Document Type

Article

Department

Haematology and Oncology, East Africa

Abstract

Aim: Hypertension and diabetes mellitus (DM) are major causes of morbidity andmortality, with growing burdens in low-income countries where they are underdiag-nosed and undertreated. Advances in machine learning may provide opportunities toenhance diagnostics in settings with limited medical infrastructure.

Materials and Methods: A non-interventional study was conducted to develop andvalidate a machine learning algorithm to estimate cardiovascular clinical and labora-tory parameters. At two sites in Kenya, digital retinal fundus photographs were col-lected alongside blood pressure (BP), laboratory measures and medical history. Theperformance of machine learning models, originally trained using data from the UKBiobank, were evaluated for their ability to estimate BP, glycated haemoglobin, esti-mated glomerular filtration rate and diagnoses from fundus images.

Results: In total, 301 participants were enrolled. Compared with the UK Biobankpopulation used for algorithm development, participants from Kenya were youngerand would probably report Black/African ethnicity, with a higher body mass indexand prevalence of DM and hypertension. The mean absolute error was comparable orslightly greater for systolic BP, diastolic BP, glycated haemoglobin and estimated glo-merular filtration rate. The model trained to identify DM had an area under thereceiver operating curve of 0.762 (0.818 in the UK Biobank) and the hypertensionmodel had an area under the receiver operating curve of 0.765 (0.738 in the UKBiobank).

Conclusions: In a Kenyan population, machine learning models estimated cardiovas-cular parameters with comparable or slightly lower accuracy than in the populationwhere they were trained, suggesting model recalibration may be appropriate. Thisstudy represents an incremental step toward leveraging machine learning to makeearly cardiovascular screening more accessible, particularly in resource-limitedsettings.

Publication (Name of Journal)

Diabetes, Obesity and Metabolism

DOI

https://doi.org/10.1111/dom.15587

Included in

Cardiology Commons

Share

COinS