Multistate clinical prediction models in renal replacement therapy

Lead Research Organisation: University of Manchester
Department Name: School of Health Sciences


A clinical prediction model (CPM) is a tool to predict future risk for patients. Existing CPMs are generally based on a single outcome or decision point. For example, the future risk of cardiovascular events or death during a surgical procedure. However, in many situations we are faced with a complex sequence of events and decisions, so we would like to predict multiple outcomes to optimise treatment. One such example is renal replacement therapy (RRT), which is a general term for treatments available to patients with kidney failure. These patients are faced with many treatment options (including transplant and dialysis) and may move between different treatments over time. Statistically,Ttis can be characterised by a multi-state model. Statistical methodology for clinical prediction to inform optimal treatment is underdeveloped in this area.

It is expected that this improvement to methodology for clinical prediction in multi-state models has the potential to improve treatment, particularly in RRT.

Our objectives are to:
1. Review literature of CPMs and multi-state modelling to identify the potential to apply CPMs in a multi-state framework.
2. Develop a multi-state model to describe the flows of patients through treatments in RRT, using data from the UK Renal Registry, which contains data on all patients in the UK on RRT.
3. Use the developed multi-state model to define aetiological drivers of outcomes in patients on renal replacement therapy (RRT)
4. Use the developed multi-state model, combined with CPM methodology, to derive a 'decision aid' for optimising treatment for RRT patients.
5. Externally validate the developed decision aid/CPM on international renal registries (e.g. Australia/New Zealand, Sweden).

We expect the outcomes/impact of the project to be:
Provide a unique insight into the current 'journey' of RRT patients through multi-state modelling.
Provide a tool which can be used to optimise dynamic treatment regimens for these complex patients to improve prognosis.
Statistical methodology, particularly around the link between predictive modelling and decision support, that will see wider impact in other disease and application areas.


10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/N509565/1 01/10/2016 30/09/2021
1790838 Studentship EP/N509565/1 01/10/2016 31/12/2020 Michael Barrowman