Applying machine learning algorithms to datasets to predict outcome for paediatric solid organ transplant recipients

Predicting outcomes after paediatric solid organ transplantation is challenging. Machine learning (ML) models have been developed in order to address this in the large datasets now available in registries and those generated within single centre electronic health records (EHRs). Systematic review and meta-analyses of these models following kidney and lung transplantation, predominantly from adult patients, suggest that clinician predictions on outcomes can be enhanced by information from these models, and certain models can outperform clinicians. Due to the extensive data collection routinely taking place continually in the Great Ormond Street Hospital for Children NHS Foundation Trust electronic health record (>3,000 variables per patient post lung transplant), and the co-location of three paediatric solid organ transplant programmes on one site (kidney, heart, lung), we have the opportunity to both validate existing ML models and to determine new variables that may have superior
sensitivity and specificity when predicting future outcomes.

Hypothesis and/or Aims:

• To validate existing machine learning (ML) tools in paediatric-only registry datasets


• Using a series of ML reinforcement learning approaches, including deep learning, determine a novel model both in organ specific and in "all population" analyses.


• To develop software tools that can be updated prospectively as new patients go through the transplantation programme.


Research and Policy outputs:

• Systematic review of literature to date (aim to publish as review article)


• New insights into ML model approaches to small registry and hospital EHR data


• Software that can augment existing tools or be deployed into an EHR to aid clinical decision making.


References:

• Ravindhran B, Chandak P, Schafer N, Kundalia K, Hwang W, Antoniadis S, et al. Machine learning models in predicting graft survival in kidney transplantation: meta-analysis. BJS Open. 2023;7(2).



• Gholamzadeh M, Abtahi H, Safdari R. Machine learning-based techniques to improve lung transplantation outcomes and complications: a systematic review. BMC Med Res Methodol. 2022;22(1):331.



• Divard G, Raynaud M, Tatapudi VS, Abdalla B, Bailly E, Assayag M, et al. Comparison of artificial intelligence and human-based prediction and stratification of the risk of long-term kidney allograft failure. Commun Med (Lond). 2022;2(1):150.



• Lisboa PJG, Jayabalan M, Ortega-Martorell S, Olier I, Medved D, Nilsson J. Enhanced survival prediction using explainable artificial intelligence in heart transplantation. Sci Rep. 2022;12(1):19525.



• Ivanics T, So D, Claasen M, Wallace D, Patel MS, Gravely A, et al. Machine learning-based mortality prediction models using national liver transplantation registries are feasible but have limited utility across countries. Am J Transplant. 2023;23(1):64-71.
  
  Applicants should have, or expect to receive an upper second-class Bachelor's degree and a Master's degree (or equivalent work experience) in a relevant discipline or an overseas qualification of an equivalent standard.
  
  A 3-year PhD Studentship in healthcare data science funded by GOSH Children's Charity is available within University College London Great Ormond Street Institute of Child Health. The studentship will commence from September 2024 onwards, under the supervision of Prof Stephen Marks, Dr Rossa Brugha, and supported by Prof Mario Cortina Borja., The student will be based in the UCL Institute of Child Health, and will work alongside a team of Data Engineers and Data Scientists from the NHS, Academia, and Industry, through the Clinical Informatics Research Programme at GOSH DRIVE (https://www.goshdrive.com/)


The student will learn about all aspects of healthcare related "big data" including national registries and modern hospital records, systems for data sharing (including the OMOP common data model, FIHR), as well as testing and developing ML models on real world datasets, working alongside data scientists and clinicians. At the end of the PhD, we expect the student to be ready for independent work with healthcare data sets to develop tools that leverage large data resources to improve patient care.

This Studentship presents a unique opportunity to conduct supervised research at and be a part of the research community, being an integral part of the exciting and thriving research team.

This studentship provides a starting stipend of £21,237 per annum and covers the cost of tuition fees based on the UK (Home) rate. Non-UK students can apply but if they are not eligible for UK/Home fees status, will have to personally fund the difference between the UK (Home) rate and the Overseas rate.