Growing Evidence Shows Importance of AI for Health Care
Artificial intelligence (AI) will likely have an impact on virtually every sector of the economy, but one of its greatest benefits will come from deploying it in health care. While some researchers have anticipated these benefits for years, there is an emerging body of scientific evidence that these capabilities may soon be ready to move from theory to practice. Given this progress, policymakers should be looking closely at how to integrate AI into the health-care system safely and effectively.
According to a 2019 literature review in Health Policy and Technology, AI supports health care in four ways: assessment of disease and treatment, management or alleviation of complications, patient-care assistance during a treatment or procedure, and research aimed at discovery or treatment of disease. Such interventions have the potential to reduce wait times, by freeing up doctors’ availability, while also helping drive innovations that lower costs.
AI may perform better than doctors, particularly in cases where the model can identify patterns in medical images that human eyes cannot discern. A 2019 meta-study found that “current AI development has a diagnostic performance that is comparable with medical experts, especially in image recognition-related fields.” For example, AI performs better than dermatologists on average for applications like melanoma detection. These benefits may extend beyond image detection, as AI increasingly appears to be a useful tool for patients, offering an immediately responsive first line of defense for medical questions. Or, as researchers in the American Association for Physician Leadership wrote in 2019,
IBM’s Watson diagnoses heart disease better than cardiologists do. Chatbots dispense medical advice for the United Kingdom’s National Health Service in lieu of nurses. Smartphone apps now detect skin cancer with expert accuracy. Algorithms identify eye diseases just as well as specialized physicians.
AI may also be useful as a complement to human examination and diagnosis. Diagnostic error underlies about 10 percent of adverse events occurring in hospital practices. In many cases, misdiagnosis can emerge due to medical specialization, or else the lack of specialized knowledge held by general practitioners. In both cases, misdiagnosis is a problem of information. And by helping address this issue, AI may support greater health-care efficiency, reducing the costs—human and institutional—of misdiagnosis.
To be clear, not every study about AI’s benefits in health care holds up to scrutiny. A 2020 paper in the British Medical Journal criticized some research findings about AI diagnostic success for poor research designs and a lack of randomized trials. However, the initial results are still promising.
Indeed, many doctors are optimistic about AI’s health-care benefits. Polling of physicians by the American Medical Association in 2023 found that a majority of physicians thought there would be advantages to using AI in health care, including diagnostic ability (cited by 72 percent of respondents), workflow efficiency (69 percent), and clinical outcomes (61 percent). Physician enthusiasm was highest for AI tools that help reduce administrative burdens, including documentation (54 percent) and prior authorization (48 percent). All of this could free up doctors’ time, allowing them to meet with more patients and potentially reducing costs. Meanwhile, health-care-led large language models have the potential to support as a first line of defense, offering basic diagnosis for nonurgent concerns, and further reducing doctor visits.
Evidence about the benefits of AI in health care is mounting. As policymakers consider which sectors should prioritize AI adoption, health care should be at the top of the list. The deployment of AI into health care is not inevitable. Adoption lags, in particular, present a challenge, even as the medical and pharmaceutical industries invest heavily in AI capital. Indeed, A 2022 systemic review of scholarship found that, while a majority of physicians and medical students were aware of AI’s applications, only between 10 to 30 percent had used AI. These findings are corroborated by the American Medical Association’s polling: under two-thirds of survey respondents reported using AI.
Yet interest in AI in health care is growing. Governments themselves have recognized the capacity of AI to help create capacity within the health-care sector. The United Kingdom, for instance, is looking to address the current deficiencies of its National Health Services (NHS) through investments in AI. In 2023 the UK committed £21 million to deploy AI to support swifter NHS diagnosis of patient conditions, including cancers strokes, and heart conditions. And in a more ambitious project, they devoted £100 million to support the development of precision treatments for dementia, in addition to diagnostic support for Britons struggling with their mental health.
AI is not likely to be a panacea to all health-care system ailments. Yet it is increasingly viewed as a critical investment for countries looking to improve the delivery and scale of quality health-care support. And as demographic change and aging populations in many Western countries entail higher relative health-care burdens, AI’s support in diagnosis, drug development, and health-care operations may serve as a much-needed remedy.
