Routine use of artificial intelligence in colonoscopies may erode doctors’ skills, a new study has suggested, stirring concerns about whether AI could weaken doctors’ clinical abilities as the technology spreads into medicine.

The study by researchers in Europe has found that the rate at which experienced doctors detected precancerous growths in the colon through colonoscopies without AI assistance decreased by 20 per cent several months after routine introduction of AI.

The researchers tracked 1,443 colonoscopies — 795 before and 648 after the adoption of regular AI — performed by 19 experienced endoscopists in four hospitals in Poland, each of whom had conducted over 2,000 colonoscopies before the study.

The average detection rate for possible precancerous growths, called adenomas, fell from 28 per cent before AI exposure to 22 per cent afterwards — a relative reduction of 20 per cent. In the colonoscopies performed with AI, the detection rate was 25 per cent.

“We were surprised that the observed deskilling was high,” said Marin Romanczyk, assistant professor of gastroenterology at the Academy of Silesia in Poland and a member of the study team. “But it may reflect a pattern seen with other technologies — for instance, reliance on online maps can weaken the ability to navigate with paper maps,” he told The Telegraph.

The study, published on Tuesday in The Lancet Gastroenterology and Hepatology, is the first to examine whether prolonged exposure to AI might inadvertently erode skills of endoscopists. Earlier studies, including one in India, have found that AI can boost adenoma detection rates.

Omer Ahmad, a consultant gastroenterologist at University College London who was not involved in the research, said the findings highlight the importance of weighing possible unintended consequences in the rapid adoption of AI in medicine.

The results “provide the first real-world clinical evidence for the phenomenon of deskilling, potentially affecting patient-related outcomes”, Ahmad wrote in an accompanying commentary in the journal.

Some doctors urged caution in interpreting the results.

“Any suggestions about deskilling based on these results would be premature,” said Mohinish Chhabra, director of gastroenterology and therapeutic endoscopy at Fortis Hospital in Mohali, Punjab, who earlier this year completed India’s first study evaluating AI-assisted colonoscopies.

Chhabra said the main factor influencing adenoma detection rates is the time spent inspecting the colon — whether with or without AI — called the colonoscopy withdrawal time. The longer the inspection, the higher the detection rate.

AI-assisted colonoscopy is one example of the growing use of AI in medicine, from image analysis to diagnostic decision support to robotic surgery. The expansion has also raised concerns among some patients’ rights advocates that certain diagnostic procedures, including colonoscopies, may sometimes be driven by commercial incentives.

Romanczyk said the findings were particularly relevant amid AI’s rapid uptake in clinical practice. They underscore the need for more research into how the technology affects the skills of healthcare professionals in different fields and how doctors and AI can work together more effectively, he said.

Chhabra agreed that more research is needed, but for a different reason. “We’ll need studies that follow many doctors over time to confirm whether deskilling is happening,” he said.

The Fortis study, which screened 501 patients over two years, found that AI-assisted colonoscopies detected polyps in 51 per cent of patients compared with 27 per cent using standard colonoscopies, and adenomas in 41 per cent compared with 17 per cent.