No full text
Unpublished conference/Abstract (Scientific congresses and symposiums)
Comprehensive Clustering Analysis for Incomplete COPD Dataset
Nekoee Zahraei, Halehsadat; LOUIS, Renaud; Donneau, Anne-Françoise
2019Journée des Doctorants de l’Ecole Doctorale « SPSS »
 

Files


Full Text
No document available.

Send to



Details



Keywords :
Chronic obstructive pulmonary disease (COPD); Consensus Clustering; Multiple Imputation; Cluster Analysis; Factor Analysis
Abstract :
[en] Introduction : Chronic obstructive pulmonary disease (COPD) is a complex, multidimensional and heterogeneous disease. In the past years, application of classification methods in COPD context has been developed based on clinical observation with a limited number of variables on incomplete dataset with missing values. In such studies, selection of variables included in the analysis and dealing with missing data have a high effect on results. The main purpose of this study is to identify clinical phenotypes among adults. In this study, missing data and dimension-reduction, which are present in any large dataset of observational data, were handled. Méthodologie : In this application, 178 patients were described by 86 mixed and huge sets of variables. A common occurrence in clinical study is missing value. In various literature, we can find many methods to deal with missing value. Among several methods for imputing missing values, multiple imputation is widely used to handle missing data. A very limited number of studies combining these two important issues, cluster analysis and multiple imputation. Therefore, in this study, difficulties of multiple imputing missing values in cluster analysis is characterized and in final step, patients are classified into homogeneous distinct groups. After imputation step, the methodology of HCPC (Hierarchical Clustering on Principal Components) is used. Factor analysis of mixed data (FAMD) is applied for reducing the complexity of huge dimensional data. After this step, hierarchical clustering is performed using Ward's criterion on the selected principal components. In the final step, consensus clustering is used to assign each individual to cluster. All statistical analyses were performed using R software. Résultats : Three different phenotypes were defined in COPD. These clusters were identified as: phenotype 1 included women with moderate COPD and mild atopic traits (n=65). phenotype 2 comprised severe men with exacerbation-prone, bacterial colonization/neutrophilic and systemic inflammation (n=52) and phenotype 3 included men moderate COPD with emphysema (n=61). Conclusions : In the past years, classification methods in COPD have been applied based on ignoring missing values with limited or selected number of variables. In this study, these two issues are solved. Then, with advanced statistical methods, patients are divided into three distinct clusters. These clinically meaningful clusters of patients with common characteristics can be used to predict outcomes of patients with COPD, to aid in development of personalized therapy.
Disciplines :
Public health, health care sciences & services
Author, co-author :
Nekoee Zahraei, Halehsadat ;  Université de Liège - ULiège > Département des sciences de la santé publique > Biostatistique
LOUIS, Renaud ;  Centre Hospitalier Universitaire de Liège - CHU > Département des Services Logistiques > Secteur gardiennage
Donneau, Anne-Françoise ;  Université de Liège - ULiège > Département des sciences de la santé publique > Biostatistique
Language :
English
Title :
Comprehensive Clustering Analysis for Incomplete COPD Dataset
Publication date :
03 December 2019
Event name :
Journée des Doctorants de l’Ecole Doctorale « SPSS »
Event organizer :
Université catholique de Louvain | UCLouvain
Event place :
Louvain, Belgium
Event date :
3-12-2019
Available on ORBi :
since 28 October 2019

Statistics


Number of views
130 (15 by ULiège)
Number of downloads
0 (0 by ULiège)

Bibliography


Similar publications



Contact ORBi