JARVIS-MACE: AI-driven risk prediction model for patients with Type 2 diabetes

Project Reference :


Institution :

National University of Singapore (NUS), Singapore General Hospital (SGH)

Principal Investigator :

Prof Wynne Hsu (NUS) and Dr Amanda Lam (SGH)

Technology Readiness :

4 (Technology validated in lab)

Technology Categories :

AI - Deep Learning

Background/Problem Statement

Diabetes mellitus is a condition reaching epidemic proportions worldwide and in Asia. Individuals with type 2 diabetes have a twofold increased risk for cardiovascular disease (CVD) such as myocardial infarction, stroke and peripheral vascular disease. CVD is also the principal cause of death in type 2 diabetes patients. Having a means of estimating risk of diabetes complications will aid population health management, by allowing healthcare providers to allocate valuable resources towards patients at high risk of complications. Presently, risk prediction engines for cardiovascular risk in diabetes patients derive their predictions from cross-sectional data at a single time point in a patient’s journey, which is less than ideal, given the longitudinal contributions of various risk factors to the risk of cardiovascular complications.


Developed in close collaboration with the Singapore General Hospital, JARVIS-MACE Risk Prediction Model (JMRPM) is an AI-driven longitudinal risk-prediction model for major adverse cardiovascular events (MACE) that utilizes a recurrent temporal time process network to learn a latent representation of patient visit history to derive the survival probability function to obtain the risk of an event occurring and to estimate the time to next event.

JMRPM addresses the lack of a method to estimate risk of MACE in patients with type 2 diabetes mellitus, especially one based on longitudinal rather than cross-sectional data, and based on a predominantly Asian population.

JMRPM achieved the highest C-index score of 0.765 compared to other similar models trained on the T2DM dataset for a first event prediction.

Link to publication: Recurrent temporal point process network for first and repeated clinical events


Compared to other existing approaches that only consider patient cross-sectional data, our approach takes patient longitudinal record as inputs. This means that our approach considers the temporal trajectory of patient over multiple visits. It is important to consider the disease trajectory of patients, especially in chronic diseases such as type 2 diabetes mellitus, as their outcomes mainly depend on how their relevant indicators, such as blood glucose, change over time.

When used as a counselling tool to explain an individual patient’s risk to them during a clinical encounter, this tool may allow patients and providers to overcome treatment inertia.

Potential Application(s)

The use of JMRPM by primary healthcare providers to estimate risk of MACE in type 2 diabetes mellitus patients will aid population health management, by allowing healthcare providers to allocate valuable resources towards patients at high risk of complications. This will in turn enable improved management of diabetes mellitus and relevant comorbidities to achieve superior outcomes.

We welcome interest from the industry for collaboration/ co-development / customisation of the technology into a new product or service. If you have any enquiries or are keen to collaborate, please contact us.