In an ageing population, and higher survival rates for many conditions more people will develop and live with multiple diseases, leading to decreased quality of life, increased numbers of prescribed medications and higher health care costs. Currently, most clinical guidelines and research focus on single diseases in isolation. This makes it difficult for clinicians to treat patients with co-existing diseases as there is insufficient evidence of which medications for one condition may be beneficial or harmful for others. We propose to study disease patterns and clusters in the UK population to identify key combinations of illnesses which lead to severe adverse outcomes such as death. Patients with or at high risk of these illness clusters can provide the basis for future clinical trials to determine the efficacy and safety profiles of medications in the multiple disease setting. These disease patterns and clusters might also help predict which patients have a higher risk of mortality. This work will help scope the possibility of preventing progression to severe disease outcomes by prescribing the right medications at the right time for the right patients, helping to address the feasibility of the emerging stratified / precision medicine agenda.
We aim to describe disease patterns and within-person disease clusters in a representative sample of the UK population using electronic health records from 1997 to 2010 in the CALIBER (CArdiovascular research using LInked Bespoke studies and Electronic health Records) programme. Diseases will be aggregated into broad diagnosis groups modified from the Clinical Classifications Software (CCS) categorisation scheme developed by the Agency for Healthcare Research and Quality (AHRQ). Single and co-occurring prevalence rates for major diseases will be described by age, gender, ethnicity and deprivation index. Comorbidity scores that quantify the strength of disease co-occurrence will be calculated for disease pairs. Using the comorbidity score and p values based on the Benjamini-Hochberg false discovery rate, we will rank and select disease pairs which capture the highest correlations between different disorders. Patients will be stratified based on the similarity of their disease classifications using novel unsupervised machine learning methods involving clustering algorithms. This will allow us to identify the key disease characteristics in each cluster. We will then investigate which combination of key diagnoses lead to severe outcomes in the disease clusters using network analytic measures such as connectivity and lethality.
Multi-morbidity and related outcomes (including mortality)
Aroon Hingorani - Chief Investigator - University College London ( UCL )
Valerie Kuan - Corresponding Applicant - University College London ( UCL )
Caroline Dale - Collaborator - University College London ( UCL )
Eda Bilici Ozyigit - Collaborator - University College London ( UCL )
Harry Hemingway - Collaborator - University College London ( UCL )
Jorgen Engmann - Collaborator - University College London ( UCL )
Juan Pablo Casas Romero - Collaborator - University College London ( UCL )
Matias Fuentealba - Collaborator - University College London ( UCL )
Rini Veeravalli - Collaborator - University College London ( UCL )
Rohan Takhar - Collaborator - University College London ( UCL )
Samuel Kim - Collaborator - University College London ( UCL )
Spiros Denaxas - Collaborator - University College London ( UCL )
Stefanie Mueller - Collaborator - University College London ( UCL )
Alireza Moayyeri - Collaborator - UCB Celltech
David Prieto-Merino - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
HES Admitted Patient Care;ONS Death Registration Data;Patient Level Index of Multiple Deprivation;Practice Level Index of Multiple Deprivation