Dengue is an infection caused by a flavivirus, and affects up to 100 million people each year worldwide. Although most patients will experience fever and other symptoms which resolve after a few days, a small proportion of patients (up to 5%) develop severe disease with bleeding, shock or organ failure. The healthcare and socioeconomic burden of dengue is immense, and healthcare facilities are frequently placed under high demand during epidemics.

Being able to identify and monitor patients with dengue carefully means they can receive prompt life-saving treatment. In our study, our group at Imperial looked at whether using a low cost wearable device (costing < $100 USD) in hospital could do just this. At the moment, the current gold standard for monitoring involves frequent, invasive blood tests to measure haematocrit.

We carried out a prospective clinical study at the Hospital for Tropical Diseases, a specialist referral hospital in Southern Vietnam which sees thousands of dengue patients a year. We recruited 250 patients with dengue and asked each of them to wear a non-invasive device which uses photoplethysmography (PPG) in the fingerprobe to capture information about the patient for up to 24 hours. Time series machine learning analysis including multi-modal spatio-temporal fusion transformer models were developed, and we found that it was possible to predict if patients were to become unwell, up to two hours ahead in time with an area under the receiver operator curve (AUROC) of 0.83 and area under the precision recall curve (AUPRC) of 0.67.

With the right implementation this could mean that patients can be safely monitored in our setting without the need for repeated observations or blood tests – adding an additional layer of safety to care. The continuous nature of monitoring also means that any changes in clinical state can be picked up without the need for nursing staff or clinicians to physically attend.

Since these findings, we have taken the research forward and developed the D-SCAPE (Dengue Shock Classification And Prediction wEarable) device and hardware at Imperial College London and aim to carry out focused future clinical studies, and see how we can improve dengue management. With smaller wearables, a big question is whether patients can avoid having to come into hospital at all, whilst being managed safely at home.