Research interests:
Dr Starkey’s research interests centre around how electronic health records can be used to better understand medication-related harm and identify people most at risk as part of the HDR UK Medicines in Acute and Chronic Care Driver Programme. As part of this work, he is working on developing and validating a clinical prediction model to identify individuals at the highest risk of severe bleeding events while taking direct oral anticoagulants, a class of blood-thinning medications.
He has extensive experience of health data analytics including the use of statistical, epidemiological and data science methodologies in population-level datasets to answer complex questions and inform clinical guidelines. Thomas completed his PhD at The University of Birmingham, where he used statistical approaches and electronic health records to evaluate the impact of COVID-19 in people living with cancer including responses to COVID-19 vaccination and disease outcomes.
Project title
Developing and validating a multivariable model to predict severe bleeding in patients using direct oral anticoagulants (DOACs)
The project
Direct oral anticoagulants (DOACs) are blood-thinning medications prescribed for preventing and treating blood clots. This includes prevention of stroke in individuals with atrial fibrillation (irregular and often fast heart rate) and treating deep vein thrombosis (blood clots) in the leg. However, despite becoming the blood thinners of choice, DOACs may lead to medication-related harm, especially when given to patients with other risk factors. These factors may include the co-existence of two or more long-term illnesses (multimorbidity), multiple concurrent medications (polypharmacy), and interactions between medications (drug-drug interactions).
The aim of this project is to develop and validate a model to predict severe bleeding in patients using DOACs resulting in hospital admission and/or death in the UK. This will incorporate electronic health records from primary care settings such as GP surgeries linked to hospital admission and death records. Novel statistical and machine-learning techniques will be evaluated and compared for in predicting severe bleeding during model development and candidate predictor selection guided by clinical relevance.
A model or risk score that can accurately predict severe bleeding in patients using DOACs will help identify those patients who may require enhanced monitoring or medication review. This should not only reduce DOAC-related harm but also save the NHS millions of pounds that are spent on treating this medication-related harm.
Why did you want to undertake this project?
Medication-related harm results in adverse outcomes for patients, particularly for those with multiple long-term conditions or diseases on several different medicines. More specifically, people prescribed with direct oral anticoagulants (DOACs) are often older and with multiple long-term health conditions who may be more susceptible to severe harm from prescription medications. A model or risk score that can accurately predict severe bleeding in patients using DOACs may therefore help identify those patients who require enhanced monitoring or medication review. This should not only reduce DOAC-related harm but also save the NHS millions of pounds that are spent on treating this medication-related harm at a time when national healthcare resources are scarce.
Support from the HDR UK Medicines in Acute and Chronic Care Driver Programme for this project will have a considerable impact on my research by enabling me to share expertise within a diverse team of experts, including clinicians, academic researchers, and patients. This collaborative approach is important for developing innovative solutions to address complex clinical questions and, perhaps more importantly, involving and engaging with patients and the public as part of the HDR UK ‘Patient and Public Involvement and Engagement’ mission.
