Figure 1
The flow chart of study population enrollment.
Table 1
Baseline characteristics between non-AKI group and AKI group.
| VARIABLE | NON-AKI (N = 1196) | AKI (N = 300) | TOTAL (N = 1496) | p-VALUE |
|---|---|---|---|---|
| General information | ||||
| Male, n (%) | 774 (65) | 216 (72) | 990 (66) | 0.021 |
| Age, y | 67 (63, 72) | 68 (64, 74) | 67 (63, 72) | 0.05 |
| History | ||||
| Smoking history, n (%) | 310 (26) | 84 (28) | 394 (26) | 0.51 |
| Valvular Heart Disease, n (%) | 521 (44) | 124 (41) | 645 (43) | 0.528 |
| Cerebral stroke, n (%) | 76 (6) | 30 (10) | 106 (7) | 0.038 |
| NYHA Class IV heart failure, n (%) | 154 (13) | 58 (19) | 212 (14) | 0.006 |
| Gout, n (%) | 48 (4) | 35 (12) | 83 (6) | <0.001 |
| Hypertension, n (%) | 395 (33) | 109 (36) | 504 (34) | 0.31 |
| Diabetes, n (%) | 247 (21) | 85 (28) | 332 (22) | 0.005 |
| Rheumatism, n (%) | 10 (1) | 5 (2) | 15 (1) | 0.199 |
| Parameters | ||||
| Serum Inorganic phosphorus, mmol/L | 1.18 (1.06, 1.32) | 1.2 (1.07, 1.45) | 1.18 (1.06, 1.34) | 0.006 |
| Urea, mmol/L | 6.9 (5.4, 8.9) | 9.93 (7.08, 15.03) | 7.3 (5.6, 9.81) | <0.001 |
| Uric acid, μmol/L | 449.1 (361.39, 551.7) | 534.05 (417.8, 662.92) | 464.65 (373, 575.02) | <0.001 |
| LDH, U/L | 198 (169, 237) | 230.9 (189.75, 291.25) | 204 (171, 249) | <0.001 |
| CK, U/L | 78 (55, 115) | 91 (60.83, 138.5) | 80 (56, 120) | <0.001 |
| CK-MB, U/L | 10.2 (7.77, 13.9) | 11.1 (8.4, 15.5) | 10.25 (7.88, 14.1) | 0.023 |
| AST, U/L | 26 (20.17, 34) | 28 (22, 43) | 26 (21, 35) | <0.001 |
| TC, mmol/L | 4.37 (3.67, 5.13) | 3.99 (3.24, 4.69) | 4.3 (3.59, 5.08) | <0.001 |
| LDL-C, mmol/L | 2.71 (2.2, 3.31) | 2.6 (1.97, 3.12) | 2.69 (2.16, 3.29) | 0.001 |
| ApoA-I, g/L | 1.13 (0.94, 1.32) | 1.04 (0.85, 1.19) | 1.12 (0.93, 1.3) | <0.001 |
| HDL-C, mmol/L | 1.03 (0.85, 1.23) | 0.94 (0.76, 1.13) | 1.02 (0.84, 1.21) | <0.001 |
| ADA, U/L | 10.4 (8.5, 13.22) | 11.7 (8.9, 14.72) | 10.7 (8.5, 13.7) | 0.005 |
| GGT, U/L | 43 (26, 76.88) | 49 (31, 103) | 44 (27, 81) | <0.001 |
| TBIL, μmol/L | 16.5 (12.2, 23.8) | 18.7 (12.6, 29.15) | 16.85 (12.3, 24.6) | 0.002 |
| TP, g/L | 65.52 ± 6.94 | 64.79 ± 6.8 | 65.37 ± 6.91 | 0.101 |
| TSH, mIU/L | 1.59 (0.94, 2.54) | 1.79 (1.06, 3.21) | 1.63 (0.94, 2.67) | 0.009 |
| NT-proBNP, pg/mL | 2363 (1005.75, 5343.25) | 5712 (2335.75, 15659) | 2917.5 (1172.5, 6308) | <0.001 |
| hs-cTn, ng/L | 20.4 (11.97, 36.28) | 37.35 (19.07, 69.17) | 22.3 (12.68, 42.7) | <0.001 |
| Neutrophil count, 109/L | 5.1 (3.87, 6.58) | 4.74 (3.63, 6.47) | 4.99 (3.82, 6.53) | 0.066 |
| PDW, % | 15.1 (12.2, 16.6) | 15.85 (12.6, 16.72) | 15.4 (12.3, 16.62) | 0.045 |
| Plateletcrit, % | 1.9 (1.58, 2.4) | 1.83 (1.43, 2.3) | 1.9 (1.52, 2.4) | 0.004 |
| Platelet count,109/L | 195 (158, 238) | 188 (146.45, 226) | 194 (156, 236) | 0.016 |
| WBC count, 109/L | 6.85 (5.68, 8.23) | 7.03 (5.56, 8.48) | 6.87 (5.67, 8.29) | 0.523 |
| RDW, % | 0.14 (0.13, 0.15) | 0.14 (0.13, 0.16) | 0.14 (0.13, 0.15) | 0.028 |
| RBC count, 1012/L | 4.45 (4.05, 4.89) | 4.38 (3.98, 4.84) | 4.44 (4.02, 4.89) | 0.07 |
| Lymphocyte count, 109/L | 1.57 (1.19, 2.05) | 1.52 (1.17, 2.06) | 1.57 (1.19, 2.06) | 0.606 |
| Eosinophil count, 109/L | 0.12 (0.06, 0.21) | 0.14 (0.06, 0.24) | 0.12 (0.06, 0.22) | 0.158 |
| TT,s | 16.4 (15.6, 17.3) | 16.4 (15.7, 17.1) | 16.4 (15.7, 17.2) | 0.245 |
| Fibrinogen, g/L | 3.54 (2.97, 4.31) | 3.48 (3.01, 4.03) | 3.51 (2.98, 4.29) | 0.242 |
| PCT, ng/ml | 0.05 (0.05, 0.06) | 0.05 (0.05, 0.1) | 0.05 (0.05, 0.07) | 0.112 |
| Echocardiogram | ||||
| Left Atrial Size, mm | 43 (38, 48) | 45 (40, 51) | 43 (38.75, 48) | <0.001 |
| RVVD, mm | 58 (52, 63) | 61 (54.75, 65) | 59 (53, 64) | <0.001 |
| LVEF, % | 29 (24, 33) | 29 (23.5, 34) | 29 (24, 33.45) | 0.606 |
Figure 2
Univariate analysis showed the association between individual variables and the incidence of AKI.
Figure 3
Collinearity analysis of the independent variables. The results highlight the presence of significant multicollinearity, suggesting the need for careful consideration when incorporating these variables into subsequent machine learning analyses to mitigate potential biases arising from collinearity. *P < 0.05.
Figure 4
LASSO regression analysis for variable selection in predicting AKI in elderly patients with heart failure with reduced ejection fraction. Panel A shows the deviance (binomial) as a function of the log of the regularization parameter λ. The red dots represent the deviance at each λ value, with vertical dotted lines indicating the selected λ values based on cross-validation. Panel B shows the subset of variables with non-zero coefficients, identifying four potential risk factors associated with the outcome: NT-proBNP, Uric Acid (UA), C-reactive protein (CRP), and Urea. These variables were subsequently included in the multivariable analysis for further evaluation.
Figure 5
Nomogram representation for predicting the probability of the outcome. This visual tool incorporates the four selected variables, enabling individualized risk assessment to facilitate clinical decision-making. The nomogram offers a user-friendly interface, allowing healthcare professionals to gauge the likelihood of the outcome based on the specific values of the incorporated predictors.
Figure 6
Receiver operating characteristic (ROC) curves evaluating the discriminative ability of the developed model. The ROC curve was used to evaluate the predictive performance of the model. The dataset was randomly divided into training and testing cohorts. In the training set, the area under the curve (AUC) was 0.754, indicating good discrimination. To assess the model’s generalizability to unseen data, it was further evaluated in the testing set, yielding an AUC of 0.721. These results demonstrate that the model has satisfactory predictive ability and generalization performance in both datasets.
Figure 7
Calibration curve for evaluating the agreement between the model’s predicted probabilities and the observed outcomes. The calibration analysis resulted in a C-index of 0.754, signifying a robust concordance between predicted and actual outcomes. This indicates the model’s efficacy in accurately estimating the probabilities and ensuring reliable predictions for the outcome of interest.
Figure 8
The DCA plots of the Nomogram for the risk of AKI in HFrEF patients.