Fig. 3

Hierarchical clustering and correlation matrix with significant cardiometabolic biomarkers. A heatmap (top left) and correlation matrix (top right and bottom) for the 31 protein biomarkers significantly associated with ICU/death (P < 0.05/116 hospital laboratory tests and biomarkers = 4 × 10^–4). The correlation matrix shows how the protein biomarkers, ordered based on hierarchical clustering, correlate with one another (top right) and how they correlate with the demographic factors, clinical variables, and hospital laboratory tests (bottom). The color reflects the magnitude and direction of the Pearson correlation coefficient. The cells corresponding to correlations with P > 0.05 were left blank. The P values and odds ratios (OR) reported for the association of each variable with ICU/death are the same as those shown in Fig. 2. Box A shows the association of the largest cluster, comprised of 16 biomarkers, with type 2 diabetes, chronic kidney disease (CKD), and cardiac disease. Boxes B and C show how this cluster correlates with the hospital labs. Finally, Box D shows correlations between the hospital laboratory tests and a smaller cluster, comprising the five biomarkers that were negatively associated with ICU/death. SD Standard deviation, CI confidence interval, AA African American, COPD chronic obstructive pulmonary disease, CAD coronary artery disease, HFpEF heart failure with preserved ejection fraction, HFrEF heart failure with reduced ejection fraction, BUN blood urea nitrogen, ERS erythrocyte sedimentation rate, LDH lactate dehydrogenase, AST aspartate aminotransferase, WBC white blood cells, CRP C-reactive protein, ALC absolute lymphocyte count, eGFR estimated glomerular filtration rate