Heart failure is an increasing problem in the community with an ageing population. The treatment of heart failure has improved, resulting in fewer deaths. Consequently patients can remain on potent drugs for many years. This has its own problems - our hospital admission study showed that kidney damage caused by drugs used to remove fluids was the second most common adverse reaction (ADR) causing hospital admission. Potentially this is preventable through regular monitoring of how the patient’s kidneys are working. However there are no clear clinical guidelines on how or where (with the patient’s GP or in specialist clinics) this checking should be carried out. Even if population based guidelines were available, this is likely to be wasteful in terms of unnecessary testing for patients increasing the economic burden and also impairing the quality of life of patients.
We aim to develop personalised kidney function monitoring guidelines based on the characteristics of individual patients (for example: Age, surgical interventions, drugs being taken and other diseases that they might have). The availability of such monitoring protocols would change clinical care pathways, and potentially reduce hospital admissions, reduce cost and improve patient quality of life.
The main objective of this study is to inform the development of personalised renal function guidelines in heart failure patients. In order to facilitate a personalised approach, a data-driven method is required to analyse individual factors which amplify the risk of renal decline with diuretics.
The method will involve extraction of patient information from all patients with a diagnosis of heart failure and associated clinical codes. Within this patient group we will be looking at occurrence of renal impairment via change in eGFR and creatinine biomarkers. We will examine the frequency of renal function monitoring during the study observation period. These results will then be mapped to patient factors which may conceivably impact renal function. Therefore co-variates would include a full list of prescribed medications in primary care, co-morbidities, other relevant investigations and interventions.
Due to the large number of estimated variables, we will incorporate machine learning algorithms for guideline generation. In particular, but not exclusively, we will combine Gaussian processes and change-point analyses as well as penalized linear methods to identify key predicting variables and time-points. This methodology will allow us to discover patient factors hitherto unexplored which impact patient risk of renal decline with medication, creating a truly personalised guidance to improve patient outcomes.
Darren Ashcroft - Chief Investigator - University of Manchester
Ahmed Al-Naher - Corresponding Applicant - University of Liverpool
Bertram Muller-Myhsok - Collaborator - University of Liverpool
Heather A Robinson - Collaborator - University of Manchester
Jennifer Downing - Collaborator - University of Liverpool
Munir Pirmohamed - Collaborator - University of Liverpool
Niels Peek - Collaborator - University of Manchester
Paolo Fraccaro - Collaborator - University of Manchester
Tjeerd van Staa - Collaborator - University of Manchester
Patient Level Index of Multiple Deprivation