Identifying clinically important COPD sub-types using data-driven approaches in primary care population based electronic health records

Abstract

Background COPD is a highly heterogeneous disease composed of different phenotypes with different aetiological and prognostic profiles and current classification systems do not fully capture this heterogeneity. In this study we sought to discover, describe and validate COPD subtypes using cluster analysis on data derived from electronic health records. Methods We applied two unsupervised learning algorithms (k-means and hierarchical clustering) in 30,961 current and former smokers diagnosed with COPD, using linked national structured electronic health records in England available through the CALIBER resource. We used 15 clinical features, including risk factors and comorbidities and performed dimensionality reduction using multiple correspondence analysis. We compared the association between cluster membership and COPD exacerbations and respiratory and cardiovascular death with 10,736 deaths recorded over 146,466 person-years of follow-up. We also implemented and tested a process to assign unseen patients into clusters using a decision tree classifier. Results We identified and characterized five COPD patient clusters with distinct patient characteristics with respect to demographics, comorbidities, risk of death and exacerbations. The four subgroups were associated with 1) anxiety/depression; 2) severe airflow obstruction and frailty; 3) cardiovascular disease and diabetes and 4) obesity/atopy. A fifth cluster was associated with low prevalence of most comorbid conditions. Conclusions COPD patients can be sub-classified into groups with differing risk factors, comorbidities, and prognosis, based on data included in their primary care records. The identified clusters confirm findings of previous clustering studies and draw attention to anxiety and depression as important drivers of the disease in young, female patients. Electronic supplementary material The online version of this article (10.1186/s12911-019-0805-0) contains supplementary material, which is available to authorized users.

Identifying damage clusters in patients with systemic lupus erythematosus

“…It creates the most statistically valid clusters, minimizing distances within clusters and maximizing distances between clusters, such that the subgroups are based only on the data themselves, without intervention of the researcher . Recently, the feasibility of cluster analyses for untangling disease heterogeneity and identifying subgroups has gained interest …”

A Machine Learning Approach for Studying the Comorbidities of Complex Diagnoses

“…Although this does not mean that it is free of limitations and biases, this method is currently being applied in many health sciences fields. In this particular concept of grouping patients according to their diagnoses, though, a large number of studies on chronic obstructive pulmonary disease (COPD) stands out [3,[5][6][7][8]. With these studies, knowledge of diagnoses associated with COPD has not only improved, but it has also allowed an improvement of the statistical methodology to assist studies regarding high-dimensionality healthcare data.…”

“…With diagnostic variables, which are those that raise our concerns, we often find all of these problems: a large number of dichotomous and possible unnecessary variables and, very likely, high collinearity. The latter is especially important, since it could dominate patient assignments into clusters [5].…”

“…These problems can be addressed by applying dimensionality reduction techniques, which aim to preserve the main structure in the data while reducing its dimensionality to a low-dimensional projection [12]. Theses methods allow us to: (a) easily discover hidden structures in the data, enabling an easier representation of them, (b) simplify the dataset by reducing the number of variables, and (c) use the outcome of orthogonal (and, therefore, uncorrelated) vectors as an input for clustering algorithms, thus avoiding collinearity issues [5].…”

“…They both project the data such that the new coordinate system best preserves the variance in the data [13], the main difference between them beingthat the former is applied to to quantitative data while MCA is intended for categorical variables. They are both very prevalent in literature, especially on COPD comorbidity studies [5][6][7]14] and gene expression research [15].…”