Use this URL to cite or link to this record in EThOS:
Title: ML-EWS - Machine Learning Early Warning System : the application of machine learning to predict in-hospital patient deterioration
Author: Nangalia, V.
ISNI:       0000 0004 7225 3146
Awarding Body: UCL (University College London)
Current Institution: University College London (University of London)
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Preventing hospitalised patients from suffering adverse event (AEs) (unexpected cardiac, arrest, intensive care unit admission, surgery or death) is a priority in healthcare. Almost 50% of these AEs, caused by mistakes/poor standards of care, are thought to be preventable. The identification and referral of a patient at risk of an AE to a dedicated rapid response team is a key mechanism for their reduction. Focussing on variables that are routinely collected and electronically stored (blood test data, and administrative data: demographics, date and method of admission, and co-morbidities), along with their trends, I have collected data on ~8 million admissions. I have explained how to navigate the complex ethical and legal landscape of performing such an ambitious data linkage and collection project. Analysing data on ~2 million hospital admissions with an in-hospital blood test result, I have 1. described how these variables (particularly urea and creatinine blood tests, method of admission, and date of admission) influence in-hospital mortality rate in different groups of patient. 2. created four machine learning (ML) models that have the highest accuracy yet described for identifying a patient at risk of an SAE, while at the same time capturing the majority of patients likely to die (high sensitivity). These models ML-Dehydration, ML-AKI, ML-Admission, and ML-Two- Tests, can be applied to admissions with limited data, specific syndromes, or on all patients in hospital at different time points in their hospital trajectory respectively. Their area under the receiver operator curves are 79.6%, 85.9%, 93% and 90.6% respectively. 3. built and deployed a technology platform Patient Rescue that allows for the automated application of any model in any hospital, as well as the communication of rich patient level reports to clinicians, all in real-time. The ML models and the Patient Rescue platform together form the ML – Early Warning System.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available