Skip to main content
Fig. 2 | Journal of Intensive Care

Fig. 2

From: Advanced echocardiographic phenotyping of critically ill patients with coronavirus-19 sepsis: a prospective cohort study

Fig. 2

Hierarchical clustering (a) and matrix correlation (b) of contractility and loading condition indices in critically ill patients with coronavirus-19 sepsis. In a, the parameters were reordered using computerized hierarchical clustering with the corrplot package of R statistical environment. Hierarchical clustering is a statistical method for finding comparatively homogeneous clusters of cases based on measured characteristics. The analysis starts with each case as a separate cluster (i.e., there are as many clusters as cases), and then combines the clusters sequentially, reducing the number of clusters at each step. The clustering method uses the dissimilarities between objects. The algorithm uses a set of dissimilarities or distances between cases when constructing the clusters and proceeds iteratively to join the most similar cases. Distances between clusters were recomputed by the Lance-Williams dissimilarity update formula according to the complete linkage method. In b, the four big squares drawn in the chart are based on the results of hierarchical clustering and each contains the members of a cluster (LV afterload cluster in the upper-left corner, cardiac preload cluster in the middle upper-left, LV contractility cluster in the middle lower-right, and RV function cluster in the lower-right corner). Numbers and the blue-white-red color spectrum denote the Spearman correlation coefficients (with Benjamini-Hochberg correction to control the false discovery rate at 0.05 level); positive correlations are represented in a blue scale; negative correlations are in a red scale. The surface areas of the colored pixels and their color intensity show the absolute value of corresponding correlation coefficients; non-significant coefficients are left blank. There was a strong correlation between most indices within the LV contractility cluster (blue pixels in the middle lower-right cluster) and within the LV afterload cluster (blue pixels in the upper-left cluster). In addition, some LV contractility indices were negatively correlated with an afterload parameter (red pixels above and to the left of the middle lower-right cluster), and positively correlated with RV function indices (blue pixels below and to the right of the middle lower-right cluster), but not with preload indices. IVC, maximal diameter of inferior vena cava in mm; EA, ratio of early to late diastolic wave velocity at the mitral valve; Ee, ratio of early pulsed-wave Doppler to early tissue Doppler diastolic wave velocity at the lateral mitral valve annulus; EF, LV ejection fraction in %; AS, absolute values of global LV longitudinal peak systolic strain in %; sm, tissue Doppler peak systolic wave at lateral mitral annulus in cm s−1; VAC, ventricular-arterial coupling; ME, LV end-systolic maximal elastance in mmHg mL−1; AE, end-systolic arterial elastance in mmHg mL−1; SVR, systemic vascular resistance in mmHg L−1 min; DAP, diastolic arterial pressure in mmHg; PCD, pulmonary circulatory dysfunction; TAPSE, tricuspid annulus plane systolic excursion in mm; st, tissue Doppler peak systolic wave at tricuspid lateral annulus in cm s−1

Back to article page