UCL Institute of Cardiovascular Science


Disease Patterns

This work package investigates the sequence and patterns of co-occurrence of multiple conditions in the same individual using data from linked electronic health records.

With anonymised data on many individuals obtainable from linked electronic health records across the life course we are able to collate age specific incidence patterns for around 1000 index diseases.  With this information we are able to identify which diseases co-occur in the same person by chance, as well as to identify which co-occur in the same person at rates higher than expected by chance.  Diseases that co-occur in the same individual at rates consistent with what you would expect by chance often do not share the same cause since they occur independently, but understanding these will help in planning service provision and help to estimate risks of adverse outcomes due to  polypharmacy (use of many medicines by a single individual). 

Diseases co-occurring at rates higher than you would expect by chance would likely share common causes such as common environmental causes (e.g. smoking triggering coagulation and DNA damage, causing stroke and cancer respectively;  one condition causing another (e.g. atrial fibrillation leading to ischaemic stroke); the treatment of one condition causing another (e.g. anticoagulation for pulmonary embolism causing subdural haemorrhage); or from common genetically determined mechanisms.  Identifying these inform treatment plans and suggest diseases that can be targeted by the same drugs.  This work builds on similar research on 308 conditions defined from Read codes in the Clinical Practice Research Data Link, ICD-10 codes and OPCS-4 codes from Hospital Episode Statistics that was released via the CALIBER data portal, now part of the HDRUK Phenotype library.  Insights will be contributed to the Health Data Research UK (HDRUK) national implementation projects in Phenomics and Multimorbidity.