Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.665465
Title: Multiple imputation of missing data and prognostic survival modelling for incident patients starting dialysis in England, Wales and Northern Ireland
Author: Steenkamp, Margaretha
ISNI:       0000 0004 5349 4921
Awarding Body: University of Bristol
Current Institution: University of Bristol
Date of Award: 2014
Availability of Full Text:
Access through EThOS:
Abstract:
Background The UK Renal Registry (UKRR) collects data, performs clinical audit and reports on achievement of clinical standards and outcomes for patients on renal replacement therapy (RRT). Mortality predictions for patients starting dialysis are important, especially for elderly patients with an increased risk of death. Studies have shown that patients who are considering starting dialysis, do want prognostic information to aid informed decision making. The main aim of the prognostic model developed in this research, is to provide realistic and objective prognostic information to clinicians when discussing prognosis with patients in preparation for starting dialysis. Methods The research described in this thesis investigates the application of multiple imputation by chained equations for handling of missing data in the UKRR database and the development of the prognostic model in the 20 complete imputed data sets. Internal validation was based on a non-randomly geographical split of renal centres and the prognostic model was externally validated in the Australia and New Zealand Dialysis and Transplant Registry (ANZDATA) data. Main results Prognostic model discrimination at 1 year after start of dialysis was good (c-statistic=0.79). The results from the prognostic model in terms of calibration, discrimination and variance explained did not dilute much over follow-up time. The geographical internal validation showed very good calibration and discrimination with little degradation of model fit compared to the full dataset. Although the predictive ability of the externally validated prognostic sub-model in the ANZDATA cohort was reasonable, calibration . in the 2 most at risk patient groups was not adequate. Recommendations The transportability of the prognostic model developed in the UKRR data was not proven in the external validation and it is not advisable to use the prognostic model in populations other than the UK. The prognostic model will need to be calibrated before use in other populations.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.665465  DOI: Not available
Share: