Skip to navigation Skip to main content Skip to footer

Approved Research

Utilising machine learning in predicting adverse drug events in older people.

Principal Investigator: Dr Prasad Nishtala
Approved Research ID: 89800
Approval date: October 27th 2022

Lay summary

Adverse Drug Events (ADEs) in older adults !65 years old are common and have a significant effect on the patient's functionality and quality of life. Previous research in this field has focused on recognisable ADEs such as bleeding from anticoagulants, however, a high proportion of ADEs such as falls are attributed to age and a natural decline in health and are under-reported. Additionally, studies have found that 50% of ADE-related hospital admissions are preventable. There is an urgent need to develop methods to accurately forecast clinical outcomes associated with ADEs. Traditional statistical analysis of data has the drawback of not being able to consider multiple factors when assessing risk. Advances in Machine Learning (ML) approaches have given a new scope to accurately forecast clinical outcomes associated with ADEs from large healthcare databases in an attempt to address the high prevalence of ADEs in older adults.

Our study, which will last approximately 36 months, aims to perform an analysis of UK Biobank data to identify populations of older adults at risk of ADEs, such as hospitalisations and cognitive impairment. We will use various ML approaches to build predictive models which will accurately predict ADEs in older adults. The predictive models will be built using a statistical coding language, R. The main outcomes in this study will be the multiple ML model metrics which will give an indication of how accurate and robust our predictive model is. In the later stages of the project, we will attempt to develop and optimise various novel ML techniques to predict ADEs in older adults. This will take advantage of the vast amount of information contained in narratives of ADE reports held by pharmacovigilance regulatory agencies.

If associations between ADEs and high-risk drugs were accurately identified and predicted, changes in prescribing policy could result in fewer instances of ADEs in older adults. In turn, such an outcome would have significant positive effects on general public health. Moreover, a decrease in ADEs among older adults would allow limited NHS resources to be redirected to other clinical areas.