Turning care data into patient avatars to aid clinical decision making
Research type
Research Study
Full title
Turning Paediatric Intensive Care data into patient Avatars to aid clinical decision making: A clustering and causal prediction approach
IRAS ID
345909
Contact name
Kelly Davies
Contact email
Sponsor organisation
Alder Hey Children's Hospital NHS Foundation Trust
Duration of Study in the UK
1 years, 0 months, 1 days
Research summary
Paediatric Intensive Care (PIC) treats critically ill children with complex and often rare medical conditions. Patients admitted to PICUs vary greatly, and even children with similar illnesses often respond differently to treatment, making it difficult for doctors to predict outcomes or choose the most effective care.
Intensive care units collect large volumes of patient information, including vital signs, blood tests, ventilator measurements, diagnostic imaging and demographic data. However, interpreting such vast data to recognise useful patterns among patients is challenging and often underutilised due to its complexity and volume.
Identifying patterns within patient data could help clinicians understand why some children respond differently to the same treatments, enabling personalised care and improved outcomes.
This exploratory study aims to identify groups of patients who share similar risks or characteristics. Using advanced computer-based methods (e.g. clustering algorithms, causal prediction), we will analyse patient data to automatically group similar cases together. These algorithms detect patterns that may not be easily recognised by clinicians through routine observation alone. This research is based on the idea of developing a 'patient avatar' - a digital model that represents a child's condition and how it changes over time.
We will study data from approximately 2500 children admitted over three years to a large PICU in the UK. The analysis will include a subset of routinely collected data from an ICU admission.
We will focus on two primary outcomes: survival during ICU stay and length of stay in intensive care. These outcomes were selected because they are critical indicators of patient severity. We will then investigate applying these models to a continuously collected, normalised dataset.
Ultimately, this research seeks to determine whether grouping patients using these data-driven methods can help clinicians better anticipate illness progression, inform clinical decisions during ICU stays, and stimulate further research into tailoring interventions to improve outcomes for critically ill children.REC name
London - Westminster Research Ethics Committee
REC reference
25/PR/1406
Date of REC Opinion
28 Oct 2025
REC opinion
Favourable Opinion