Abstract
Introduction
Kidney injury is an important manifestation of post-resuscitation syndrome and a significant factor leading to high mortality rates after cardiopulmonary resuscitation (CPR).This study aimed to develop and validate a multivariable nomogram to predict estimated glomerular filtration rate (eGFR) after CPR to assess the degree of kidney injury and provide protective strategies.
Methods
The clinical data of patients after CPR admitted to Tianjin Medical University General Hospital from January 2017 to June 2024 and Tianjin Medical University General Hospital Airport Hospital from January 2017 to December 2019 were retrospectively analyzed. The patients those who met the inclusion criteria were randomly divided into training and validation cohorts at a ratio of 7∶3.We obtained clinical data from January 2021 to June 2023 at First Affiliated Hospital of Hebei North University as external validation. Univariate and multivariate linear regression methods were used to identify independent risk factors for 7d-eGFR after CPR, develop and validate (internal and external) a multivariate nomogram model. Calibration curve, Bland-Altman plot, and paired-T validation were used to validate the predictive performance of the model.
Results
We included 439 patients after CPR, of whom 307 were in training cohort and 132 were in validation cohort. And 105 patients were included as an external validation cohort. Multivariable linear analysis showed that age (beta coefficient [β], 95% confidence interval: −0.344 [−0.528, −0.160]), hypertension (−3.610 [−5.968, −1.252]), diabetes mellitus (−2.992 [−5.295, −0.689]), no flow time (−0.577 [−0.996, −0.158]), baseline eGFR (0.349 [0.269∼0.429]), ACR (−0.042 [−0.073, −0.011]), lactic acid (−0.650 [−1.214,−0.086]) were the independent risk factors for eGFR after CPR. A composite nomogram predicted eGFR with good accuracy in training (97.07%), internal validation (95.45%), and external validation (91.08%) cohorts. The nomogram model has good predictive ability for AKI and CKD in training (AUC = 0.933 and 0.882), internal validation (AUC = 0.915 and 0.859), and external validation (AUC = 0.823 and 0.784) cohorts.
Conclusion
The developed nomogram could be used to predict 7d-eGFR after CPR, which helped to accurately quantify kidney function levels and early predict the probability of AKI and CKD progression, achieving early detection and intervention, thereby improving the prognosis of patients after CPR.
Keywords: Glomerular filtration rate, Acute kidney injury, Chronic kidney disease, Nomogram prediction model, Post-resuscitation syndrome
Introduction
Cardiac arrest (CA) remains a global clinical challenge with high post-resuscitation mortality, largely attributed to post-resuscitation syndrome, among which acute kidney injury (AKI) and its progression to chronic kidney disease (CKD) are key determinants of poor prognosis [1–3]. The transition from AKI to CKD is neither incidental nor inconsequential. Epidemiological data indicate that AKI survivors manifest a 2.5-fold increased risk of developing CKD within 5 years, with a significant subset progressing to end-stage renal disease requiring chronic dialysis or transplantation [4–6]. Despite this, current clinical practice is constrained by diagnostic mode that are reactive rather than anticipatory.
Therefore, early assessment of kidney function levels after resuscitation and prediction of high-risk patients with AKI and progression to CKD can provide an early intervention “time window” for treatment, which is of great significance for reducing mortality after resuscitation. At present, Kidney Disease Improving Global Outcomes (KDIGO) guidelines have a diagnostic time window of 48 h even 7 days for AKI and 3 months for CKD, thereby forfeiting critical early intervention opportunities during the initial 24–72 h post-CPR when kidney injury trajectories are most malleable [7, 8].
The glomerular filtration rate (GFR) is the gold standard for evaluating the overall kidney functional status and plays an important role in the prediction and diagnosis of kidney diseases [9, 10]. However, its accurate estimation in the post-CA is confounded by hemodynamic lability, systemic inflammatory responses, and the absence of validated predictive models that synthesize pre-arrest comorbidities with resuscitation specific variables [11, 12]. Consequently, there is a strong clinical demand for a predictive tool with the capable of stratifying patients at risk of significant kidney injury within the hyperacute phase post-CPR.
So, we hypothesized that a multivariable nomogram integrating readily available demographic, clinical, and laboratory parameters could accurately predict the 7-day estimated GFR (7d-eGFR) following CPR, thereby serving as a surrogate for early kidney injury severity. Through this model, we sought to shift the post-CPR kidney care timeline from reactive damage control to proactive preservation, ultimately predicted AKI and the probability of progression to CKD, achieving early detection and intervention, thereby delaying disease progression and improving the prognosis of CPR patients.
Materials and Methods
Data Source
We retrospectively collected clinical data of patients admitted to Tianjin Medical University General Hospital from January 2017 to June 2024 and Tianjin Medical University General Hospital Airport Hospital from January 2017 to December 2019. The two outcome variables of this study were AKI and CKD. AKI was determined based on elevated serum creatinine (Scr) levels according to the diagnostic criteria of KDIGO [13]. These criteria included: (1) an increase in Scr ≥26.5 μmol/L within 48h; (2) an increase in Scr exceeding 1.5 times the baseline value, known or inferred to have occurred within 7d [14, 15]. Patients with an estimated glomerular filtration rate (eGFR) below 60 mL/min/1.73 m2 for more than 3 months were defined as having CKD [16]. Furthermore, eGFR was calculated for all patients using the Chronic Kidney Disease Epidemiology Collaboration creatinine (CKD-EPI) equation [17].
Data Collection and Preprocessing
All patients in the study were required to satisfy each of the following inclusion criteria: (1) age ≥18 years; (2) diagnosis of CA, length of stay more than 24 h. Patients were excluded if they met any of the following criteria: (1) patients with end-stage kidney disease who required hemodialysis or had an eGFR below 60 mL/min/1.73 m2; (2) patients with incomplete data >15%; (3) trauma-induced CA. According to the same inclusion and exclusion criteria, we obtained clinical data from January 2021 to June 2023 at First Affiliated Hospital of Hebei North University as external validation.
We used the CARET package in R software for data preprocessing, removing variables that had strong correlations with other variables. Thirty four variables were selected, including (1) demographic information: age, sex and body mass index; (2) basic diseases and personal family history: cardiovascular history, diabetes, hypertension; (3) related indicators of CPR: no blood flow time, time to reach ROSC, adrenaline dosage; (4) first laboratory validation results within 24 h after resuscitation: white blood cells, hemoglobin, platelets, total cholesterol, triglycerides, albumin, globulin, alanine aminotransferase, aspartate aminotransferase, baseline eGFR, serum creatinine, urea nitrogen, uric acid, blood glucose, glycated hemoglobin, blood potassium, blood sodium, blood calcium, lactate, brain natriuretic peptide, troponin, D-dimer, proteinuria, 24-h urine output, and albumin to creatinine ratio (ACR). The dependent variable was the eGFR value on the 7th day (7d-eGFR). Two follow-up outcome variables (1) diagnosed AKI cases 7 days later; (2) diagnosed CKD cases 3 months later (online suppl. Table S1; for all online suppl. material, see https://doi.org/10.1159/000551922).
Model Development and Validation
Patients who met the inclusion criteria were randomly divided into training and validation cohorts at a ratio of 7∶3. Univariate linear analysis was conducted and variables with p < 0.05 were included in the multivariate linear regression model. Then the backward method was used to select variables with p < 0.05 as independent influencing variables, and the variance inflation factor (VIF) was used to evaluate the relationships between variables (VIF >5 was considered multicollinearity) [18]. A nomogram prediction model was constructed based on independent influencing variables. We evaluated the accuracy of the training and validation cohorts by calibration curve and mean absolute error (MAE). Bland-Altman plot and paired-T validation were used to assess correlation analysis between observed and predicted 7d-eGFR. And the predictive performance of the nomogram model for AKI and CKD was assessed by receiver operating characteristic (ROC) curves and area under the ROC curve (AUC). Although our primary modeling target was the7d-eGFR value, we secondarily assessed the clinical utility of the model by evaluating its ability to classify patients into AKI and CKD categories based on standard diagnostic thresholds applied to the predicted eGFR. All predictors were collected within 24 h of ROSC, ensuring temporal separation from the outcome.
Statistical Analysis
SPSS 26.0 software and R software (version 4.3.2) were used for data statistics and analysis. Continuous variables were presented as median with interquartile range and compared using the Mann-Whitney U validation, while categorical variables were represented by numerical values and corresponding percentages (n, %) and compared using the chi-square validation. p < 0.05 was considered statistically significant.
Results
Characteristics of the Study Cohort
According to the inclusion and exclusion criteria, a total of 439 patients were included. Among them, there were 212 AKI patients and 116 CKD patients. The total cohort was divided randomly into training cohort (n = 307) and validation cohort (n = 132) according to a 7:3 ratio (shown in Fig. 1). The comparison of characteristics between the training and validation cohorts showed no statistically significant differences in variable indicators (p > 0.05) (shown in Table 1).
Fig. 1.
Flowchart of training and validating the prediction model.
Table 1.
Comparison of clinical baseline characteristics between training cohort and validation cohort
| Variables | Total cohort (n = 439) | Training cohort (n = 307) | Validation cohort (n = 132) | p value |
|---|---|---|---|---|
| Age, years | 66 (62, 70) | 65 (62, 75) | 66 (63, 75) | 0.178 |
| Sex: male (%) | 231 (52.6) | 164 (53.4) | 67 (50.8) | 0.683 |
| Cardiovascular history (%) | 215 (49.0) | 154 (50.2) | 61 (42.6) | 0.512 |
| Diabetes mellitus (%) | 194 (44.2) | 135 (44.0) | 59 (44.7) | 0.972 |
| Hypertension (%) | 179 (40.8) | 129 (42.1) | 50 (37.9) | 0.482 |
| BMI, kg/m2 | 24.6 (23.3, 25.5) | 24.2 (23.1, 25.6) | 24.3 (22.9, 25.3) | 0.146 |
| No flow time, min | 5.0 (4, 6) | 5.0 (4, 6.5) | 5.0 (4.5, 6.0) | 0.480 |
| ROSC time, min | 20.5 (9.0, 28.5) | 20.5 (8.5, 29.5) | 21.0 (9.5, 28.0) | 0.452 |
| Adrenaline dosage, mg | 4 (3, 5) | 4 (3, 5) | 3 (3, 5) | 0.561 |
| WBC, cells ×109/L | 8.5 (7.5, 9.6) | 8.6 (7.5,10.3) | 8.3 (7.7, 9.8) | 0.447 |
| Hb, g/L | 105 (103, 120) | 106 (102, 117) | 104 (102, 121) | 0.153 |
| PLT, cells ×109/L | 223 (215, 242) | 225 (213, 246) | 220 (215, 253) | 0.197 |
| Total cholesterol, mmol/L | 4.4 (4.2, 4.7) | 4.4 (4.2, 4.9) | 4.3 (4.1, 4.7) | 0.913 |
| Triglycerides, mmol/L | 1.5 (1.5, 1.9) | 1.5 (1.5, 1.9) | 1.7 (1.4, 1.9) | 0.596 |
| Alb, g/L | 48 (45, 55) | 48 (45, 53) | 49 (46, 55) | 0.118 |
| Glb, g/L | 33 (32, 37) | 32 (31, 37) | 33 (32, 36) | 0.686 |
| ALT,U/L | 42 (34, 50) | 40 (33, 52) | 42 (35, 51) | 0.552 |
| AST, U/L | 33 (30, 36) | 32 (30, 35) | 34 (31, 36) | 0.727 |
| Baseline eGFR, mL/min/1.73 m2 | 90 (79, 99) | 91 (80, 100) | 88 (77, 98) | 0.209 |
| Scr, µmol/L | 94 (88, 105) | 93 (88, 104) | 95 (89, 107) | 0.993 |
| BUN, mmol/L | 9.2 (7.9, 11.3) | 9.2 (7.6, 11.9) | 8.9 (7.7, 11.5) | 0.588 |
| Uric acid, µmol/L | 314 (290, 347) | 315 (292, 350) | 313 (280, 349) | 0.584 |
| Glucose, mmol/L | 7.5 (6.3, 8.7) | 7.6 (6.4, 8.8) | 7.5 (6.2, 8.4) | 0.381 |
| HbA1c, % | 6.4 (5.6, 7.3) | 6.5 (5.8, 7.2) | 6.4 (5.5, 7.4) | 0.912 |
| Blood potassium, mmol/L | 4.8 (4.5, 5.4) | 4.9 (4.5, 5.5) | 4.8 (4.4, 5.3) | 0.204 |
| Blood sodium, mmol/L | 142 (138, 152) | 141 (136, 152) | 143 (139, 150) | 0.345 |
| Blood calcium, mmol/L | 2.32 (2.20, 2.37) | 2.33 (2.20, 2.39) | 2.32 (2.21, 2.35) | 0.406 |
| Lactic acid, mmol/L | 5.6 (4.6, 6.8) | 5.5 (4.6, 6.7) | 5.7 (4.5, 6.9) | 0.961 |
| BNP, ng/L | 142 (130, 221) | 144 (133, 217) | 141 (126, 222) | 0.468 |
| TnT, ng/mL | 0.27 (0.16, 0.41) | 0.28 (0.15, 0.42) | 0.27 (0.18, 0.40) | 0.581 |
| D-dimer, ng/mL | 808 (667, 1,253) | 810 (677, 1,261) | 807 (653, 1,192) | 0.503 |
| Proteinuria (%) | 251 (57.2) | 179 (58.3) | 72 (54.5) | 0.532 |
| 24h urine volume, mL | 1,590 (1270, 1840) | 1,600 (1,300, 1,860) | 1,580 (1,230, 1,810) | 0.790 |
| ACR, mg/g | 57 (37, 74) | 57 (37, 71) | 60 (38, 77) | 0.308 |
| 7d-eGFR, mL/min/1.73 m2 | 85 (76, 95) | 86 (77, 95) | 85 (75, 94) | 0.651 |
| AKI (%) | 212 (48.3) | 152 (48.9) | 60 (45.5) | 0.499 |
| CKD (%) | 116 (26.4) | 83 (27.0) | 33 (25.0) | 0.745 |
Data were presented as median (interquartile range) for continuous variables and n (%) for categorical variables.
BMI, body mass index; ROSC, return of spontaneous circulation; WBC, white blood cells; Hb, hemoglobin; PLT, platelets; Alb, albumin; Glb, globulin; ALT, alanine aminotransferase; AST, aspartate transaminase; eGFR, estimated glomerular filtration rate; BUN, blood urea nitrogen; BNP, brain natriuretic peptide; TnT, troponinT; ACR, albumin-to-creatinine ratio; AKI, acute kidney injury; CKD, chronic kidney disease.
Univariate linear regression analysis screened out 11 variables. Multivariate linear regression analysis showed that age (β = −0.344, 95% confidence interval [CI]: −0.528∼−0.160, p < 0.001), hypertension (β = −3.610, 95% CI: −5.968∼−1.252, p < 0.01), diabetes (β = −2.992, 95% CI: −5.295∼−0.689, p < 0.05), no flow time (β = −0.577, 95% CI: −0.996∼−0.158, p < 0.01), baseline eGFR (β = 0.349, 95%CI: 0.269∼0.429, p < 0.001), ACR (β = −0.042, 95%CI: −0.073∼−0.011,p < 0.05), lactic acid (β = −0.650, 95% CI: −1.214∼−0.086, p < 0.05) were independent influencing factors (Table 2).
Table 2.
Univariate and multivariate linear regression analysis
| | Univariable | Multivariable | ||
|---|---|---|---|---|
| beta coefficients (95% CI) | p value | beta coefficients (95% CI) | p value | |
| Age | −0.629 (−0.845∼−0.413) | <0.001 | −0.344 (−0.528∼−0.160) | <0.001 |
| Hypertension | −8.215 (−10.896∼−5.534) | <0.001 | −3.610 (−5.968∼−1.252) | <0.01 |
| Diabetes mellitus | −6.991 (−9.700∼−4.282) | <0.001 | −2.992 (−5.295∼−0.689) | <0.05 |
| No flow time | −1.181 (−1.687∼−0.675) | <0.001 | −0.577 (−0.996∼−0.158) | <0.01 |
| Adrenaline dosage | −2.872 (−3.772∼−1.972) | <0.001 | − | − |
| Baseline eGFR | 0.429 (0.341∼0.517) | <0.001 | 0.349 (0.269∼0.429) | <0.001 |
| Scr | −3.543 (−4.090∼−2.996) | <0.001 | − | − |
| ACR | −0.093 (−0.132∼−0.054) | <0.001 | −0.042 (−0.073∼−0.011) | <0.05 |
| Proteinuria | 3.019 (0.201∼5.837) | <0.05 | − | − |
| Lactic acid | −1.827 (−2.466∼−1.188) | <0.001 | −0.650 (−1.214∼−0.086) | <0.05 |
| Blood potassium | −14.059 (−16.595∼−11.523) | <0.001 | − | − |
eGFR, estimated glomerular filtration rate; ACR, albumin-to-creatinine ratio.
Linear regression equation is as follows: Y = 89.439 − 0.344 × age −3.610 × hypertension −2.992 × diabetes − 0.577 × no flow time + 0.0349 × baseline eGFR −0.042 × ACR − 0.650 × lactic acid. In addition, no multicollinearity was observed among the variables (VIF <5): age VIF = 1.105; hypertension VIF = 1.157; diabetes VIF = 1.116; no flow time VIF = 1.078; baseline eGFR VIF = 1.043; ACR VIF = 1.091; lactic acid VIF = 1.184.
Model Development and Validation
A nomogram prediction model was constructed based on multivariate linear regression (shown in Fig. 2). The nomogram was used by first giving each variable a score on the “Points” scale. The scores for all variables were then added to obtain the total score and a vertical line was drawn from the “Total Points” row to estimate the 7d-eGFR value.
Fig. 2.
Nomogram for predicting the 7d-estimated glomerular filtration rate (7d-eGFR).
Scatter plot of calibration curve was used to directly compare the observed versus predicted 7d-eGFR values for each individual patient. The nomogram predicted 7d-eGFR with good accuracy in training cohort (R = 0.928, 95% CI: 0.902–0.946; 97.07% accuracy; p < 0.001) (shown in Fig. 3a) and validation cohort (R = 0.893, 95% CI: 0.876–0.924; 95.45% accuracy; p < 0.001) (shown in Fig. 3b). The MAE between the predicted and actual 7d-eGFR values was 8.546 in training cohort and 8.465 in validation cohort.
Fig. 3.
Observed versus predicted 7d-estimated glomerular filtration rate (7d-eGFR) after CPR. a Observed versus predicted 7d-eGFR in training cohort (n = 307). b Observed versus predicted 7d-eGFR in validation cohort (n = 132).
In Bland-Altman plot, the upper and lower blue dashed lines represented the upper and lower limits of the 95% consistency limit, which was 1.96 times the standard deviation. The red solid line in the middle represented the mean of the difference. The red dashed line represented the position where the average difference was 0. The higher the consistency between the measurement results of the two methods, the closer the line representing the average difference (red solid line) was to the line representing the average difference of 0 (red dashed line). The mean value of the predicted and actual in training and validation cohorts were 4.5 (shown in Fig. 4a) and 3.6 (shown in Fig. 4b). Most of the differences between the two groups were within the 95% consistency limit, indicating good consistency between the predicted results of both groups. Paired-T validation results showed that there were no statistically significant differences between the actual and predictive values of the model: p = 0.89 in training cohort (shown in Fig. Fig. 5a) and p = 0.63 in validation cohort (shown in Fig. 5b), which suggested that the model had a good predictive ability.
Fig. 4.
Plot of differences between predicted and actual 7d-eGFR versus the mean of the predicted and actual. a Plot of differences in training cohort. b Plot of differences in validation cohort. The red solid line represented the mean of the differences. The two blue dotted lines represented the limits of agreement, ±1.96σ.
Fig. 5.
Paired-T results between the actual and predictive values of 7d-eGFR. a Paired-T results in training cohort. b Paired-T results in validation cohort.
Model Prediction of AKI and CKD
ROC curve analysis showed that the prediction ability for AKI was AUC = 0.933 (95% CI: 0.902–0.961) in training cohort and AUC = 0.915 (95% CI: 0.894–0.945) in validation cohort. The results of both cohorts indicated that the model had excellent predictive performance. Two cohorts of AUC were subjected to Delong testing, and the difference was not statistically significant (p = 0.773) (shown in Fig. 6a).
Fig. 6.
ROC curves for predicting outcome in training and validation cohorts. a ROC curves for predicting AKI. b ROC curves for predicting CKD.
ROC curve analysis showed that the prediction ability for CKD was AUC = 0.882 (95% CI: 0.857–0.925) in the training cohort and AUC = 0.859 (95% CI: 0.843–0.891) in the validation cohort. The results of both cohorts indicated that the model had good predictive performance. Two validation cohorts of AUC were subjected to Delong testing, and the difference was not statistically significant (p = 0.492) (shown in Fig. 6b).
Model External Validation
We obtained clinical data of 105 patients from First Affiliated Hospital of Hebei North University as external validation cohort. Scatter plot of calibration curve showed that the model predicted 7d-eGFR with good accuracy in external cohort (R = 0.499 [95% CI: 0.386–0.577]; 91.08% accuracy; p < 0.001) (shown in Fig. 7). The MAE between the predicted and actual 7d-eGFR values was 6.215. ROC curve analysis showed that the model in external validation had AUC = 0.823 (95% CI: 0.794∼0.846) and AUC = 0.784 (95% CI: 0.720∼0.822) for predicting AKI and CKD, indicating the model had good generalization ability (shown in Fig. 8; online suppl. Table S2).
Fig. 7.
Observed versus predicted 7d-eGFR after CPR in external validation cohort (n = 105).
Fig. 8.
ROC curves for predicting outcome in external validation cohort. a ROC curve for predicting AKI. b ROC curve for predicting CKD.
Discussion
The study developed a multiple linear regression-based nomogram model to predict 7d-eGFR within 24 h after CPR and further validated its efficacy in predicting AKI and CKD. Our analysis identified seven independent predictors of 7d-eGFR: age, hypertension, diabetes, no flow time, lactic acid, ACR, and baseline eGFR. The nomogram model demonstrated high accuracy in both internal validation, with strong correlation between predicted and observed values. External validation result showed that the model achieved a moderate correlation between predicted and observed 7d-eGFR, indicating limited but statistically significant predictive capacity beyond chance. The potential reasons for the reduced performance in the external validation might include: first differences in patient demographics and baseline kidney function; second temporal and geographic heterogeneity in clinical practice patterns affecting post-discharge follow-up and creatinine measurement timing; third potential overfitting due to unmeasured confounders or dataset-specific feature distributions.
The model showed excellent discriminative ability for AKI and CKD, confirming its robustness across populations. It should be noted that the continuous 7d-eGFR prediction model serves as the foundation, while the evaluations of AKI and CKD demonstrated its potential clinical applicability. These performances compared favorably with or exceed recent post-resuscitation prognostic tools that either focused on mortality or required complex physiological indicators unavailable in most centers [19, 20]. Previous studies on post-resuscitation kidney function have primarily focused on single biomarkers or delayed assessment, lacking the ability to integrate multi-dimensional clinical data for early prediction [21–23]. For instance, a cohort study by Geri et al. [24] reported that serum creatinine at 48 h post-CPR correlates with AKI risk, but this metric fails to capture the dynamic interplay of demographic, comorbid, and resuscitation-related factors.
Our findings were consistent with the post-resuscitation syndrome in which prolonged whole body ischemia-reperfusion injury synergistically amplified endothelial dysfunction, oxidative stress, and tubular damage [25–27]. Older age, hypertension, and diabetes well-recognized accelerators of glomerulosclerosis were additionally retained [28], reinforcing the concept that premorbid vulnerability determined the trajectory from early AKI to irreversible CKD. The negative coefficients for no flow time and serum lactic acid mirrored the duration and severity of systemic hypoperfusion, whereas baseline eGFR and ACR reflect pre-existing kidney reserve and microvascular integrity. Importantly, traditional resuscitation metrics such as adrenaline dose and time to ROSC were excluded during multivariable refinement, suggesting that kidney prognosis was dictated more by the depth of initial insult and baseline susceptibility than by in-hospital resuscitation intensity.
The nomogram prediction model had the following advantages. First, the model enabled early quantification of kidney function within 24 h after CPR, addressing the limitation of current KDIGO guidelines. This “window” for early intervention was critical as timely adjustments of dialysis timing and drug dosages based on predicted GFR could mitigate kidney injury progression. The simplicity nomogram was another advantage [29]. Unlike complex machine learning models that required specialized software, our nomogram allowed clinicians to calculate predicted 7d-eGFR via visual scoring, enhancing its applicability in resource-limited settings. A key strength of this study is its strict methodological design, including multicenter data collection, strict inclusion/exclusion criteria, and stepwise validation (internal and external). The use of VIF <5 ruled out multicollinearity, ensuring the stability of independent predictors [30].
However, several limitations should be acknowledged. First, the retrospective design might introduce selection bias as patients with incomplete data (>15%) were excluded. Second, the external validation cohort was limited to one center, and further validation across different ethnicities or regions was needed. Third, unmeasured variables, such as genetic polymorphisms or detailed post-resuscitation care protocols, were not included, which might affect model performance. Fourth, the study lacked data on the etiologies of CA, a critical variable that could have influenced subsequent renal outcomes. Finally, the long-term predictive value for end-stage renal disease of the model remained unvalidated, warranting extended follow-up. Future research should focus on three areas: (1) prospective multicenter studies to validate the model in broader populations, including pediatric or elderly subgroups; (2) integration of emerging biomarkers such as neutrophil gelatinase-associated lipocalin or imaging data to improve predictive accuracy; (3) exploration of utility of the model in guiding personalized interventions, such as targeted kidney perfusion optimization or early nephroprotective drug administration.
In conclusion, we had developed and externally validated the clinical model that early predicted 7d-eGFR, AKI, and subsequent CKD after CPR using seven universally accessible variables. Pending prospective interventional confirmation, this tool could facilitate early individualized risk stratification, guide timely nephroprotective measures and ultimately improved long-term kidney and patient-centered outcomes among survivors of CA.
Statement of Ethics
This study was approved by the Ethics Committee of Tianjin Medical University General Hospital (Approval No. IRB2021-KY-300) and the Ethics Committee of Tianjin Medical University General Hospital Airport Hospital (Approval No. IRB2025-005-01). Given the retrospective nature of the study, the requirement for informed consent was exempted by Prof. Zhixiang Zhang of the Ethics Committee of Tianjin Medical University General Hospital and the Ethics Committee of Tianjin Medical University General Hospital Airport Hospital.
Conflict of Interest Statement
The authors declare that there are no conflicts of interest.
Funding Sources
This work was supported by supported by Key Laboratory of Medical Rescue Key Technology and Equipment, Ministry of Emergency Management (Grant No. YJBKFKT202510), National Ministry of Industry and Information Technology-2024 Special Project for Engineering Technology and Equipment for Natural Disaster Prevention and Control (Grant No. 0747-2461SCCZA303-10), and Sanming Project of Medicine in Shenzhen (Grant No. SZSM202411032). The funders had no role in the study design, execution and analysis, manuscript conception, planning, writing, and decision to publish.
Author Contributions
Jinxiang Wang, Heng Jin and Yanfen Chai were responsible for the conception and design of the study. Jinxiang Wang drafted the manuscript. Guowu Xu, Jinxuan Liu, Qin Li, Zhongliang Ji, and Qianlong Xue were responsible for data acquisition and analysis. Wei Han reviewed the analyses and provide financial support. Qi Lv, Shike Hou, and Haojun Fan contributed to interpreting the findings and critically reviewed the manuscript. All authors read and approved the final manuscript.
Funding Statement
This work was supported by supported by Key Laboratory of Medical Rescue Key Technology and Equipment, Ministry of Emergency Management (Grant No. YJBKFKT202510), National Ministry of Industry and Information Technology-2024 Special Project for Engineering Technology and Equipment for Natural Disaster Prevention and Control (Grant No. 0747-2461SCCZA303-10), and Sanming Project of Medicine in Shenzhen (Grant No. SZSM202411032). The funders had no role in the study design, execution and analysis, manuscript conception, planning, writing, and decision to publish.
Data Availability Statement
All data generated or analyzed during this study are included in this article and its online supplementary material. Further inquiries can be directed to the corresponding author.
Supplementary Material.
Supplementary Material.
References
- 1. Rundgren M, Ullén S, Morgan MPG, Glover G, Cranshaw J, Al-Subaie N, et al. Renal function after out-of-hospital cardiac arrest; the influence of temperature management and coronary angiography, a post hoc study of the target temperature management trial. Crit Care. 2019;23(1):163. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Zheng J, Lv C, Zheng W, Zhang G, Tan H, Ma Y, et al. Incidence, process of care, and outcomes of out-of-hospital cardiac arrest in China: a prospective study of the BASIC-OHCA registry. Lancet Public Health. 2023;8(12):e923–32. [DOI] [PubMed] [Google Scholar]
- 3. Rasmussen SB, Jeppesen KK, Kjaergaard J, Hassager C, Schmidt H, Mølstrøm S, et al. Blood pressure and oxygen targets on kidney injury after cardiac arrest. Circulation. 2023;148(23):1860–9. [DOI] [PubMed] [Google Scholar]
- 4. Kovesdy CP, Coresh J, Ballew SH, Woodward M, Levin A, Naimark DM, et al. CKD Prognosis Consortium. Past decline versus current eGFR and subsequent ESRD risk. J Am Soc Nephrol. 2016;27(8):2447–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Levey AS. Defining AKD: the spectrum of AKI, AKD, and CKD. Nephron. 2022;146(3):302–5. [DOI] [PubMed] [Google Scholar]
- 6. Niculae A, Gherghina ME, Peride I, Tiglis M, Nechita AM, Checherita IA. Pathway from acute kidney injury to chronic kidney disease: molecules involved in renal fibrosis. Int J Mol Sci. 2023;24(18):14019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Cheung AK, Chang TI, Cushman WC, Furth SL, Hou FF, Ix JH, et al. Executive summary of the KDIGO 2021 clinical practice guideline for the management of blood pressure in chronic kidney disease. Kidney Int. 2021;99(3):559–69. [DOI] [PubMed] [Google Scholar]
- 8. Levey AS, Eckardt KU, Dorman NM, Christiansen SL, Hoorn EJ, Ingelfinger JR, et al. Nomenclature for kidney function and disease: report of a kidney disease: improving Global Outcomes (KDIGO) Consensus Conference. Kidney Int. 2020;97(6):1117–29. [DOI] [PubMed] [Google Scholar]
- 9. KDIGO Conference Participants . Genetics in chronic kidney disease: conclusions from a kidney disease: improving Global Outcomes (KDIGO) Controversies Conference. Kidney Int. 2022;101(6):1126–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Shlipak MG, Tummalapalli SL, Boulware LE, Grams ME, Ix JH, Jha V, et al. The case for early identification and intervention of chronic kidney disease: conclusions from a kidney disease – improving Global Outcomes (KDIGO) Controversies Conference. Kidney Int. 2021;99(1):34–47. [DOI] [PubMed] [Google Scholar]
- 11. Kuroki N, Abe D, Iwama T, Sugiyama K, Akashi A, Hamabe Y, et al. Prognostic effect of estimated glomerular filtration rate in patients with cardiogenic shock or cardiac arrest undergoing percutaneous veno-arterial extracorporeal membrane oxygenation. J Cardiol. 2016;68(5):439–46. [DOI] [PubMed] [Google Scholar]
- 12. Shroff GR, Garcia S, Schmidt C, Okeson B, Tannenbaum E, Pacheco R, et al. Renal impairment and mortality in patients with STEMI and cardiogenic shock/cardiac arrest. Catheter Cardiovasc Interv. 2023;102(2):179–90. [DOI] [PubMed] [Google Scholar]
- 13. Kellum J, Lameire N, KDIGO AKI Guideline Work Group . Diagnosis, evaluation, and management of acute kidney injury: a KDIGO summary (Part 1). Crit Care. 2013;17(1):204. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Thomas ME, Blaine C, Dawnay A, Devonald MA, Ftouh S, Laing C, et al. The definition of acute kidney injury and its use in practice. Kidney Int. 2015;87(1):62–73. [DOI] [PubMed] [Google Scholar]
- 15. Yang L, Xing G, Wang L, Wu Y, Li S, Xu G, et al. Acute kidney injury in China: a cross-sectional survey. Lancet. 2015;386(10002):1465–71. [DOI] [PubMed] [Google Scholar]
- 16. Nelson RG, Grams ME, Ballew SH, Sang Y, Azizi F, Chadban SJ, et al. CKD prognosis consortium. Development of risk prediction equations for incident chronic kidney disease. JAMA. 2019;322(21):2104–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. GBD 2017 Disease and Injury Incidence and Prevalence Collaborators . Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet. 2018;392(10159):1789–858. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Li F, Yang M, Li Y, Zhang M, Wang W, Yuan D, et al. An improved clear cell renal cell carcinoma stage prediction model based on gene sets. BMC Bioinformatics. 2020;21(1):232. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Grams ME, Sang Y, Ballew SH, Gansevoort RT, Kimm H, Kovesdy CP, et al. CKD Prognosis Consortium. A meta-analysis of the association of estimated GFR, albuminuria, age, race, and sex with acute kidney injury. Am J Kidney Dis. 2015;66(4):591–601. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Fraccaro P, van der Veer S, Brown B, Prosperi M, O’Donoghue D, Collins GS, et al. An external validation of models to predict the onset of chronic kidney disease using population-based electronic health records from Salford, UK. BMC Med. 2016;14:104. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Tangri N, Grams ME, Levey AS, Coresh J, Appel LJ, Astor BC, et al. CKD Prognosis Consortium. Multinational assessment of accuracy of equations for predicting risk of kidney failure: a meta-analysis. JAMA. 2016;315(2):164–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Jou C, Shah R, Figueroa A, Patel JK. The role of inflammatory cytokines in cardiac arrest. J Intensive Care Med. 2020;35(3):219–24. [DOI] [PubMed] [Google Scholar]
- 23. Cunningham CA, Coppler PJ, Skolnik AB. The immunology of the post-cardiac arrest syndrome. Resuscitation. 2022;179:116–23. [DOI] [PubMed] [Google Scholar]
- 24. Geri G, Guillemet L, Dumas F, Charpentier J, Antona M, Lemiale V, et al. Acute kidney injury after out-of-hospital cardiac arrest: risk factors and prognosis in a large cohort. Intensive Care Med. 2015;41(7):1273–80. [DOI] [PubMed] [Google Scholar]
- 25. Oh CJ, Kim MJ, Lee JM, Kim DH, Kim IY, Park S, et al. Inhibition of pyruvate dehydrogenase kinase 4 ameliorates kidney ischemia-reperfusion injury by reducing succinate accumulation during ischemia and preserving mitochondrial function during reperfusion. Kidney Int. 2023;104(4):724–39. [DOI] [PubMed] [Google Scholar]
- 26. Mah KE, Alten JA, Cornell TT, Selewski DT, Askenazi D, Fitzgerald JC, et al. Acute kidney injury after in-hospital cardiac arrest. Resuscitation. 2021;160:49–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Yang WS, Kim YJ, Ryoo SM, Kim WY. Independent risk factors for sepsis-associated cardiac arrest in patients with septic shock. Int J Environ Res Public Health. 2021;18(9):4971. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Kanbay M, Copur S, Yildiz AB, Covic A, Covic A, Ciceri P, et al. Intrauterine life to adulthood: a potential risk factor for chronic kidney disease. Nephrol Dial Transplant. 2023;38(12):2675–84. [DOI] [PubMed] [Google Scholar]
- 29. Bandini M, Fossati N, Briganti A. Nomograms in urologic oncology, advantages and disadvantages. Curr Opin Urol. 2019;29(1):42–51. [DOI] [PubMed] [Google Scholar]
- 30. Zhu ZY, Ren KJ, Duan XW, Hu XL, Lv Y, Wang D, et al. Fibrosis-4 index as a predictor of all-cause and cardiovascular mortality in patients with chronic kidney disease. PLoS One. 2025;20(8):e0329315. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
All data generated or analyzed during this study are included in this article and its online supplementary material. Further inquiries can be directed to the corresponding author.