Despite Singapore's AI push, some doctors say cost and access keep the technology out of reach

SINGAPORE: Dr Song Majinyang uses ChatGPT's free trial to draft emails, standard operating procedures and workflows. But for now, she has no plans to pay for clinical AI tools to help with patient care.

Her reasons are straightforward: cost, access and questions about value. The Ministry of Health announced a S$200 million (US$155 million) investment in October 2024 to bring AI to public healthcare institutions over five years. But many independent GPs find themselves paying out of pocket or going without.

Of five GPs CNA spoke to, two actively use AI for clinical work, while the others limit it to basic administrative tasks or avoid it altogether. For those who have adopted AI, the expenses come from their own pockets. 

"If it's not easily accessible to GPs, then the hurdle will be very high," said Dr Song, who runs MyCare Medical Clinic in Chai Chee. "We're quite a small clinic, there's no incentive to try this out."

WHAT CLINICS ARE SPENDING

Dr Joanne Koay, chief medical officer of Assure Family Clinic in Bukit Merah, said her clinic spends up to S$2,200 annually on a Ministry of Health-approved AI platform. The technology supports health screening, genomics-based wellness assessments for customised care plans and early cancer detection.

Dr Joshua Chua of Cavenagh Medical Clinic in Bukit Timah subscribes to four different AI platforms at roughly S$100 per month combined. He uses them for dictation after consultations, translation of foreign medical reports and generating Excel formulas to organise patient data for upcoming vaccinations and laboratory tests.

While clinics generally absorb these costs, Dr Koay acknowledged that some of it may eventually trickle down to patients if fees become too high.

For small, independent practices, the question becomes whether to invest in technology that may or may not improve patient outcomes.

Dr Roland Xu of Procare GP Clinic in Ang Mo Kio, who currently does not use any AI tools, said transcription tools could help clinics with high patient loads, but "singleton" clinics like his – those that do not belong to a chain – can still comfortably manage documentation without AI.

"If you have to adopt newer technology whatsoever, be it AI or not, the cost is usually harder to bear for singleton GPs," he said. 

Cost isn't the only obstacle. Dr Song points to a more fundamental issue: commercial platforms like ChatGPT offer generic responses to medical questions, and the sophisticated clinical AI models she is aware of are either restricted to pilot programmes or not readily available.

Dr Song is also concerned about confidentiality, and said doctors need reassurance that patient data stays secure on AI platforms, without external parties "reading behind the content".

Cavenagh Medical's Dr Chua said he is careful to "never, ever" put patient data into these platforms and leaves out any identifiers.

For RadLink Group, a diagnostic imaging provider serving GPs and specialists, the solution has been to keep data local. Medical director Dr Eng Chee Way said the company uses AI for chest X-rays, mammograms and CT lung scans, with imaging data stored locally and not uploaded to cloud platforms or shared with AI vendors.

IS AI NECESSARY FOR DOCTORS?

Opinions were split on whether AI meaningfully improves patient care.

Dr Xu said AI in healthcare focuses primarily on diagnostics, which is more relevant for tertiary care providers such as hospitals that handle complex cases. He also questioned whether there would be sufficient return on investment for such tools.

"If patients themselves don't see the need to even ask whether the doctors that they are seeing are, in fact, utilising any form of AI models, then I don't really see the benefit nor the requirements for us to be adopting AI at this current juncture," he said. 

Dr Song observed that most clinics where she works as a locum do not use AI. Some older GPs still struggle with existing computer systems, making AI an even bigger leap. But these clinics did not have AI in the past, and have survived without it, she said. 

To her, AI is "an add-on, it's a good-to-have, but it's not a necessity".

Others see AI as increasingly valuable in modern medical practice, such as in cancer screening. In the past, patients with normal-looking scans were simply sent home, said Dr Koay. Today, AI can estimate a patient’s risk to a certain degree even when no abnormalities are visible, allowing doctors to monitor them more closely for early warning signs.

She said she believes prevention is better than cure, and that identifying risks early is far more effective than waiting for a disease to develop.

Dr Eng described himself as a "big proponent" of AI and believes medical professionals should not avoid using it.

"If you don't use it, you are doing your patient a disadvantage, a disservice at this juncture," he said. 

But he also acknowledged its limitations. AI hallucinations – where the system generates incorrect or misleading information – remain a risk, and the technology can miss important details since it is only as robust as the data it is trained on.

The advantage, he said, is that unlike humans, AI does not fatigue and is "very, very consistent". 

He does not think medical imaging is ready to go fully autonomous yet. Instead, humans and AI should work together and make up for each other's "blind spots". 

"It's more like flying a plane, where you have auto navigation and a pilot at the same time," said Dr Eng. That, he added, is the "optimal outcome".