Our study aims to characterise known and novel clusters of multimorbidity (known and novel) in common allergic disorders (asthma and atopic eczema).
We will identify anonymised records of adults with asthma or eczema in CPRD-GOLD using algorithms based on morbidity codes in primary care data. We will apply data-driven techniques to identify disease clusters (multimorbidity, defined using Read code chapters) in people with common allergic diseases, including previously described morbidity (e.g., cardiovascular diseases, fractures). This exploratory research will offer insights into pathways to multimorbidity, potentially allowing us to identify novel disease clusters not previously linked to allergic conditions (i.e., non-random clustering), highlighting avenues for subsequent research.
We will use logistic regression with individual level data to derive associations (Odds Ratios) between comorbidities. By modelling individuals together (people with asthma and atopic eczema separately [cases] matched on age, sex and practice with individuals without these conditions [control]) and including a comorbidity variable as the outcome (e.g. cardiovascular) and another comorbidity variables as exposures (e.g. respiratory, renal, etc.). By including an interaction term between exposure comorbidity and individual case-control status, we will test if associations between comorbidities varies across groups. We will visualise associations between comorbidities as networks in each patient group and identify clusters of highly associated comorbidities within networks. Clusters will be reviewed by subject-matter experts and patients for clinical plausibility. Clusters might suggest temporally-ordered sequences of diagnoses that we will visualise to illustrate disease progression over time. Logistic models can be adapted to estimate associations of sequence of events (e.g. A -> B) by defining appropriately event B as the outcome and only preceding A events as exposures. Sequential associations can be visualised using directed networks. We will compare networks and clusters between groups with a differential analysis of the edges and comparing overall network metrics (centrality, connectivity, etc.).
Our final outcomes summaries are associations between comorbidities- multiple health conditions including allergic and non-allergic morbidity (with and without directionality) and patterns of comorbidity clusters, which we will compare between exposed and unexposed groups that can vary depending on asthma and atopic eczema status. To find these clusters, we will use population-level metrics derived from regression analyses on individual-level data.
Sinead Langan - Chief Investigator - London School of Hygiene & Tropical Medicine ( LSHTM )
Amy Mulick - Corresponding Applicant - London School of Hygiene & Tropical Medicine ( LSHTM )
Alasdair Henderson - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
Aziz Sheikh - Collaborator - University of Edinburgh
David McAllister - Collaborator - University of Glasgow
David Prieto-Merino - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
Dorothea Nitsch - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
Jennifer Quint - Collaborator - Imperial College London
Kathryn Mansfield - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
Liam Smeeth - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
Ronan Lyons - Collaborator - Swansea University
Spiros Denaxas - Collaborator - University College London ( UCL )