Documentation in Health Care
By Susan Chapman, MA, MFA, PGYT
For The Record
Vol. 36 No. 1 P. 16

Technology can relieve the burden of documentation for clinicians and help mitigate burnout.

Physicians and nurses often become overwhelmed by the administrative demands that come with providing high-quality care. In fact, many clinicians are citing documentation and administrative tasks as a contributing factor to an increase in clinician burnout syndrome.1 “I think when we talk about the administrative burden, certainly the documentation burden, and the technical workflows comprising many sources of data, all of the multiple clinical activities that are going on, including distractions and fatigue, all of that can become very challenging for providers,” says Whende Carroll, MSN, RN, NI-BC, FHIMSS, clinical informatics advisor at HIMSS.

One way the health care industry has sought to address those challenges is through advances in HIM. Among those innovations is automation, which can significantly reduce the burden of documentation on clinicians.

The Power of Automation
Natural language processing (NLP), a technology that converts voice to text, is a form of artificial intelligence (AI) and critical to HIM automation. “NLP can summarize large amounts of sources of data—that could be textual, video, or audio—and get them into other digital health technologies and electronic health records where clinicians spend most of their time doing documentation,” Carroll says.

“This technology is at work while the physician is in the process of documenting the note and can ping them about the specificity required to capture the acuity of the patient,” adds Tami McMasters Gomez, director of coding and clinical documentation integrity (CDI) services, HIM division at the University of California, Davis (UC Davis). “We’ve customized this process to be more in line with our clinical practice. There is some evidence-based reasoning that exists with [NLP] that looks for abnormal lab and clinical findings and suggests that the doctor add a corresponding diagnosis. The technology is available in both the outpatient and inpatient space, but both require clinical indicators, risk factors, and treatment—the compliance component. It doesn’t eliminate the need for human intervention but can reduce the number of retrospective queries that require providers to refamiliarize themselves with information after the patient encounter has closed,” she says.

According to Kathleen McGrow, DNP, chief nursing information officer at Microsoft, “Automation not only can reduce the number of retrospective coding queries but it also can decrease turnaround time to respond to these queries and the number of ‘unable to determine’ coding queries. This results in fewer ‘discharged but not final billed’ open charts, increased provider engagement, increased satisfaction of coding specialists, and identification of additional opportunities for documentation improvements.”

This use of AI could help address the critical coder shortage that health care faces. According to the American Medical Association, across the United States there’s a 30% shortage in medical coders, which may be less due to turnover and more likely due to the growth of the health care industry.2

Most organizations already employ computer-assisted coding, a form of AI that has the ability to learn from the user. “We’re now moving into a space for autonomous coding,” McMasters Gomez says. “In the coding space, the administrative burden is less on the physicians, but it’s more about efficiencies and meeting the need for the staffing shortage with coders. It also gives coders time to spend on more complex cases. I see autonomous coding being more for professional evaluation and management levels and ancillary services like radiology, lab, and for some emergency medicine physicians. And I see it allowing coders to serve as more second-level reviewers to ensure the integrity of the final coded encounter. Still, as far as physician administrative burden when organizations require physicians to do their own coding, autonomous coding will definitely remove the administrative burden for providers.”

UC Davis is using an AI product called Code Confidence, which has the ability to learn from its users. “This product offers a threshold that you can turn on. You can be conservative, somewhere in the middle, or very aggressive,” McMasters Gomez explains. “It auto-assigns diagnosis codes based on historical information; for instance, machine learning says that 90% of the time or greater when these symptoms are documented, this code is accurate. It then automatically assigns the code list for the coder, and the coder can still engage to agree or disagree. They can give a thumbs up or remove the code, respectively. It’s actually more than a simple computer-assisted coding application.”

Ambient technology is another arm of AI and a form of NLP. Voice-assisted technology and hardware systems are able to capture video and voice and summarize large amounts of data. “Ambient scribe documentation is an example of this technology,” McMasters Gomez states. “Providers could walk into a room, and while they’re walking in, this AI technology will pick up on the conversation with the patient and document everything.”

This technology additionally can be used to further automate clinical documentation, enabling providers to streamline the creation of medical documentation and improve clinical efficiency. “Ambient technology can play a key role in reducing burnout and freeing providers’ time to focus on delivering high-quality care,” McGrow notes.

“Predictive modeling is a form of AI that can be very beneficial,” she adds. “These tools can suggest possible inputs based on patterns in the data, reducing the time spent on data entry. For example, missed appointments not only lead to a decline in patients’ health but they can have a negative impact on clinic operations due to overstaffing. Microsoft Cloud for Healthcare offers a missed appointments prediction AI solution that helps providers predict patient no-shows, allowing staff to reach out to the patients to help ensure they make it to their appointments. Other predictive modeling tools can help with patient scheduling to help providers better adjust appointments and optimize staff-resource allocation so that the right nurse is in the right place at the right time.”

There are numerous other tools that also can be used throughout the revenue cycle and in patient care. For example, there are AI algorithms that are able to read documentation to gauge the mean length of stay to help manage patient stays. “In the web-cycle space, there is a great deal of automation around billing and preauthorizations,” McMasters Gomez says. “These are all things that generally require human beings to do the work. There are even bots that communicate with insurance companies, and insurance companies also have bots. Health care bots will dispute with another bot; they’ll go back and forth. That mitigates the need to have a human being on the phone eight hours a day conversing with another person on the payer side.”

Generative AI—which can create new output from existing data—a technology currently in its infancy, can be used like NLP. This technology can offer prompts and help clinicians obtain resources such as policies, procedures, protocols, and guidelines from multiple sources to help with patient education, developing care plans, and other tasks. “Many of those activities are efficiency and productivity breakers—a real bottleneck to productivity,” Carroll explains. “Generative AI can also help with deduplication of sources and information. If you have one contextual prompt, you can get the answers that you need.”

There are many other applications for automation that can ease the burden for providers. AI can help complete routine tasks, such as administrative work, data entry, billing, charting, and scheduling, which can allow clinicians to focus on their patients. It can also be significant in reducing documentation errors, which can lead to improved patient safety and quality of care.

Integrating the Technology
Carroll reports that there are many products, solutions, and inventions on the market that are being offered by different companies. Some vendors are startups, and some are highly recognizable names. Each organization must choose which vendors and products are appropriate for their needs.

McMasters Gomez points out that UC Davis works closely with vendors and curates clinical criteria. “There is an algorithm for a nudge, it’s specific clinical criteria, and you have to use language that delineates between words like ‘and’ or ‘or,’ so it’s very specific. For example, if a patient presents with low sodium, we have to curate the technology’s logic to be very specific about when we want it to alert the clinician and when we want to ask for documentation of a diagnosis of hyponatremia. So, we spend a lot of time creating the code logic. Usually, a CDI manager and analyst develop the clinical criteria that we want to use and how we wish it to work. We send that information to the vendor and physician champion for review and approval; the vendor then puts it into a test environment. We test it, decide if we need to adjust the logic or use it as it is, and then we move on. We can do that with any diagnosis. But, getting physician buy-in is key,” McMasters Gomez says.

Throughout this process, UC Davis is not only adopting new technologies, honing them, and creating best practices, the vendors it works with are also integrating what they are learning through this collaborative process. “We’re serving as a model for other organizations that can then benefit not only from the example, but from what UC Davis has been able to help vendors curate,” McMasters Gomez shares. “We’re helping to further develop the products, and the process so they can benefit the whole health care industry as opposed to just one institution or one system of hospitals.”

AI is only as good as the documentation available and an organization’s practices. Therefore, governance is an important aspect of the technology’s implementation and overall success. The conversation around AI’s safe implementation in health care has brought with it ethical and regulatory concerns. Such issues center around transparency, particularly as it involves AI algorithms, bias, and the safeguarding of data used to train the technology as AI is rapidly integrated into the health care system.

A 2019 article in the Journal of the American Medical Informatics Association aimed to move the topic from the conversation stage and into that of implementation. The authors cite ethical challenges facing AI implementation in health care, most notably, “potential biases in AI models, protection of patient privacy, and gaining the trust of clinicians and the general public in the use of AI in health care.”3 They concede “the ethical integrity and public role of the health professions relies on maintaining broad public trust,” and that those health professions using AI had to address those ethical concerns in order to maintain their reputations and integrity. To that end, the authors offer recommendations to address AI governance and they propose a model for governance in health care based on fairness, transparency, trustworthiness, and accountability. The tools the authors suggest included data governance panels, education, and patient consent and support.3

“Among the safeguards that can help AI be implemented securely, there needs to be a committee that provides strategic oversight to ensure that we don’t, in fact, take the entry-level people out of the mix because when we bring somebody on, somebody’s going to be able to manage the AI and do some of that remedial work that’s necessary to ensure safe implementation,” McMasters Gomez says. “Teaching that AI entry-level processes is going to be key so we can ease the burden downstream. But also, I think having a very thoughtful approach to it so that we don’t end up causing more burden is also important. In some instances, AI could require physicians to interact more than they currently do, perhaps to get the technology to where we need it to be. Having a thoughtful process around how we develop, curate, govern, and strategically roll out AI is critical, especially in the physician realm.”

Improving Quality of Life and Patient Care
An American Medical Association 2016 study found that for each hour a physician spent with a patient, some two additional hours were spent on the EHR and other administrative work.4 Because automation can help relieve much of that administrative burden, it’s expected to improve quality of life overall for overworked providers. “These technologies can free up resources for nurses who spend 20% to 40% of their time documenting in charts and not being with patients, and then also what they call ‘pajama time’ for providers who are working after hours. So, these technologies can help efficiencies, reduce waste, and increase productivity. They can also create balance and well-being for clinicians at a time when that is really critical to the challenges of the workforce, the shortages that we’re seeing, and those who are just burned out and dissatisfied,” Carroll says.

And while McGrow agrees that HIM automation can alleviate some of the documentation burdens on clinicians, it’s not a complete solution. “The burden of documentation is a multifactorial problem, resulting from complicated billing guidelines, insurance rules, historical processes that clinicians are accustomed to, and more,” she says. “Therefore, a comprehensive approach that addresses these factors is necessary to effectively reduce the documentation burden on clinicians.”

McMasters Gomez sees the benefits beyond those afforded providers as well. “Integrating AI efficiently will benefit everyone involved, the clinicians, the coders, the CDI, at every level of a health care organization,” she says. “Ultimately, though, it offers an improvement for patients, who are better served during their hospital stays and even after discharge. They may have less back and forth with their insurance companies, for instance, if things are more accurate. Things like that can have a big impact on patient care and the patient experience as a whole.”

— Susan Chapman, MA, MFA, PGYT, is a Los Angeles-based freelance writer and editor.


1. Budd J. Burnout Related to electronic health record use in primary care. J Prim Care Community Health. 2023;14:21501319231166921.

2. Lubell, J. Addressing another health care shortage: medical coders. American Medical Association website. https://www.ama-assn.org/about/leadership/addressing-another-health-care-shortage-medical-coders. Published April 19, 2023.

3. Reddy S, Allan S, Coghlan S, Cooper P. A governance model for the application of AI in health care. J Am Med Inform Assoc. 2020;27(3):491-497.

4. Robinson KE, Kersey JA. Novel electronic health record (EHR) education intervention in large healthcare organization improves quality, efficiency, time, and impact on burnout. Medicine (Baltimore). 2018;97(38):e12319.