HomeNovelNovel multimorbidity clusters in people with eczema and asthma: a population-based cluster analysis
Novel multimorbidity clusters in people with eczema and asthma: a population-based cluster analysis
December 18, 2022
We included 434,422 individuals with eczema (58% female, median age 47 years) and 1,333,281 matched controls without eczema (55% female, 47 years old) and 517,712 individuals with asthma (53% female, 44 years old) and 1,601,210 Matched controls without asthma (53% female, age 45) (Table 1, Supplementary Figure S2). Many controls (n = 696,623) were matched with both subjects with eczema (52% of total) and subjects with asthma (57% of total) and contributed to both analyses. Similarly, many people were identified as having both eczema and asthma (n = 115,685, representing 27% of people with eczema and 22% of people with asthma) and contributed to both analyses.
Excluding skin diseases (Read Chapter M), the most common morbidity codes ever recorded in individuals with eczema were for respiratory and nervous system conditions (Chapters H and F, 86% and 79%). For matched controls without eczema, including all chapters, were for respiratory and musculoskeletal conditions (H and N, 67% and 66%) (Table 1, Supplementary Fig. S3), although this varied by age and gender (Fig.1) .
Excluding respiratory diseases (Chapter H), the most common codes ever recorded for people with asthma were for skin and musculoskeletal conditions (Chapters M and N, 76% and 72%). Matched controls without asthma were the same but less prevalent (64% and 63%). All chapters were recorded more commonly in people with eczema and asthma than in their controls.
Comparison of multimorbidity
Summary statistics of 171 regression models (one for each pairwise combination of reading chapters) for each condition are available in the supplementary figures. S4-S5. People with eczema or asthma generally had their first diagnosis recorded from both reading chapters earlier than people without, more often experienced regressive outcomes from both diagnoses, and were followed up for a longer time (Supplementary Figures S4-S5 ). Gender was well balanced between populations with and without eczema and asthma.
We found larger multimorbidity clusters in people with eczema and people with asthma than in controls. Detailed images of networks and dendrograms are available in the supplementary figures. S6–S13.
Figure 2 summarizes 16 undirected networks of reading chapters by (modeled) age at first condition recorded, gender, and exposure/control status for the eczema and asthma groups. Reading chapters are pinned in the same place for all groups so that morbidity associations can easily be compared visually between groups. There was greater interconnectedness in the eczema and asthma populations than controls, in women than men, and in older than younger diagnostic ages.
The greatest connectivity between Read chapters was in women with eczema or asthma (between 30 and 37 borders, Panels ii, iv, x, xii), followed by men with eczema or asthma (9-24 borders, Panels i, iii , ix, xi), female controls (14-20, panels vi, viii, xiv, xvi), and male controls (2 or 3, panels v, vii, xiii, xv). In general, 50-year-olds with their first recorded condition had higher five-year multimorbidity than 18-year-olds of the same sex.
Analysis of hierarchical clusters
Table 2 and Fig. 3 present the results of the hierarchical cluster analyses. In Fig. 3, the closely related Read chapters are grouped together so that the clusters can be easily viewed.
In men with eczema, modeled with the first condition occurring at age 18, we identified a five-year multimorbidity cluster including cutaneous, respiratory, infectious/parasitic, and neurological conditions and injury/poisoning (see chapters M/H/ A/F/S ; Panel i). In their controls without eczema, we identified a smaller group containing only chapters for cutaneous, respiratory, and infectious/parasitic conditions (M/H/A; Panel v). When the first condition was modeled at age 50, we identified two clusters in men with eczema (panel iii) containing all conditions from the previous diagnostic age: infectious/parasitic diseases and injury/poisoning (A/S; cluster smaller) and skin, respiratory, and neurological conditions (M/H/F; largest cluster), as well as musculoskeletal, ill-defined, and digestive conditions (N/R/J) in the largest cluster. In their controls, we found a single, small group of cutaneous and musculoskeletal (M/N) conditions (panel vii).
In women with eczema modeled with the first condition occurring at age 18, we identified two clusters of multimorbidity at five years (Panel ii). The largest cluster contained many of the same five conditions as the clusters of men with eczema (skin, respiratory, infectious/parasitic, neurological, and injury/poisoning; chapters M/H/A/F/S), plus musculoskeletal and genitourinary disorders conditions (N/K). The smallest contained pregnancy/delivery/puerperium (E/L) and mental health conditions. In their controls we also identified two clusters but with different patterns (Panel vi). One was a subset of the larger eczema group, containing cutaneous, respiratory, and infectious/parasitic (M/H/A) conditions; the other contained pregnancy/delivery/puerperium, genitourinary, and musculoskeletal (L/K/N) conditions. When the first condition was modeled at age 50, we identified a single large cluster of nine reading chapters; Cutaneous, Respiratory, Infectious/Parasitic, Injury/Poisoning, Neurological, Musculoskeletal and Genitourinary (M/H/A/S/F/N/K, i.e. the same conditions that occur in the younger diagnostic age), so such as Defined Diseases Conditions and Disorders of the Digestive System (R/J) (Panel iv). In controls we found a larger cluster of multimorbidity than men or control women first diagnosed at age 18, but smaller than their matches, including skin, respiratory, neurological, musculoskeletal, genitourinary conditions and ill-defined (M/H/F /N/K/R) (Panel viii).
The clusters identified in people with asthma were very similar to those identified in people with eczema (panels ix-xii), and the clusters identified in controls with asthma were very similar to those identified in controls with eczema (panels xiii-xvi). Table 2 details the differences, all due to chapters or groupings of chapters falling just below the 30% probability of co-occurrence in eczema and just above in asthma, or vice versa.
For both the eczema and asthma populations and their controls, increasing the duration of follow-up generally increased the likelihood of each pair of co-occurrences, producing larger clusters and/or more at the 30% cutoff ( Complementary figures S14A-H and S15A–H). The switch to Ward’s linkage algorithms in the hierarchical clustering method had no material impact on the identified clusters (Suppl. Fig. S16).