Evidence from real world data
Interactive analysis and visualization
The aim of this large-scale population study based on real-world data from the region of Aragon (Spain), was to analyze, visualize and compare the multimorbidity patterns of two key diseases in the aging process such as heart failure (HF) and chronic obstructive pulmonary disease (COPD) in men and women using network analysis.
Our findings could help us to better understand the multimorbidity and clinical epidemiology of HF and COPD, especially with respect to four key aspects: i) by facilitating the analysis and visualization of this project in an interactive and visual way; ii) by identifying specific comorbidities that could be considered key for each index disease in terms of their position and influence in the network; iii) by describing the existence of clusters. The identification of COPD- or HF- specific disease clusters and highly connected diseases might be useful for the design of disease prevention and management strategies; and iv) by studying the role of sex in the multimorbidity of these index conditions, in line with the European recommendations of including the sex perspective in all publicly funded research projects.
We aimed to compare the multimorbidity patterns of HF and COPD in men and women using network analysis. Individuals aged 40 years or older on 2015 of the EpiChron Cohort (Aragon, Spain) were stratified by sex and as having COPD (n= 28,608), HF (n= 13,414), or COPD and HF (n= 3,952). We constructed one network per group by obtaining age-adjusted phi correlations between comorbidities
We conducted a cross-sectional, observational study in the EpiChron Cohort (Prados-Torres et al., 2018). This cohort links demographic and clinical anonymized information based on real-world data from the electronic health records (EHRs) and clinical-administrative databases of public health system users of the Spanish region of Aragón (1.3 M inhabitants). We included in this study all individuals aged 40 years and older on January 1, 2015 (n= 673,059), and we selected those patients with a diagnosis of COPD and/or HF in their primary and/or hospital care EHRs. We stratified the study population by sex and as having COPD (n= 28,608), HF (n= 13,414), or COPD+HF (n= 3,952). The Clinical Research Ethics Committee of Aragón (CEICA) approved the research protocol of this study (PI16/0136) and waived the requirement to obtain informed consent from patients since the information used was anonymized.
For each patient, we studied their sex, age, and all the chronic conditions registered in their primary and/or hospital care EHRs in the index date. Diseases were coded according to the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), or the International Classification of Primary Care (ICPC), and then grouped into 260 mutually exclusive Expanded Diagnostic Clusters (EDCs) through the Johns Hopkins ACG System (version 11.0, The Johns Hopkins University, Baltimore, MD, US). We included in the study the list of 114 EDCs defined as chronic in the study by Salisbury et al. (Salisbury et al., 2011), which has been validated in previous studies on multimorbidity (Poblador-Plou et al., 2014; Prados-Torres et al., 2012). Comorbidities were included in the analysis as binary variables (i.e., absence/presence).
In the population 40 years of age and older of the EpiChron Cohort, the prevalence of COPD, HF and COPD+HF was 4.25%, 1.99%, and 0.59%, respectively. The demographic characteristics with each index condition (COPD, HF, or COPD+HF) are shown above. Women were almost twice as frequent than men in the population with HF (64% vs. 32% and 37% in the COPD and COPD+HF groups, respectively). The HF group was on average 10 years older than the COPD group (80.3 ± 11.0 vs. 70.7 ± 12.1 years) and presented a higher mean number of comorbidities (8.1 vs. 5.4 conditions). Hypertension, dyslipidemia, arthritis, and diabetes mellitus appeared within the most frequent chronic conditions in all groups.
We performed network analysis to construct one network per index condition (COPD, HF, or COPD+HF) and sex, with a total of six networks. To do this, we obtained age-adjusted phi correlations between binary diagnostic variables. We performed multiple correlation testing (p<0.01) to correct family wise-error rate due to multiple comparisons. Only diseases with 10 or more cases were included in the multiple correlation analysis. In the networks, each node represents a disease, and the links or edges mean statistically significant correlations between the nodes.