Dear Doctors: According to a 2024 study, approximately 200,000 deaths occur each year in the United States due to medical errors. Could artificial intelligence play a role in reducing these incidents? I’ve learned it may assist with issues like ventilator management, omissions in patient instructions, and the detection of diagnostic mistakes.
Dear Reader: Your inquiry about the use of artificial intelligence in healthcare opens up an important discussion that evokes both cautious optimism and legitimate concerns. While the conversation around AI as a medical tool may seem contemporary, its origins date back to the mid-20th century. What began as a theoretical concept in the 1950s evolved by the turn of the century into various algorithms and pattern-recognition models aimed at aiding areas such as imaging analysis, diagnosis, and certain clinical practices. It wasn’t until the advances in data processing and computing power during the 2010s that these deep-learning techniques started to transition from theory to application.
Before delving into the potential benefits of AI in healthcare, let’s clarify the statistics you’ve mentioned. Although there is a strong desire for precise figures regarding deaths from medical errors, estimates can vary tremendously based on how these errors are defined. Some studies focus strictly on point-of-care mistakes, whereas others include systemic failures like inadequate care coordination, limited access to necessary services, issues with insurance, and poor follow-up care. The analysis that supports the 200,000 figure takes a broader perspective, encompassing more than just point-of-care errors.
Advocates for AI suggest that these tools can substantially enhance medical care. For instance, pattern-recognition models have shown effectiveness in analyzing medical imaging scans, often achieving improvements in both speed and accuracy. A study conducted at UCLA discovered that AI models could detect pancreatic tumors in CT scans months—sometimes even years—before a human observer could recognize them. Another prominent application is in screening for diabetic retinopathy, a condition where elevated blood sugar levels harm retinal blood vessels. AI systems have also emerged as promising early warning indicators for sepsis, help in identifying heart arrhythmias in ECGs, and assist in reading mammograms during breast cancer screenings.
Despite the promising potential of these advanced tools, many healthcare professionals urge a cautious approach in their adoption. Medicine is inherently a personal field, where context plays a critical role. Consequently, most experts argue that artificial intelligence should serve as an assistant rather than be treated as a sole authority.
When implemented thoughtfully and under the right supervision, AI systems hold the promise of improving patient care. They can analyze vast datasets rapidly, uncover patterns indicative of early-stage diseases, enhance imaging accuracy, detect subtle changes in vital signs that may precede complications, identify lapses in follow-up care, and highlight coordination failures before they result in harm. As AI technologies continue to advance, retaining the human element in the process will be essential for their success and broader acceptance.
(Send your questions to [email protected], or write: Ask the Doctors, c/o UCLA Health Sciences Media Relations, 10960 Wilshire Blvd., Suite 1955, Los Angeles, CA, 90024. Due to the high volume of correspondence, individual replies cannot be guaranteed.)
