Auto-risk: Hospital-embedded personalised risk trajectories for 'patients like me'

Lead Research Organisation: University College London
Department Name: Institute of Health Informatics


Data-driven methods are poised to revolutionise healthcare by learning from a cohort of patients far exceeding that of an individual clinician, or indeed institution's, lifetime experience. However most models only capture a snapshot of the patient's current state on admission and fail to exploit the rich temporal evolution of an individual's healthcare record. Furthermore, there is commonly a disconnect between healthcare data from GPs and hospitals, owing to the lack of a universal data access platform that bridges primary and secondary care, meaning that much information on patients' past medical history is fragmented and incomplete. Analytical approaches utilising datasets from diverse sources (i.e., primary care, secondary care and specialist registries), may thus facilitate better understanding of disease processes with consequent potential to accelerate diagnosis, allow for more personalised monitoring and institution of earlier treatment.
Utilising national datasets encompassing both primary (Clinical Practice Research Datalink) and secondary care (Hospital Episode Statistics) with previously validated phenotypes from the CALIBER portal, we will adopt novel machine learning methods to map disease trajectories for common medical conditions, encompassing the path from inpatient admission to discharge/death, including inpatient events and treatments, such as Critical Care admission. These machine learning techniques will then be compared against conventional statistical survival models and validated on local data from the University College London Hospitals (UCLH) electronic health records platform.
Finally, the model will seek to be the first UK demonstrator of a 'Clinical Informatics Consult' framework by integrating a personalised risk trajectories app within the Experimental Medicine Application Platform (EMAP) at UCLH. The app will be available for clinicians to query by inputting a patient's Master Reference Number to return a personalised risk


10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/S021612/1 01/04/2019 30/09/2027
2418748 Studentship EP/S021612/1 28/09/2020 30/09/2024 Christopher Tomlinson