Pushing AI to the front of house in healthcare

At the CHI Innovate 2025 conference, SingHealth’s Director of Innovation for Radiology Dr Charlene Liew raised an important point.

“Can we actually work alongside artificial intelligence (AI), each of us doing our own tasks, separately but synergistically.

"We’ve been working on this research ourselves, and would propose a new model: Instead of human in the loop, we can have the human in the parallel loop,” she said.

At the conference, organised by NHG Health on July 10 and 11, speakers like Dr Liew highlighted how AI is not just a high-tech tool, but a partner that can be leveraged for deeply “human” sectors like healthcare and social work.

Dr Liew is also Changi General Hospital (CGH)’s Director for the AI Office, as well as Duke-NUS Medical School’s Clinical Assistant Professor.

She noted that in response to a study done by CGH. It was found that AI was able to independently diagnose patients with up to 90 per cent accuracy.

Physicians who used AI only achieved a 76 per cent accuracy rate, and an accuracy of 74 per cent without using AI.

Dr Liew was speaking at the panel AI for all: From Rarefied to Regular, alongside speakers from nursing, medical social work and community care.

Working in parallel

To illustrate how humans can work in parallel with AI, Dr Liew shared about a trial conducted to assess the accuracy of AI systems in classifying chest X-rays.

Forty-three radiologists provided the reference standards (normal, non-urgent and urgent) against which the AI’s performance would be measured.

The trial showed that the AI system was 98 per cent accurate in confirming that the patient was healthy.

CGH now has a system in place to allow radiologist to cross check the results given by AI and approve them for “normal” chest x-rays, enabling them to work in parallel with AI systems.

Dr Liew added that CGH was also exploring using generative AI (GenAI) tools like Pair with medical residents, as well as in using agentic AI to analyse medical records.

Smarter rostering

IHH Healthcare Singapore’s Director of Nursing, Sally Lim, shared how the institution has been using AI for its nurse rostering system, NurseShift.ai, since June 2024.

Traditional rostering was tedious, and it often struggled to accommodate the needs of multiple stakeholders, she noted.

However, the use of AI has streamlined this process. According to Lim, NurseShift.ai has reduced the average time taken to do roster plans by half.

“Up to 87 per cent of the shift requests are now being fulfilled... The difficulty in ensuring appropriate skill levels has decreased from 89 to 23 [percent],” shared Lim.

In this way, real-time, flexible scheduling has also increased nurse satisfaction. By lowering administrative burden and overhead costs, nurses can focus their time and energy on caring for the patients, Lim said.

To subscribe to the GovInsider bulletin, click here.  

Enabling patient-centric care

AI has found its way even in the “soft sciences” of social work.

Woodlands Health’s Senior Principal Medical Social Worker (MSW), Melissa Chew, shared that MSWs spend a lot of time documenting sessions with patients, on top of actual sessions with patients.

This leads to burnout among social workers due to working overtime, shared Chew.

Besides, she noted that both patient and caregivers might find it distracting when the MSW interacts with them while typing the case notes and retrieving the case information.

To address this problem, Open Government Products (OGP) developed Scribe, an AI transcription and summarisation tool.

In particular, the generated summaries can be exported directly to Care360 (a patient management system developed by OGP that is also used by MSWs), and both its speech recognition and large language models do not rely on external programs: sensitive patient data never leaves the system.

According to Chew, Scribe can recognise most local languages and dialects and generates effective summaries of sessions with patients, including areas for potential follow-up.

The outcome?

A 38.7 per cent reduction in documentation time, or approximately 18 minutes less per session, said Chew.

Beyond MSWs, Scribe was also used in community based social work and even by other healthcare professionals coming from other clinical and allied health functions, as well as operations, continued Chew.

Likewise, Lions Befrienders’ Executive Director, Karen Wee, shared that complementing human counsellors with AI can help speed up the early assessment of seniors’ mental health.

“We could cut out that 10 minutes or 15 minutes and that's only because it's early screening. If it’s severe or suicidal, we would have to refer [the seniors] back to the hospitals,” she explained.

However, the counsellors would still manage traditionally high-touch aspects of the therapeutic relationships between client and therapist, complex diagnoses and culturally sensitive engagements.

“AI today is not that ready, but it might be even able to reach that in ten years’ time,” she shared.

Ultimately though, Wee emphasised that using AI in healthcare was “not about us and the clients, per se. It's about teamwork. It's about collaboration. It's about having an open mind.”

Editor's note (Sept 29): The article previously highlighted that Care360 and OGP worked together to build Scribe. This is incorrect, and has been corrected. Both Scribe and Care360 are systems developed by OGP.