Over the past decade, physician burnout has evolved from a serious concern to a troubling epidemic, affecting 50 percent of physicians and residents. Excessive workloads, process inefficiencies, and administrative burdens associated with electronic medical records (EMR) hinder productivity and significantly impact physician well-being.

In the United States, physicians spend between 34 and 55 percent of their workday compiling clinical documentation and reviewing EMRs. While some of this contributes to ongoing patient care, much of it is done in the name of billing documentation, litigation defense, and regulatory compliance. These excessive documentation requirements also strain the patient-physician relationship, reducing the time patients spend with their doctors and impeding effective communication and care.

“Many healthcare institutions are turning to artificial intelligence (AI) as a potential solution to physician overwork and burnout.”

Studies have shown that medical scribes are efficient substitutes for reducing administrative burden and increasing physician productivity. However, they are not widely used due to high initial costs, patient discomfort, and high turnover.

Now, many healthcare organizations are turning to artificial intelligence (AI) as a potential solution to physician overwork and burnout. After all, AI offers long-awaited benefits for clinicians, such as reduced time spent on clinical documentation and improved patient focus during visits, plus more accurate visit notes.

Unfortunately, clinicians’ well-being is not always the highest priority for hospital administrators. Currently, most U.S. hospitals generate revenue based on “fee-for-service” payment models, in which providers pay a set price for each procedure performed or patient seen, with no explicit rewards for the quality of service provided. Internally, this model translates into an incentive system that prioritizes the number of patients over the value of the patient. That is, physicians are incentivized to maximize revenue by seeing and treating as many patients as possible.

Rewards for quality and value creation are rare, if present at all, in physician compensation plans. As the time previously spent compiling EMR documentation becomes available, institutions may increase their expectations for the patient and procedure volumes that physicians perform. That is, AI could inadvertently exacerbate the patient volume problem rather than easing physician workloads and improving the patient experience.

Healthcare is at a crossroads: we need policies and organizational practices that align clinician well-being with patient value, rather than pitting the two against each other, to harness AI for the greater good.

Many healthcare institutions are starting to use AI

AI, which includes machines and software capable of replicating human behavior in solving complex problems, can be trained on large datasets and perform tasks involving search, reasoning, and learning. In healthcare, AI helps physicians diagnose diseases, prescribe optimal treatment plans, and improve patient engagement with personalized approaches. In addition, AI can handle administrative tasks, promising to reduce clinicians’ workloads, increase job satisfaction, reduce burnout, and improve the quality of patient care.

Many healthcare institutions are beginning to harness the transformative potential of AI. For example:

• CommonSpirit Health recently announced the launch of ‘Insightli’, an AI tool aimed at streamlining workflows and creating customized content.
• In 2022, Amazon introduced Amazon Clinic, a virtual healthcare service that uses AI to provide affordable treatments for common ailments.
• According to the organization, by October 2023, 10,000 Permanente Medical Group clinicians and employees will have started using AI to reduce the time clinicians spend documenting EMRs and enable “more personalized, meaningful and effective interactions with patients.”
• Several hospitals across the United States have tested AI scribes to tackle the growing burden of data entry and clinical documentation. Typically, these tools use smartphone microphones to transcribe physician-patient encounters, then leverage machine learning and natural language processing to convert verbal interactions into summarized, accurate visit notes and suggest billing codes, all without retaining audio recordings to protect sensitive information.

AI’s ability to improve clinical documentation, billing, and reimbursement processes while allowing physicians to focus on patients has the potential to reduce burnout rates, increase job satisfaction, and improve patient care. Given these early successes, broader adoption of AI seems only a matter of time.

What’s standing in the way of progress?

But there’s a problem: Most American physicians don’t work within systems that prioritize the well-being of physicians or even patients. The prevalence of fee-for-service payment models often leads to incentive systems that prioritize patient volume over patient value in order to maximize hospital revenue and physician compensation.

Unfortunately, trying to save time with an AI tool when taking visit notes could actually translate into higher patient loads. If volume-driven financial incentives remain unchanged, there may be requests to use that freed up time to see even more patients. Therefore, AI could perpetuate a cycle of higher workloads, leading to burnout, clinical errors, and poorer patient outcomes. History has shown that payers are reluctant to self-regulate when changing their profit-driven models, necessitating intervention from other parts of the healthcare system.

In a country that primarily uses the fee-for-service model and struggles with cost containment, there is also the risk of questionable practices being built into AI algorithms. For example, upcoding, when providers inflate the severity of a patient’s condition to obtain a higher fee, could become institutionalized in AI systems trained on past behaviors. This could lead to AI writers being biased toward upcoding, reflecting volume-based revenue maximization incentives.

What needs to be done?

AI’s promise to improve provider well-being and align the healthcare system with patient value is unlikely to materialize without changing payment models and incentives to promote value over volume. Achieving such a seismic shift in incentive structures will take time, energy, and concerted effort.

As the technology advances, so too must the policies and systems that define and regulate its use, including updating the associated incentive mechanisms. For example, European countries, notably the UK’s National Health Service, have gradually introduced bundled payments to prioritise patient outcomes and best practice in their reimbursement structures. Other countries may wish to follow suit.

“AI automation of clinical documentation offers significant benefits for physician productivity and well-being.”

Critics of the push from volume to value often argue that de-emphasizing patient volume can impact access to care. Therefore, it is imperative that clinicians are expected to strike a healthy balance between volume and value. Without proper boundaries, a short-sighted response to volume maximization, even when justified as a much-needed expansion of access, is likely to come at the cost of increasing burnout, leading to a long-term physician shortage.

AI automation of clinical documentation offers significant benefits for clinician productivity and well-being. However, to ensure that AI does more good than harm, additional changes to financial incentive systems must accompany its implementation. We cannot afford to wait for payers to take the first step. Healthcare organizations must lead the way by updating technologies and internal incentive structures to protect providers and patients now and in the future.

Susanna Gallani is the Tai Family Associate Professor of Business Administration in the Accounting and Management Unit at Harvard Business School, where she studies healthcare performance management systems and value-based healthcare implementation strategies. Lidia Moura is a clinical neurologist and director of neurological population health at Massachusetts General Hospital (MGH) and an associate professor of neurology at Harvard Medical School. She is also an OpEd Project Fellow and director of the Center for Value-based Healthcare and Sciences at MGH. Katie Sonnefeldt is a researcher at HBS and supports research into performance management in healthcare and health equity.

You might also like:

Feedback or ideas to share? Email the Working Knowledge team at [email protected].

Image: Image by HBSWK with images from AdobeStock/沈军贡, AdobeStock/Only PNG, and AdobeStock/MCGORIE