Abstract
Health care costs now comprise nearly one-fifth of the United States’ gross domestic product, with the last 25 years marked by rising administrative costs, a lack of labor productivity growth, and rising patient and physician dissatisfaction. Policy experts have responded with a series of reforms that have – ironically - increased patient and physician administrative burden with little meaningful effect on cost and quality. Artificial intelligence (AI), a topic of great consternation, can serve as the “wheat thresher” for health care delivery, empowering and freeing both patients and physicians by decreasing administrative burden and improving labor productivity. In this Viewpoint, we discuss three principal areas where AI poses an unprecedented opportunity to reduce cost, improve care, and markedly enhance the patient and physician experience: (1) automation of administrative process, (2) augmentation of clinical practice, and (3) automation of elements of clinical practice.
Explore related subjects
Discover the latest articles, news and stories from top researchers in related subjects.Avoid common mistakes on your manuscript.
Applications of Artificial Intelligence in Health Care Delivery
Health care costs now comprise nearly one-fifth of the United States’ gross domestic products, with the last 25 years marked by rising administrative costs, a lack of labor productivity growth, and rising patient and physician dissatisfaction [1]. Policy experts have responded with a series of reforms that have – ironically - increased patient and physician administrative burden with little meaningful effect on cost and quality. Some programs, such as Accountable Care Organizations have driven consolidation [2] while other long-touted interventions such as comprehensive primary care showed no systematic impact on cost or quality [3]. With decades of reliance on micro policy adjustments not yielding the desired outcomes, now is the time to think differently. Artificial intelligence (AI), a topic of great consternation, can serve as the “wheat thresher” for health care delivery, empowering and freeing both patients and physicians. In this Viewpoint, we discuss three principal areas where AI poses an unprecedented opportunity to reduce cost, improve care, and markedly enhance the patient and physician experience: (1) automation of the mundane, (2) augmentation of human-driven clinical practice, and (3) automation of elements of clinical practice.
Automation of the Mundane
Nearly two-thirds of physicians suffer from burnout, with many citing electronic health record documentation and inbox messages, prior authorization, and other routine administrative tasks [4]. This collective experience is mirrored in empirical research: a recent time motion study of internal medicine residents demonstrated that only 13% of the average day is spent in face-to-face contact with patients [5] while other work demonstrates that a typical primary care physicians spend 6 h per clinic day writing patient notes [6].
AI offers the opportunity to restore joy in clinical practice through automation of the mundane – or manual, human capital driven tasks. With companies such as Augmedix, DeepScribe and Amazon presenting early attempts at automating clinical notetaking, eventually the burden of recordkeeping should shift from the clinician to technology. Ultimately, clinicians should be able to review and sign AI-generated notes as opposed to spending their limited time during and after patient encounters documenting visits. Early evidence suggests that automation of simple responses to patient messages may not only decrease burden, but also increase satisfaction as AI may be perceived in some circumstances as having greater empathy [7]. Finally, the arduous painful tasks of coding and billing could be similarly automated, saving clinicians from struggling with over 70,000 unique diagnosis codes.
A process that could arguably not be worse for all those involved, prior authorization offers another obvious opportunity for improvement using AI. The average oncology practice has over 6 full-time staff to handle prior authorizations, [8] while physicians generally report filling out an average of 37 prior authorization forms per week with 75% reporting this as a “high” or “extremely high” burden [9]. Using AI to automate the submission of data as the first level of review for some prior authorization processes based upon clear and transparent guidelines could eliminate the need for physician involvement. Eventually, AI could help facilitate near real-time adjudication of prior authorization requests during a clinical visit, freeing up both clinicians and health plan staff to engage in more meaningful, complex reviews.
Augmentation of human-driven Clinical Practice
AI can transform how clinicians use information and make decisions, driving increasingly personalized care. AI technology [10] drives dynamic displays in Teslas, alerting drivers of potential harm thus facilitating quick decision making and real-time interventions while BlueCruise in Fords facilitates hands-free highway driving. Medical care is no different, requiring split second decisions in an environment of countless informational inputs. AI-driven technology can offer meaningful clinical decision support, enabling clinicians in acute care environments such as intensivists, anesthesiologists, and nurses to manage patients more efficiently and effectively by highlighting concerning trends, gaps in care, or suggesting actions that would otherwise be omitted or delayed.
Technological innovations like magnetic resonance imaging and genetic sequencing have provided clinicians with additional diagnostic tools to customize and personalize care, with AI and automation offering a similar opportunity to upskill clinicians while improving productivity. Through algorithmic pattern recognition, AI applications may more accurately read CT scans, [11] mammography, [12] or pathology slides, [13] catching diagnoses that might have been missed and enhancing the capability of the supervising clinicians. We expect that these tools will, for example, enable pathologists to spend time reviewing suggested diagnoses of routine cases and devote more time to selecting assays, improving overall lab performance, and addressing complex cases. For some cancers, AI tools will eventually be able to suggest treatment options for oncologists to review and then discuss with their patients. This will enhance the consistency of care and accelerate the adoption of newly generated scientific evidence which today often takes decades to come into routine clinical practice.
Automation of Elements of Clinical Practice
While fully autonomous AI-driven medical care is likely decades away, automation of specific narrow tasks is already underway. Technology like IDx-DR [14] (Digital Diagnostics Inc, Iowa) can streamline screening for diabetic retinopathy without physician interpretation and with such precision that the company holds liability insurance. In resource-limited settings, point-of-care digital cytology with AI is being used for cervical cancer screening [15]. After being trained on 30,493 electroencephalograms, an AI model was able to achieve diagnostic performance similar to human experts [16].
These initial narrow and specific autonomous AI applications offer insight into the future of innovation. For example, as radiological AI applications reach the market, images will be read increasingly by machine learning algorithms with diagnoses being determined with a pre-specified degree of certainty, enabling radiologists to review cases with a low degree of certainty. The transition to automation will be gradual with human practitioners remaining relevant and in the loop supervising the algorithms while finding new avenues to care for patients, requiring physicians to develop the skills to oversee and manage the output and activities of AI-enabled technologies.
Policy Opportunities
With technological advances comes the fear of the unknown. The wheat thresher was alleged to destroy the agricultural industry and put farmers out of business. In reality, it provided much needed efficiencies in agriculture and allowed farmers to scale and meet increasing demand. AI will play a similar long-term role in that it is an innovative and flexible answer to one of the largest issues facing the healthcare sector: the lack of labor productivity growth coupled with rising administrative costs. Like technological change in order industries, AI may drive some short term costs through implementation, with eventual, significant long-term cost savings.
Important and answerable policy questions remain. Medical licensure for independently practicing products should be avoided. Instead, noting that current U.S. Food and Drug Administration regulatory pathways are ill-suited for complex and rapidly evolving software products, policymakers should consider the creation of a series of voluntary, alternative regulatory pathways fit for software as a medical device, inclusive of AI and software-driven medical devices. Liability remains a question and should be placed with the parties best positioned to mitigate the risk. Policymakers and courts will sort out the gray line between product, clinician, and consumer liability for AI-augmented and AI-driven medical care.
Regardless, the benefits of AI to patients and clinicians must not be overlooked. AI has the potential to transform care delivery augmenting physicians and the healthcare team, leading to the provision of superior care and substantial costs savings, the stated policy goals of many previously failed policy interventions. If the profession panics and fails to capitalize on the positive potential of AI, it will neglect an important opportunity to revolutionize healthcare.
The views of this paper do not necessarily represent those of the Medicare Payment Advisory Commission or the American Medical Association.
Data Availability
No data were used in this study as it is a perspective paper.
References
Private Community Hospitals Labor Productivity: U.S. Bureau of Labor Statistics. Accessed June 29, 2023. https://www.bls.gov/productivity/highlights/hospitals-labor-productivity.htm
Kanter GP, Polsky D, Werner RM. Changes In Physician Consolidation With The Spread Of Accountable Care Organizations. Health Aff (Millwood). 2019;38(11):1936–1943. doi:https://doi.org/10.1377/hlthaff.2018.05415
Comprehensive Primary Care Initiative | CMS Innovation Center. Accessed July 25, 2023. https://innovation.cms.gov/innovation-models/comprehensive-primary-care-initiative
Almost Two-Thirds of U.S. Doctors, Nurses Feel Burnt Out at Work: Poll. US News & World Report. Accessed October 26, 2023. https://www.usnews.com/news/health-news/articles/2023-02-23/almost-two-thirds-of-u-s-doctors-nurses-feel-burnt-out-at-work-poll
Chaiyachati KH, Shea JA, Asch DA, et al. Assessment of Inpatient Time Allocation Among First-Year Internal Medicine Residents Using Time-Motion Observations. JAMA Intern Med. 2019;179(6):760–767. doi:https://doi.org/10.1001/jamainternmed.2019.0095
Arndt BG, Beasley JW, Watkinson MD, et al. Tethered to the EHR: Primary Care Physician Workload Assessment Using EHR Event Log Data and Time-Motion Observations. Ann Fam Med. 2017;15(5):419–426. doi:https://doi.org/10.1370/afm.2121
Sharma A, Lin IW, Miner AS, Atkins DC, Althoff T. Human–AI collaboration enables more empathic conversations in text-based peer-to-peer mental health support. Nat Mach Intell. 2023;5(1):46–57. doi:https://doi.org/10.1038/s42256-022-00593-2
Lin NU, Bichkoff H, Hassett MJ. Increasing Burden of Prior Authorizations in the Delivery of Oncology Care in the United States. J Oncol Pract. 2018;14(9):525–528. doi:https://doi.org/10.1200/JOP.18.00428
Survey quantifies time burdens of prior authorization. American Medical Association. Published January 30, 2017. Accessed July 10, 2023. https://www.ama-assn.org/practice-management/prior-authorization/survey-quantifies-time-burdens-prior-authorization
Model S Owner’s Manual. Tesla. Accessed July 10, 2023. https://www.tesla.com/ownersmanual/models/en_us/GUID-29A7E205-A689-41D1-B69C-3AE821CB70E7.html
Ziegelmayer S, Reischl S, Havrda H, et al. Development and Validation of a Deep Learning Algorithm to Differentiate Colon Carcinoma From Acute Diverticulitis in Computed Tomography Images. JAMA Netw Open. 2023;6(1):e2253370. doi:https://doi.org/10.1001/jamanetworkopen.2022.53370
Lång K, Josefsson V, Larsson AM, et al. Artificial intelligence-supported screen reading versus standard double reading in the Mammography Screening with Artificial Intelligence trial (MASAI): a clinical safety analysis of a randomised, controlled, non-inferiority, single-blinded, screening accuracy study. Lancet Oncol. 2023;24(8):936–944. doi:https://doi.org/10.1016/S1470-2045(23)00298-X
Steiner DF, Nagpal K, Sayres R, et al. Evaluation of the Use of Combined Artificial Intelligence and Pathologist Assessment to Review and Grade Prostate Biopsies. JAMA Netw Open. 2020;3(11):e2023267. doi:https://doi.org/10.1001/jamanetworkopen.2020.23267
Abràmoff MD, Lavin PT, Birch M, Shah N, Folk JC. Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy in primary care offices. Npj Digit Med. 2018;1(1):1–8. doi:https://doi.org/10.1038/s41746-018-0040-6
Holmström O, Linder N, Kaingu H, et al. Point-of-Care Digital Cytology With Artificial Intelligence for Cervical Cancer Screening in a Resource-Limited Setting. JAMA Netw Open. 2021;4(3):e211740. doi:https://doi.org/10.1001/jamanetworkopen.2021.1740
Tveit J, Aurlien H, Plis S, et al. Automated Interpretation of Clinical Electroencephalograms Using Artificial Intelligence. JAMA Neurol. Published online June 20, 2023. doi:https://doi.org/10.1001/jamaneurol.2023.1645
Funding
None.
Author information
Authors and Affiliations
Contributions
J.S. wrote the initial draft. J.M.E. and B.J.M. contributed to critical revision and supervision. All authors reviewed the manuscript.
Corresponding author
Ethics declarations
Competing Interests
Mr. Spear has nothing to report. Dr. Ehrenfeld reports serving as the president of the American Medical Association and as a member of the American Medical Association Board of Trustees. Dr. Miller previously served as a Senior Policy Fellow for Health IT at the Federal Communications Commission in 2015 and as a Medical Officer at the U.S. Food & Drug Administration from 2016 to 2017. He reports serving as a Commissioner of the Medicare Payment Advisory Commission, and receiving fees outside the related work from the Federal Trade Commission, the Centers for Medicare & Medicaid Services, the Maryland Neurosurgical Society, the California Association of Neurosurgeons, and the Digestive Health Physicians Association.
Ethical Approval
This paper does not involve human and/or animal studies.
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Spear, J., Ehrenfeld, J.M. & Miller, B.J. Applications of Artificial Intelligence in Health Care Delivery. J Med Syst 47, 121 (2023). https://doi.org/10.1007/s10916-023-02018-y
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10916-023-02018-y