Narrowing the gap between technological capability and human confidence in the context of artificial intelligence (AI) in emergency care still needs a lot of work and close collaboration among developers, clinicians, and patients, suggests a Singapore study.
“Emergency departments (EDs) worldwide are facing increasing patient loads, staffing challenges, and heightened demands for efficiency and safety. AI has emerged as a promising solution to these pressures,” said authors Dr Hoon Chin Steven Lim, Department of Emergency Medicine, Changi General Hospital, Singapore, and Dr Siew Feng Rachel Teng, MOH Holdings Pte Ltd, Elementum, Singapore.
A survey was conducted among 447 attendees of the 2025 Society for Emergency Medicine in Singapore Scientific Meeting at Ng Teng Fong Centre for Healthcare Innovation. Participants were asked about their confidence in AI enhanced care from a patient’s perspective and their openness to using AI applications in the ED, including medical imaging, clinical decision support, predictive analytics, and others.
Applications
Of the respondents, 118 completed the survey. The majority reported slight (n=59) or no (n=8) confidence in AI-enhanced care as patients, and only a few expressed high confidence levels. [Proc Singap Healthc 2026;doi:10.1177/20101058261472917]
The most preferred AI applications were medical imaging and diagnostics (n=79) and clinical decision support systems (n=56). Other applications, including predictive analytics, robotic process automation, and AI in surgical procedures, earned less interest.
Many emergency care providers acknowledged the potential of AI, but they also expressed caution and limited confidence in its use. The preferences of the respondents showed openness to AI tools that directly improve existing diagnostic and clinical decision-making tasks, which is consistent with the current landscape of AI development in emergency medicine.
“Achieving success in leading domains and carefully selected use cases will grow confidence and trust, paving the way toward positive AI-enhanced emergency care,” Lim and Teng said.
“Through close collaboration and feedback between developers and users, AI can realize its full potential as a means to support resuscitation and save lives, without compromising the trust that underpins all patient-provider relationships,” they added.
Traceability
In the Clinical AI Report by Stanford‒Harvard Research Network, experts stated the need for benchmarking on multi-turn unstructured real-world data and understanding the consequences of model errors in real-world settings. [https://arise-ai.org/report]
The report also expressed the need for processes that enable traceability of AI contributions to patient adverse events and that are capable of catching up with the speed at which AI is developing. [JAMA Health Forum 2024;5:e235514]
Furthermore, Lim and Teng stressed the benefit of AI in precision health, particularly in organizing, analysing, and interpreting data, which allow physicians to deliver targeted diagnoses, prognoses, and treatment recommendations. [https://www.ama-assn.org/practice-management/digital-health/future-ai-and-precision-health-what-stands-way]
“We need to ask whether the AI used is generalizable and suitable for the patient seated in front of us,” they said. “Datasets shaping that AI need to be diverse, longitudinal, and representative of our patient population, reducing bias and promoting equity.” [Lancet 2023;402:503]
Moreover, fully automated AI applications must be transparent when making interpretive clinical decisions without human intervention, according to Lim and Teng.
“When both patients and clinicians are made aware of such background applications, they can make an informed choice concerning the use of such systems and their output, thereby protecting their autonomy,” they said.
“It is imperative that we address these factors during the design, development, and implementation of AI. In addition, the understanding of the domains where providers are most comfortable experimenting with AI will help us to focus our efforts as trust is built,” they added.