Dépôt numérique

Comparison of Grouping Methods for the identification of Asthma Phenotypes in the Quebec Birth Cohort on Immunity and Health

Medisor, Miceline; Benedetti, Andrea; Rousseau, Marie-Claude . Comparison of Grouping Methods for the identification of Asthma Phenotypes in the Quebec Birth Cohort on Immunity and Health In: Society for Epidemiological Research’s 2016 Epidemiology Congress of the Americas, 21-24 Juin 2016, Miami.

Ce document n'est pas hébergé sur EspaceINRS.


Common among children, asthma is a chronic disease which also affects adolescents and young adults. Its manifestations are diverse and the identification of asthma phenotypes may contribute to elucidating its etiology. In this study, we identified asthma phenotypes during three periods of life (childhood, adolescence, young adulthood) using three different statistical approaches and compared the results. The Quebec Birth Cohort on Immunity and Health includes 81,496 individuals born in 1974 among whom health services (medical services and hospitalizations) for asthma and allergic diseases were documented from 1983 to 1994. Individuals who had at least one medical service or hospitalization for asthma during follow-up were included in the analysis (n=9989, 12% of study population). Three grouping methods were used: k-means clustering, Ward’s method and latent profile analysis. Variables considered for phenotype identification included an indicator of asthma (Y/N) and an indicator of first asthma-related health service (Y/N) during the period, the number of medical services and of hospitalizations for asthma, and the numbers of health services for rhinitis, eczema, urticaria, and other allergies respectively. In the analytical sample, 40% of the subjects had at least one health service for asthma during childhood (8-12 years), 55% during adolescence (13-17 years), and 45% in young adulthood (18-20 years). In general, a larger number of phenotypes was identified with k-means (9 in each period), an intermediate number with Ward (6 or 7), and a smaller number with latent profiles (5 or 6). Some of the phenotypes were identified with the three methods, although with different proportions of subjects. K-means and Ward’s methods resulted in similar findings, whereas results from latent profiles tended to differ. Our results underscore the importance of grouping method selection when studying asthma phenotypes, and suggest that the robustness of results should be evaluated.

Type de document: Document issu d'une conférence ou d'un atelier
Mots-clés libres: -
Centre: Centre INRS-Institut Armand Frappier
Date de dépôt: 25 juin 2018 16:37
Dernière modification: 25 juin 2018 16:37
URI: http://espace.inrs.ca/id/eprint/5869

Actions (Identification requise)

Modifier la notice Modifier la notice