The Transformative Potential of AI in Health Care Delivery & Administration

This is an excerpt from a recent white paper written with Dennis Chornenky, Chief AI Advisor, UC Davis Health  and David Lubarsky, Vice Chancellor of Human Health Sciences and Chief Executive Officer, UC Davis Health. To read the full white paper, click here.

An effective AI governance model will allow health systems to take advantage of innovative technologies, while mitigating risks to patients. There have been a lot of attention-grabbing headlines about AI replacing physicians, but the more likely scenario is that it functions as an aid—amplifying the capacity of physicians and care teams to serve more patients in more tailored ways. The infographic below highlights a few key ways this may happen.

	Address Information Asymmetry and enable greater health literacy and patient self-advocacy by providing rapid plain-language information to patients on their condition as well as access to second opinions. Bridge Language and Cultural Barriers by making translation tools more readily accessible at the point of care. Improve Access to Care in Under-Served Communities: AI chatbots, diagnostic algorithms and remote patient monitoring tools can be used to augment access to primary and specialty care in geographies that experience provider shortages.
	Enhance Ability to Predict and Prevent Disease: AI tools can analyze vast amounts of medical data, enabling more accurate predictions of disease or readmission based on risk factors that may not be readily apparent to human observers. Improve Patient Engagement by providing rapid access to tailored information on risk factors and strategies for patients to prevent escalation—in language that is easily accessible. Allow for More Rapid, Data-Driven Public Health Needs Assessments and Monitoring to Guide Policy Decisions and Resource Allocation: AI tools can be useful in predicting outbreaks and forecasting demand for public health services.
	Enhance Diagnostic Accuracy: Misdiagnosis is a significant issue in health care—the prevalence ranges from 5% to 20%, depending on the disease or condition.1 Misdiagnosis can have serious consequences, leading to delayed or inappropriate treatments, unnecessary procedures, increased health care costs and patient harm. Studies have shown that AI tools can exceed the diagnostic accuracy of human doctors for certain conditions. With AI, physicians have the benefit of a trained “co-pilot,” and patients can benefit from timelier and data-backed diagnoses. Increase Access to Specialty Diagnostics: AI algorithms can rapidly assess diagnostic images and biomarkers for disease risk factors, allowing for greater throughput of diagnostic screenings and freeing up clinician time to focus on true positives and complex cases. People in rural, remote or otherwise under-served areas could access specialist-level review of their diagnostic tests without having to travel, thanks to AI. One health system is developing an AI-driven diagnostic tool that analyzes voice biomarkers to predict coronary artery disease.2
	Provide Efficient and Personalized Treatment Planning: AI-driven tools have also been able to speed up the treatment planning process—for example, reducing radiation therapy treatment planning from days to mere minutes.3 Enable to Access to Bedside Decision Support: The authors of “Foundation models for generalist medical artificial intelligence” describe a potential future where AI can not only provide an early warning for an adverse event, but also provide the rationale and data points to back its assessment, and recommend a course of intervention, driven by the latest clinical evidence.4 Augment Procedures: Today, surgical robots already provide remote case proctoring. In the future, surgical robots may be able to annotate video streams of procedures in real time or raise alerts when procedure steps are omitted.5 Allow for Enhanced Remote Patient Monitoring: Companies continue to develop the capability to monitor patients with chronic conditions remotely, through the tracking of biomarker data provided through wearable technologies or phone-based image capture. These technologies can trigger patient or clinician intervention to address concerns in real-time. Provide Access to Treatment Information through Patient-Facing Chatbots: LLMs enable the creation of chatbots that can respond to patient queries and provide guidance. There are myriad liability issues to work through, but at a basic level, robust and well-tested models can be an important source of information for patients to learn about treatment options.
	Provide Note Transcription Tools: AI can be used in note transcription tools to automatically convert spoken medical information into written text, potentially enhancing efficiency and accuracy in medical documentation. Facilitate Demand Forecasting: Demand forecasting tools employ predictive modeling to estimate future patient demand, enabling health care organizations to optimize staff scheduling. Support Billing and Coding: AI can be employed to automate and streamline medical billing and coding. With AI-powered automatic coding tools, medical billing professionals can spend time checking coding rather than doing the data entry manually. Revenue cycle management is a large expense for health systems.
	Identify Patients for Clinical Trials: AI can be used to cull EHR data on patients to identify those that may be eligible for clinical trials, potentially increasing the number of historically underrepresented patients in clinical trials. Support Drug Discovery and Development: AI can be used to support drug target identification, make efficacy predictions, individualize drug response predictions and monitor drug safety and adverse events, among other things.6

_{¹ Newman-Toker, D. E., Wang, Z., Zhu, Y., Nassery, N., Saber Tehrani, A. S., Schaffer, A. C., Yu-Moe, C. W., Clemens, G. D., Fanai, M., & Siegal, D. (2021), Rate of diagnostic errors and serious misdiagnosis-related harms for major vascular events, infections, and cancers: Toward a national incidence estimate using the “Big Three.” Diagnosis (Berlin, Germany). https://pubmed.ncbi.nlm.nih.gov/32412440/}

_{² Malloy, T. (2022, April 26). AI uses voice biomarkers to predict coronary artery disease - mayo clinic news network. Mayo Clinic. https://newsnetwork.mayoclinic.org/discussion/ai-uses-voice-biomarkers-to-predict-coronary-artery-disease/}

_{³ Meskó, B., Görög, M. (2020) A short guide for medical professionals in the era of artificial intelligence. Npj Digital Medicine 3, 126. https://doi.org/10.1038/s41746-020-00333-z}

_{⁴ Moor, M., Banerjee, O., Abad, Z. S., Krumholz, H. M., Leskovec, J., Topol, E. J., & Rajpurkar, P. (2023). Foundation models for generalist medical artificial intelligence. National Library of Medicine, 616(7956), 259–265. https://doi.org/10.1038/s41586-023-05881-4}

_{⁵ Ibid.}

_{⁶ Singh, S., Kumar, R., Payra, S., & Singh, S. K. (2023). Artificial Intelligence and machine learning in pharmacological research: Bridging the gap between Data and Drug Discovery. Cureus. https://doi.org/10.7759/cureus.44359}