wOur physicians are exceptional, tireless, and often accurate. Yet, they are human. Increasingly, they face exhaustion, working extended hours under tremendous stress, and frequently with insufficient resources. Improved conditions—like more personnel and better systems—can certainly help. However, even the best-funded clinics with the most committed professionals can lack essential standards. Doctors, like all of us, often operate with a mindset reminiscent of the Stone Age. Despite extensive training, the human brain struggles to cope with the speed, pressure, and intricacies of contemporary healthcare.
Since patient care is the principal aim of medicine, what or who can best facilitate this? While AI can evoke skepticism, research increasingly illustrates how it can resolve some of the most enduring problems, including misdiagnosis, errors, and disparate access to care, and help rectify overlooked failures.
As patients, each of us will likely encounter at least one diagnostic error during our lifetime. In the UK, conservative estimates indicate that 5% of primary care visits result in an inability to diagnose correctly, putting millions at risk. In the US, diagnostic errors can lead to death or lasting harm, affecting 800,000 individuals each year. The risk of misdiagnosis is amplified for the one in ten people globally with rare diseases.
Modern medicine prides itself on being evidence-based, yet doctors don’t always adhere to what the evidence suggests. Studies reveal that evidence-based treatments are dispensed only about half the time for adults in the US. Furthermore, your doctor might not concur with the diagnosis either. In one study, reviewers providing second opinions on over 12,000 radiology images disagreed with the original assessment in roughly one-third of cases, leading to nearly 20% of treatment changes. As workloads increase, quality continues to decline, resulting in inappropriate antibiotic prescriptions and falling cancer screening rates.
While this may be surprising, there is a comprehensible reason for these errors. From another perspective, it’s remarkable that doctors often get it right. The human aspects—distraction, multitasking, even our circadian rhythms—play a significant role. However, burnout, depression, and cognitive aging affect more than just physicians; they raise the likelihood of clinical mistakes.
Additionally, medical knowledge advances more rapidly than any doctor can keep up with. By graduation, many medical students’ knowledge is already outdated, with an average of 22 hours required for a study to influence clinical practice. With a new biomedical article published every 39 seconds, even reviewing just the summaries demands a similar time investment. There are over 7,000 rare diseases, with 250 more identified each year.
In contrast, AI processes medical data at breakneck speeds, operating 24/7 without breaks. While doctors may waver, AI remains consistent. Although these tools can also make mistakes, it’s important not to underestimate the capabilities of current models. They outperform human doctors in clinical reasoning related to complex medical conditions.
AI’s superpower lies in identifying patterns often overlooked by humans, and these tools have proven surprisingly adept at recognizing rare diseases—often surpassing doctors. For instance, in a 2023 study, researchers tasked ChatGPT-4 with diagnosing 50 clinical cases, including 10 involving rare conditions. It accurately resolved all common cases by the second suggestion and achieved a 90% success rate for rare conditions by the eighth guess. Patients and their families are increasingly aware of these advantages. One child, Alex, consulted 17 doctors over three years for chronic pain, unable to find answers until his mother turned to ChatGPT, which suggested a rare condition known as tethered cord syndrome. The doctor confirmed this diagnosis, and Alex is now receiving appropriate treatment.
Next comes the issue of access. Healthcare systems are skewed. The neediest individuals—the sickest, poorest, and most marginalized—are often left behind. Overbooked schedules and inadequate public transport result in missed appointments for millions. Parents and part-time workers, particularly those in the gig economy, struggle to attend physical examinations. According to the American Time Use Survey, patients sacrifice 2 hours for a mere 20-minute doctor visit. For those with disabilities, the situation often worsens. Transportation issues, costs, and extended wait times significantly increase the likelihood of missed care in the UK. Women with disabilities are over seven times more likely to face unmet needs due to care and medication costs compared to men without disabilities.
Yet, it is uncommon to challenge the notion of waiting for a physician because it has always been the norm. AI has the potential to shift that paradigm. Imagine having a doctor in your pocket, providing assistance whenever it’s needed. The workers’ 10-year plan unveiled by Health Secretary Wes Streeting proposes that patients will be able to swiftly discuss AI and health concerns via the NHS app. This is a bold initiative, potentially offering practical clinical advice to millions much quicker.
Of course, this hinges on accessibility. While internet access is improving globally, substantial gaps remain, with 2.5 billion people still offline. In the UK, 8.5 million individuals lack basic digital skills, and 3.7 million families fall below the “minimum digital living standard.” This implies poor connectivity, obsolete devices, and limited support. Confidence is also a significant barrier; 21% of people in the UK feel they are behind in technological understanding.
Currently, AI healthcare research primarily focuses on its flaws. Evaluating biases and errors in technology is crucial. However, this focus overlooks the flaws and sometimes unsafe systems we already depend upon. A balanced assessment of AI must weigh its potential against the reality of current healthcare practices.
Charlotte Brees is a health researcher; Dr. Bott: Why Doctors Can Fail Us, and How AI Can Save LifePublished by Yale September 9th.
Read more
Deep Medicine: How Artificial Intelligence Can Make Health Care Human Eric Topol (basic book, £28)
Co-Intelligence: Life and cooperation with AI Ethan Morrick (WH Allen, £16.99)
Artificial Intelligence: A Guide to Thinking about Humans Melanie Mitchell (Pelican, £10.99)
Source: www.theguardian.com
