Palmer, Edward S.; (2021) The Secondary Use of Longitudinal Critical Care Data. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text palmer-phd-thesis.pdf - Accepted Version Download (5MB) | Preview

Abstract

Aims To examine the strengths and limitations of a novel United Kingdom (UK) critical care data resource that repurposes routinely collected physiological data for research. Exemplar clinical research studies will be developed to explore the unique longitudinal nature of the resource. Objectives - To evaluate the suitability of the National Institute for Health Research (NIHR) Critical Care theme of the Health Informatics Collaborative (CCHIC) data model as a representation of the Electronic Health Record (EHR) for secondary research use. - To conduct a data quality evaluation of data stored within the CC-HIC research database. - To use the CC-HIC research database to conduct two clinical research studies that make use of the longitudinal data supported by the CC-HIC: - The association between cumulative exposure to excess oxygen and outcomes in the critically ill. - The association between different morphologies of longitudinal physiology—in particular organ dysfunction—and outcomes in sepsis. The CC-HIC The EHR is now routinely used for the delivery of patient care throughout the United Kingdom (UK). This has presented the opportunity to learn from a large volume of routinely collected data. The CC-HIC data model represents 255 distinct clinical concepts including demographics, outcomes and granular longitudinal physiology. This model is used to harmonise EHR data of 12 contributing Intensive Care Units (ICUs). This thesis evaluates the suitability of the CC-HIC data model in this role and the quality of data within. While representing an important first step in this field, the CC-HIC data model lacks the necessary normalisation and semantic expressivity to excel in this role. The quality of the CC-HIC research database was variable between contributing sites. High levels of missing data, missing meta-data, non-standardised units and temporal drop out of submitted data are amongst the most challenging features to tackle. It is the principal finding of this thesis that the CC-HIC should transition towards implementing internationally agreed standards for interoperability. Exemplar Clinical Studies Two exemplar studies are presented, each designed to make use of the longitudinal data made available by the CC-HIC and address domains that are both contemporaneous and of importance to the critical care community. Exposure to Excess Oxygen Longitudinal data from the CC-HIC cohort were used to explore the association between the cumulative exposure to excess oxygen and outcomes in the critically ill. A small (likely less than 1% absolute risk reduction) dose-independent association was found between exposure to excess oxygen and mortality. The lack of dosedependency challenges a causal interpretation of these findings. Physiological Morphologies in Sepsis The joint modelling paradigm was applied to explore the different longitudinal profiles of organ failure in sepsis, while accounting for informative censoring from patient death. The rate of change of organ failure was found to play a more significan't role in outcomes than the absolute value of organ failure at a given moment. This has important implications for how the critical care community views the evolution of physiology in sepsis. DECOVID The Decoding COVID-19 (DECOVID) project is presented as future work. DECOVID is a collaborative data sharing project that pools clinical data from two large NHS trusts in England. Many of the lessons learnt from the prior work with the CC-HIC fed into the development of the DECOVID data model and its quality evaluation.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item