For decades, traditional pathology approaches have played an integral role in the diagnosis, classification, and assessment of disease. Technologic advances and a shift toward precision medicine, however, have resulted in the development of digital approaches, such as artificial intelligence (AI), to improve the use of data in health care in fields including patient stratification, diagnostic assays, and personalized treatment regimens.1
Although the potential of AI in health care is not yet fully understood, researchers agree that combining machine learning with physician diagnoses can greatly improve diagnostic confidence and enhance system performance.2 The rise of AI in medicine is promising and suggestive of more efficient and effective patient care.
AI improves diagnosis and diagnostics, has prognostic and predictive abilities, and generates information. It’s use in clinical practice is especially prevalent in the fields of radiology and oncology.3
Some of the foremost advances in cancer biology are in screening, targeted and immune therapies, big data, and computational methodologies, which are driving the way toward personalized care.4 For example, the development of automated cancer detection software using routine radiology or pathology scans, such as diagnosing incident lung cancer from a computed tomography (CT) scan, can be an important tool in clinical prediction.5
In addition, AI-based vs traditional tissue-based biomarkers are beginning to play an important role in the early detection and diagnosis of cancer, highlighting the robustness of this technology.6
“
Although the potential of AI in health care is not yet fully understood, researchers agree that combining machine learning with physician diagnoses can greatly improve diagnostic confidence and enhance system performance.
In discussing the incorporation of AI into neurology, Ravi B Parikh, MD, MPP, assistant professor of medicine and assistant professor of medical ethics and health policy at the Perelman School of Medicine at the University of Pennsylvania, referred to the US Food and Drug Administration (FDA)-approved reimbursable AI-based technologies that can help detect large-vessel strokes, the CT scans of which can inform radiologists more adequately than routine scans.
Overall, the FDA has approved 22 unique machine learning-enabled technologies, which perform time-saving tasks, such as segmentation, to assist in the management of stroke and intracranial hemorrhage.7
Some researchers have called generative AI (eg, ChatGPT) potentially “transformative” in that it has the potential to transform health care education, research, and clinical practice.8
Dr Parikh noted that in real-world clinic settings, “[This technology] may be useful in that it generates text to communicate with a patient or an automated, personalized response.” Further, Dr Parikh explained, generative AI can be used to “generate a clinical note, documentation, or prior authorization request.”
“This is really promising because it offers the possibility of lowering workload for doctors, nurse practitioners, and health care workers in a more robust way, allowing physicians to spend more time with their patients,” Dr Parikh added.
Although the backend use of AI (ie, managing documents or quality metrics) is popular in some clinics, many clinicians outside of specialties such as dermatology, radiology, and pathology do not frequently employ this technology.
The Use of AI in Academic Institutions vs Clinical Practice
While data protection and privacy laws in the United States impose strict regulations for AI systems that process patient data, the use of AI algorithms has the potential to alleviate administrative burdens at clinical practices, such as improving the speed of billing and claims processing.9 This, in turn, can play a role in patient satisfaction and management.
In a 2020 paper published in the Journal of the American College of Radiology, Nina Kottler, MD, described how AI in the context of radiology has primarily been used by academic institutions, not general practices. Dr Kottler herself urges the incorporation of AI by radiologists in private practice, however, to address logistic challenges and improve efficiency.10
Dr Parikh clarified the differences in perspective regarding AI use, explaining, “In academic organizations, we are testing AI-based diagnostics or digital therapeutic tools that we have developed in our labs or are working on with another industry partner to help validate. In contrast, much of the use of AI in the clinic setting is intended to improve workflow with less focus directed toward the investigational aspect and clinical trials.”
Overall, both academic- and community-based systems have considered or adopted AI, although the focus differs.
How Insurance May Impact AI Use
“Recent evidence has demonstrated that many of the FDA-approved reimbursable AI tools tend to be dominated by larger academic health systems, rather than smaller rural hospitals or clinics,” Dr Parikh said. He continued, “This may be because they often involve expensive contracts or lack requisite data to make AI work in smaller centers.”11
Another issue that AI may be able to address is clinical staff training, which Dr Parikh deemed “necessary,” in that its use can “differentiate the quality of medical institutions.”
“We need to think of solutions that do not have to be FDA-approved or have a billable code. Rather, we should be thinking about solutions that can enhance the ability to process patient visits, improve speed, gather information from electronic medical records, and improve documentation.”
Data infrastructure use should be capitalized to serve a greater number of patients and increase health care access.
Improving Efficiencies, Processes, and Patient Care With AI
Primary medicine and telemedicine have incorporated the use of AI-driven diagnostic tools that are able to assist health care professionals in making accurate diagnoses based on patient-reported symptoms, which has demonstrated potential in rendering medical care availability in remote and underserved areas more accessible.3
“Because of the COVID-19 pandemic, there is a huge investment, both private and in academia, in technologies that can ensure greater quality of remote monitoring. A large number of digital health companies, vendors, and other types of solutions are being developed for this purpose,” Dr Parikh said.
Three key areas in which AI may be useful in the setting of patient care and monitoring include AI-powered wearables for continuous monitoring, virtual nursing assistants, and telemedicine and remote patient engagement.9
Although clinicians can monitor heart failure and physical activity using remote management, device literacy among patients and providers remains a barrier, Dr Parikh noted.
The Benefit of AI for Health Care Systems
AI and machine learning can benefit health care systems in many ways, offering more precise immune therapies; improving diagnostic, prognostic, and predictive clinical decision-making; and ultimately, yielding desirable patient outcomes.
“Within the next 5 to 10 years, we are going to see AI become integrated into health care to the same extent as electronic health records were once integrated,” Dr Parikh concluded.
link