Background Access to trustworthy artificial intelligence (AI) models for clinical applications like emergency care is unevenly distributed globally due to healthcare inequities. Low-resource settings face challenges in AI model development due to limited data, small sample sizes, and inconsistent data quality. Moreover, models developed from high-resource settings are often not readily applicable in low-resource contexts. Transfer learning (TL) is an AI technology that adapts established models to new settings and offers a potential solution. This study explores the feasibility of TL in clinical contexts, using neurological outcome prediction for out-of-hospital cardiac arrest (OHCA) as a proof of concept.

Methods The Pan-Asian Resuscitation Outcomes Study (PAROS) network provides a multicenter registry for OHCA across the Asia-Pacific region. We applied TL to adapt a neurological outcome prediction model for OHCA, originally developed using a large Japanese cohort (i.e., the external model), to two PAROS registry countries: Vietnam (243 patients) and Singapore (15,916 patients). Separate TL models, calibrated with local data from Vietnam or Singapore, were developed and compared with the external model. Their predictive performance was then compared with that of the external model.

Findings The external model performed poorly on the Vietnam cohort, with an area under the receiver operating characteristic curve (AUROC) of 0·467 (95% CI: 0·141-0·785). The TL-Vietnam model significantly improved performance (AUROC = 0·807, 95% CI: 0·626-0·948). In Singapore, the TL-Singapore model demonstrated modest improvements (AUROC = 0·955, 95% CI: 0·940–0·967), up from 0·945 (95% CI: 0·929–0·958) in the external model.

Interpretation This study highlights the potential of TL to improve prediction accuracy in low-resource settings worldwide, promoting global healthcare equity.

Funding This study was supported by SingHealth Duke-NUS ACP Programme Funding, National Medical Research Council, Clinician Scientist Awards, Ministry of Health, Health Services Research Grant, Singapore, and Laerdal Foundation.

MEH Ong reports grants from the Laerdal Foundation, Laerdal Medical, and Ramsey Social Justice Foundation for funding of the Pan-Asian Resuscitation Outcomes Study; an advisory relationship with Global Healthcare Singapore (SG), a commercial entity that manufactures cooling devices. MEH Ong has a licensing agreement with ZOLL Medical Corporation and patent filed (Application no: 13/047,348) for a Method of predicting acute cardiopulmonary events and survivability of a patient. He is also the co-founder and scientific advisor of Technology Innovation in Medicine (TIIM) Healthcare, a commercial entity which develops real-time prediction and risk stratification solutions for triage. He is a member of the Editorial Board of Resuscitation. YO has received a research grant from the ZOLL Foundation and an overseas scholarship from the FUKUDA Foundation for Medical Technology and the International Medical Research Foundation. All other authors have no conflict of interest to declare.

This study was funded by SingHealth Duke-NUS ACP Programme Funding, National Medical Research Council, Clinician Scientist Awards, Ministry of Health, Health Services Research Grant, Singapore, and Laerdal Foundation.

I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained.

Yes

The details of the IRB/oversight body that provided approval or exemption for the research described are given below:

Ethics committees or Institutional Review Boards (IRBs) of all participating institutions in the Pan-Asian Resuscitation Outcomes Study (PAROS) network, including the Centralized Institutional Review Board and Domain Specific Review Board of Singapore Ministry of Health Holdings (reference numbers: 2013/604/C, 2013/00929, and 2018/2937), gave ethical approval for this work. Informed consent was waived due to the observational nature of the study, and all data were de-identified.

I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals.

Yes

I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance).

Yes

I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable.

Yes

The data supporting the findings of this study are not publicly available due to privacy and institutional data-sharing agreements.