AI in Healthcare: Promise or Illusion for Patients?

AI Takes Root in Hospitals

Artificial intelligence (AI) has become an essential player in the healthcare sector, particularly in hospitals where it is used to assist doctors with note-taking and analyzing patient records. These AI-based tools are also employed to interpret the results of medical exams and X-rays, providing valuable support to healthcare professionals.

However, despite numerous studies suggesting that these tools can deliver accurate results, the question of whether they actually improve patient health remains unanswered. Jenna Wiens, a computer scientist at the University of Michigan, and Anna Goldenberg from the University of Toronto, recently addressed this topic in an article published in Nature Medicine. They emphasize that the real impact of AI on patient health outcomes is not yet well understood.

Wiens has spent the first decade of her career trying to introduce this technology to clinicians. She has observed a notable shift in recent years, where a "switch seems to have flipped" among healthcare providers. They are now showing an increasing interest in these technologies and are rapidly deploying them in their practices. However, she notes that few providers take the time to rigorously evaluate the effectiveness of these tools.

Enthusiasm for AI Scribes

Ambient AI tools, also known as "AI scribes," which transcribe and summarize conversations between doctors and patients, are already widely adopted. A staff member at a major medical center in New York stated that doctors are thrilled with this technology, as it allows them to focus fully on their patients during consultations and reduce their administrative workload. While anecdotes and preliminary studies suggest that they reduce physician burnout, their impact on clinical decision-making remains unclear.

Wiens points out that researchers have often assessed provider and patient satisfaction, but not sufficiently how these tools influence clinical decision-making. She insists that we simply do not know how these technologies affect patient health outcomes.

Promising Tools but Uncertain Results

Other AI-based technologies are used to predict patient health trajectories or recommend treatments. However, even an accurate tool does not necessarily improve health outcomes. The impact of these tools can vary depending on hospitals and physicians, and may depend on clinical workflows.

Some research on the use of AI in education suggests that such tools can influence how people cognitively process information. For example, AI scribes could affect how doctors process patient information, which could have unforeseen consequences. Wiens stresses the need to explore these questions.

Rapid Adoption but Lack of Evaluation

A study conducted by Paige Nong from the University of Minnesota found that about 65% of American hospitals use AI-assisted predictive tools. However, only two-thirds of these hospitals have evaluated their accuracy, and even fewer have examined their biases. The number of hospitals using these tools has likely increased since then, according to Wiens.

She calls for rigorous evaluation to understand their true impact on healthcare. While she believes in the potential of AI to improve clinical care, Wiens emphasizes that she does not wish to halt the adoption of AI tools in healthcare. She simply wants more information on the impact of these tools on patients. She envisions a future where AI and traditional methods coexist in healthcare, finding a balance between the two approaches.