Discovery and characterization of heart failure sub phenotypes using cluster analysis methods in CPRD

Date of Approval: 
2018-07-20 00:00:00
Lay Summary: 
Heart failure (HF) is a long-term condition that occurs when the heart muscle does not pump blood as effectively around the body as it should. It usually gets worse over time and has an impact on patient health that is equivalent or worse than many types of cancer. Most HF treatments focus on treating symptoms and slowing down disease progression and a growing body of research shows that some HF patients respond differently to available treatments. This is partly because HF is not a single simple disease but a complex and heterogeneous one that has many hidden sub types which make accurate diagnosis and treatment challenging. This study will investigate if it is possible to discover HF subtypes using an approach called cluster analysis which puts patients into groups based on their clinical characteristics (e.g. diagnoses, lab results). Patients in the same group (called a cluster) are more similar between them than with patients in other groups. Once the clusters have been defined, we will explore differences in mortality and hospitalisation to evaluate our findings. Cluster analysis has been used before for HF, but on a much smaller scale, with smaller patient datasets and less rich data.
Technical Summary: 
Heart failure (HF) is a heterogeneous and complex syndrome with many unrecognised/undefined subtypes that respond uniquely across the range of potential therapeutic interventions currently used. Current classification relies on broad aggregate measures which may lead to overlapping groups and potentially misclassification. Since therapeutic interventions are frequently based on targeting certain patient subgroups, inadequate classification may lead to ineffective/inappropriate treatments. Improving the taxonomy of HF classification may therefore offer important clinical benefits. Whereas molecular phenotyping might theoretically provide a more rational disease description, an essential first step is to identify disease sub-types based on clinical variables that are routinely recorded in a patient’s electronic health record. In this study, we will use an analytical approach named clustering analysis (clustering) to identify HF subtypes in CPRD data and evaluate findings using mortality and hospitalisation. Clustering is an exploratory approach which uses a pre-defined set of clinical features to group patients into groups. Clustering has been extensively and successfully used in many respiratory conditions and in smaller studies to identify disease subtypes. Applying it to larger, higher-resolution data sources with longitudinal measurements of disease risk factors, symptoms and progression such as CPRD could lead to improved identification, characterization and treatment of HF subtypes.
Health Outcomes to be Measured: 
Heart Failure
Application Number: 

Spiros Denaxas - Chief Investigator - University College London ( UCL )
Spiros Denaxas - Corresponding Applicant - University College London ( UCL )
Daniel Swerdlow - Collaborator - University College London ( UCL )
Ghazaleh Fatemifar - Collaborator - University College London ( UCL )
Tom Lumbers - Collaborator - University College London ( UCL )

HES Admitted Patient Care;ONS Death Registration Data;Patient Level Index of Multiple Deprivation