Developing data driven algorithms to characterise children with multimorbidity in NHS databases

Background:
An increasing number of children are living with multimorbidity, defined as living with two or more long-term conditions.(1) Children with multimorbidity have higher mortality and are more likely to require emergency hospital admission and integrated health, education support and social care services than other children. There is great interest from families and policy makers in establishing risk factors for multimorbidity manifesting in childhood, and monitoring health outcomes for affected children.

Large, administrative health databases are ideal for these purposes. A number of classifications for identifying children with chronic or complex conditions within administrative health databases exist.(2, 3) These have been developed using literature reviews and expert clinical input. There are likely substantial advantages to using machine learning methods to classify groups of children with different types of multimorbidity in administrative health databases,(4) making full use of data contained in the longitudinal health records. An efficient, data-driven method to characterise children with multimorbidity would allow comprehensive monitoring of health outcomes to help plan services and assess the impact of new treatments.

Aims/Objectives:
The aim of this PhD project is to use machine learning to differentiate groups of children with distinct longitudinal trajectories of multimorbidity. The specific PhD objectives are to:

1) review existing classifications of multimorbidity in children in administrative health databases
2) develop a machine learning algorithm to characterise distinct groups of children with multimorbidity using longitudinal hospital admission and/or primary care data
3) use the medical multimorbidity classification to predict key health outcomes including mortality, emergency hospital admissions, or polypharmacy
4) explore risk factors for childhood multimorbidity

Methods:
The student will carry out a literature search to identify existing multimorbidity classifications using PubMed, Google Scholar and selected websites (for grey literature) to meet objective 1. For objective 2, the student will use Hospital Episode Statistics (a database of all hospital contacts in England) and the Clinical Practice Research Datalink (primary care records for ~14 million active patients in the UK, which includes linkage between mothers and babies) to develop and validate an unsupervised machine learning algorithm to characterise groups of children with similar longitudinal multimorbidity profiles. For objectives 3 &4, the student will use the algorithm to predict mortality or other health outcomes among children with multimorbidity, and examine risk factors for multimorbidity, such as maternal long-term conditions or birth characteristics. Throughout the PhD, the student will work with Great Ormond Street Hospital (GOSH) and the Luton Clinical Commissioning Group, to ensure the algorithm is applicable in different NHS settings.

This is an excellent opportunity for an individual with an MSc in medical statistics, computer science or other quantiative discipline to gain experience in applying machine learning methods to large NHS datasets, and carry out research to support children with complex conditions and their families.

References:
1.  Academy of Medical Sciences. Multimorbidity: a priority for global health research. 2018. https://acmedsci.ac.uk/file-download/82222577.
2.  Hardelid P, Dattani N, Gilbert R. Estimating the prevalence of chronic conditions in children who die in England, Scotland and Wales: a data linkage cohort study. BMJ Open. 2014;4(8):e005331.
3.  Feudtner C, Feinstein JA, Zhong W, et al. Pediatric complex chronic conditions classification system version 2: updated for ICD-10 and complex medical technology dependence and transplantation. BMC pediatrics. 2014;14:199-.
4.  Kuan V, Denaxas S, Gonzalez-Izquierdo A, et al. A chronological map of 308 physical and mental health conditions from 4 million individuals in the English National Health Service. The Lancet Digital Health. 2019;1(2):e63-e77.