*NEW* - Machine Learning for Personalised Diabetes Medication Selection using Traditionally Measured Longitudinal Clinical Variables

Type 2 diabetes (T2D) is common in Singapore and requires long-term medication management. Many patients go through several treatment changes before their blood sugar is adequately controlled, which increases the risk of complications and drives up healthcare costs.

This study will use routinely collected health data — including blood test results, prescription records, and clinic measurements — to build prediction models that estimate which diabetes medication a given patient is most likely to respond to. The aim is to move away from one-size-fits-all treatment guidelines toward medication choices that reflect each patient’s clinical profile at the time of prescribing.

Models will be developed using Singapore’s national health data platform (TRUST) and tested in separate patient cohorts. We will also estimate whether earlier, better-matched prescribing could reduce hospital admissions and costs. Findings will be published and will inform future work on data-driven diabetes care.