https://orcid.org/0000-0001-9941-9327
Abstract
Background
Acute pancreatitis (AP) is associated with a high incidence of acute kidney injury (AKI). This study aimed to develop a nomogram for predicting the early onset of AKI in AP patients admitted to the intensive care unit.
Method
Clinical data for 799 patients diagnosed with AP were extracted from the Medical Information Mart for Intensive Care IV database. Eligible AP patients were randomly divided into training and validation cohorts. The independent prognostic factors for the early development of AKI in AP patients were determined using the all-subsets regression method and multivariate logistic regression. A nomogram was constructed for predicting the early occurrence of AKI in AP patients. The performance of the nomogram was evaluated based on the area under the receiver operating characteristic curve (AUC), calibration curves and decision curve analysis (DCA).
Results
Seven independent prognostic factors were identified as predictive factors for early onset AKI in AP patients. The AUC of the nomogram in the training and validation cohorts were 0.795 (95% CI, 0.758–0.832) and 0.772 (95% CI, 0.711–0.832), respectively. The AUC of the nomogram was higher compared with that of the BISAP, Ranson, APACHE II scores. Further, the calibration curve revealed that the predicted outcome was in agreement with the actual observations. Finally, the DCA curves showed that the nomogram had a good clinical applicability value.
Conclusion
The constructed nomogram showed a good predictive ability for the early occurrence of AKI in AP patients.
Introduction
Acute pancreatitis (AP) is a common disease of sudden onset. The disease has a diverse clinical course that ranges from mild, to moderate to severe acute pancreatitis (SAP), with a high mortality rate of 36–50% [Citation1]. The pathophysiological mechanisms of AP are not well understood. However, pancreatic enzymes and various inflammatory mediators are thought to contribute to systemic inflammatory response syndrome and multi-organ failure [Citation2]. In some patients, AP may lead to acute kidney injury (AKI). AKI in AP has an incidence rate of 10–42% and a high mortality rate of about 80% [Citation3,Citation4]. Therefore, an early and accurate diagnosis of AKI in AP may help improve the disease prognosis. A few studies have explored factors and establish predictive models for AKI in AP patients. However, these studies had small sample sizes and lacked the accuracy of the predictive models [Citation5–8]. Early and accurate diagnosis of AKI in AP patients has remained challenging in clinical practice. Therefore, this study aimed to identify the prognostic factors for AKI in AP patients from a large database. In addition, we established and validated an easy-to-use prognostic nomogram, to help clinicians to prevent and treat AKI.
Methods
Data source
AP patients were obtained from the Medical Information Mart for Intensive Care-IV 1.0 (MIMIC-IV v1.0) database [Citation9]. The MIMIC-IV is a large, single-center, open-access database, containing data from 382,278 patients with 524,740 admissions to Beth Israel Deaconess Medical Center in Boston from 2008 to 2019. Relevant data, including patient demographics, hourly vital signs, laboratory microbial culture and imaging test results, surgical procedures, medication records, and survival data, were collected.
The use of data from the MIMIC-IV database was approved by the Institutional Review Board of the Beth Israel Deaconess Medical Center and Massachusetts Institute of Technology. Since the data were deidentified, there was no requirement for informed consent. To access the database, we first completed the required online courses and an examination (Record ID: 42039823).
Patients and data variables
Data were extracted using Structured Query Language (SQL) programming in PostgreSQL (version 14.0), and the SQL script codes used to extract patients' information were obtained from the GitHub website (https://github.com/MIT-LCP/mimic-code/tree/main/mimic-iv) [Citation10].
Using the International Classification of Diseases (ICD), ninth revision (ICD-9, code 577.0) and tenth revision (ICD-10, code K85%), we identified the patients diagnosed with AP from the MIMIC-IV v1.0 database. The exclusion criteria were age less than 18 years old and intensive care unit (ICU) stay < 24 h and patients with a history of chronic renal failure, including a previous history of a kidney transplant. For patients who had multiple admissions to the ICU, data were only obtained from the first admission.
After identifying eligible patients, we collected AP patient’s baseline parameters immediately after admission to the ICU, including demographic information, previous medical history, vital signs, laboratory indicators, injury factors, interventions, and disease severity scores. The vital signs and laboratory indicators are the first value within the first 24 h after ICU, and intervention measures and the severity scores of the disease are evaluated within the first 24 h after ICU.
The outcome measure was determined as the development of AKI within seven days of admission to the ICU. A diagnosis of AKI was based on the 2012 Kidney Disease: Improving Global Outcomes (KIDGO) guidelines as follows: elevated serum creatinine (SCr) after 48 h by ≥26.5 μmol/L (0.3 mg/dl); or ≥50% of the baseline SCr value (increased by 1.5 times), or a urine output of more than six hours of <0.5 mL/kg/h [Citation11]. Baseline SCr is defined as the lowest SCr in the past 7 days.
Development and validation of the prediction model
Missing data were common in the MIMIC database. In this study, variables with a deletion rate greater than 20% were excluded. The extracted variables and missing proportions of variables are shown in Appendix 1. The trimming method is used to deal with abnormal values, and the technique of multiple imputations is used to fill in missing data [Citation12].
The patients were randomly assigned to the training cohort (70%) and validation cohort (30%). Subsequently, a nomogram was established based on the training cohort and validated in the validation cohort.
Statistical analysis
Measurement data conforming to the normal distribution and homogeneity of variance were expressed as the mean ± standard deviation. Data between groups were compared using the independent t-test. The unequal variance t-test was used to compare data between groups with unequal variances. Data with skewed distribution were expressed as median and quartiles and were compared using the Mann–Whitney U test. Furthermore, categorical variables were presented with count (%) and were compared using the χ2 test.
Due to a large number of variables, the independent prognostic factors were filtered through two steps. First, a preliminary screening was conducted to identify possible predictors using the all-subsets regression method to avoid overfitting and increase the degree of the model. Subsequently, multivariate logistic regression was carried out. A nomogram was then constructed using all the independent prognostic factors.
The nomogram was validated using multiple metrics. The discriminative ability of the nomogram was evaluated using the area under the receiver operating characteristic curve (AUC). The obtained AUC values were compared with AUC values of the bedside index for the severity in acute pancreatitis (BISAP) scores, Ranson scores, and acute physiology and chronic health evaluation (APACHE) II scores. Further, a calibration curve was plotted to evaluate the calibrating ability of the nomogram. Moreover, Decision curve analysis (DCA) was used to assess the clinical significance of the nomogram. A p-value <0.05 was considered statistically significant. Data were analyzed using the Stata software (version 15.1) and R software (version 4.1.0), including tidyverse, mice, caret, leaps, glmnet, rms, pROC, and ggDCA.
Results
Baseline characteristics of the study cohort
A total of 799 patients with acute pancreatitis were identified from the MIMIC-IV database. The patients were randomly assigned to the training cohort (n = 560) and the validation cohort (n = 239). Patients were screened for acute pancreatitis as shown in . Further, the AP patients were divided into two groups (AKI and non-AKI groups) based on whether the patients developed AKI within seven days of admission into the ICU. The overall incidence rate of AKI was 62.45% (499/799). Further analysis revealed that 12.52% (100/799), 26.53% (212/799), and 23.40% (187/799) had stage 1, stage 2, and stage 3 AKI, respectively (Appendix 2).
Figure 1. A flow chart showing the screening criteria for acute pancreatitis.
Clinical data of patients in the AKI group were compared to that of patients in the non-AKI group. The results revealed that patients in the AKI group were older, with higher mortality rates and APACHE II, BISAP, and Ranson scores compared with the non-AKI group. In addition, patients in the AKI group were more likely to require mechanical ventilation and renal replacement therapy, use vasoactive drugs, and had longer ICU and hospital stays than those in the non-AKI group. Differences between the AKI and non-AKI groups were statistically significant (all p-values <0.05). See for details.
Table 1. A comparison of the baseline characteristics between the AKI and non-AKI groups.
Early potential predictors for AKI in AP patients
Variables in the training cohort with a p-value <0.2 between the AKI and non-AKI groups were selected as potential predictors for AKI. The variables were screened using the all-subsets regression model. Consequently, seven variables were identified, including age, race, total bilirubin, activated partial thromboplastin time (APTT), need for mechanical ventilation, use of vasoactive drugs, and sepsis as the potential predictors of AKI in AP, after adjusting for the maximum value of R2 (adjR2). See for details.
Figure 2. Potential predictors for AKI in AP patients.
Construction of the nomogram for predicting the risk of developing AKI
Multivariate logistic regression was performed on the seven variables identified in the all-subsets regression. The results showed that age, ethnicity, total bilirubin, activated partial thromboplastin time, need for mechanical ventilation, use of vasoactive drugs, and sepsis were all independent risk factors for early development of AKI in AP patients as shown in .
Table 2. Results of the multivariate logistic regression analysis showing predictors for the early occurrence of AKI in AP Patients.
Further, a nomogram was constructed based on the results of the multivariate logistic regression. The risk scores for each factor included in the nomogram are shown in . The higher the score, the higher the risk for AP patients to develop AKI.
Figure 3. A nomogram based on age, ethnicity, total bilirubin, activated partial thromboplastin time, need for mechanical ventilation, use of vasoactive drugs, and sepsis.
Evaluation of the effectiveness of the nomogram
In the training cohort, the AUC value was 0.795 (95% CI, 0.758–0.832). Further, the BISAP, Ranson, and APACHE II scores showed AUC values of 0.638(95% CI, 0.593–0.68), 0.706(95% CI, 0.664–0.748), and 0.734(95% CI, 0.693–0.776), respectively. The differences were statistically significant (D = 5.217, 3.094, 2.109, p < 0.05). The validation cohort had AUC values of 0.772 (95% CI, 0.711–0.832). In addition, the BISAP, Ranson, and APACHE II showed AUC values of 0.645(95% CI, 0.576–0.714), 0.680(95% CI, 0.613–0.746), and 0.717(95% CI, 0.652–0.781), respectively. The difference was not statistically significant (D = 2.715, d = 2.017, d = 1.227, p = 0.007, p = 0.044, p = 0.221). shows calibration curves for the nomogram, the training cohort, and the validation cohort. The curve is close to the front diagonal, indicating that the nomogram had good prediction performance.
Figure 4. (A,B) ROC curves of the nomogram, BISAP score, Ranson score, and APACHE II score for predicting the likelihood of developing AKI in AP patients. (A) is the training cohort, (B) is the validation cohort. (C,D) Calibrate curves of the nomogram for predicting AKI in AP patients; (C) is the training cohort, (D) is the validation cohort.
Clinical use of the nomogram
DCA was plotted with net benefit as the ordinate coordinate and high-risk threshold probability as the abscissa. The results showed that the nomogram had a good clinical value. The DCA revealed that patients had a higher net benefit than BISAP, Ranson, and APACHE II scores, as shown in .
Figure 5. DCA of the nomogram, BISAP score, Ranson score, and APACHE II score predict the early occurrence of AKI in AP patients; A is the training cohort, and B is the validation cohort. Note: The pink dotted line (None) represents the net benefit rate in the non-AKI group, while the purple dotted line (All) represents the net benefit rate of the AKI group.
Discussion
AKI is a common complication of AP patients. AKI was independently associated with a higher mortality rate in AP patients [Citation13]. Therefore, early prediction of the risk of developing AKI in acute pancreatitis may help lower the mortality rate of the disease. The MIMIC database contains a large amount of data on the clinical diagnosis and treatment of critically ill patients, thus providing data for scientific research [Citation14]. In our study, the risk factors of early AKI in AP patients were comprehensively screened using the all-subsets regression method and multivariate logistic regression. The results indicated that age, ethnicity, Total bilirubin (TBIL), activating partial thrombin time (APTT), mechanical ventilation, vasopressor and sepsis were independent risk factors for early AKI in AP patients, which was consistent with previous studies. The constructed nomogram had a good predicting ability, was based on a large sample size obtained from the MIMIC database, and had a good clinical utility.
An increase in age is associated with a decline in renal function. Patients with acute pancreatitis have a higher incidence rate of developing AKI associated with deterioration of physiological functioning with age [Citation15]. In addition, age is a known predictor of AP severity and mortality [Citation16]. For example, older patients have a higher risk of developing systemic complications, such as multiple organ failure [Citation15]. The risk of developing kidney diseases varies amongst different ethnicities. For example, African Americans have a 2–4 times higher risk of developing chronic kidney disease or end-stage renal disease than Caucasians. In addition, African Americans have a higher risk of developing AKI than Caucasians. These differences may be attributed to differences in economic conditions, social status, and genetics between African Americans and Caucasians. However, the present study revealed that Caucasians with acute pancreatitis were more likely to develop AKI than African Americans, possibly due to their diet and a higher basal metabolic rate [Citation17].
Furthermore, the present study identified hyperbilirubinemia to be an independent risk factor for the early development of AKI in AP patients. Hyperbilirubinemia may induce the formation of reactive oxygen species in mitochondria, damaging the tubular epithelial cells and exacerbating ischemic reperfusion kidney injury [Citation18,Citation19]. Bilirubin is a potent antioxidant and has been shown to be. An elevated bilirubin level is an independent risk factor for the development of AKI given that the pathogenesis of AP is closely related to oxidative stress [Citation20–22].
Coagulation dysfunction is an independent risk factor for AKI [Citation23]. In addition, a prolonged APTT has a predictive value in sepsis-associated AKI [Citation24,Citation25]. The present study revealed that an APTT value greater than 20 was associated with a poorer prognosis. According to Liu et al. differences in APTT values and thrombin time (TT) values were risk factors for organ failure in AP patients [Citation26], consistent with the present study.
The analysis revealed that mechanical ventilation (MV) predicted AKI in AP, which was consistent with findings from a previous study by Shi et al. [Citation27]. Studies have shown that acute respiratory failure due to acute pancreatitis necessitates the use of MV in patients admitted to the ICU. The use of MV can lead to acute lung injury, thus worsening hypoxia, causing vasoconstriction, decreased renal perfusion, and a decreased glomerular filtration rate. In addition, MV increases intrathoracic pressure, lower venous return and the mean arterial pressure, which may cause prerenal hypoperfusion and lead to acute renal injury [Citation28,Citation29]. This study demonstrated that vasoactive drugs can predict AKI in AP. Consistent with this finding, a previous study revealed that the need for MV, the use of vasopressor agents, and RRT were risk factors for higher mortality in AP patients [Citation30]. Critically ill patients require higher doses of vasopressor agents to help regulate blood pressure.
Up to one-third of patients with necrotizing pancreatitis develop necrotic infection [Citation31]. The incidence of SAP is biphasic and is closely related to the early and persistent presence of organ or multi-organ dysfunction in the first week of the disease course. Clinical sepsis caused by multiple organ failure syndrome due to infectious necrosis occurs later in the first week [Citation32]. A prospective study revealed that the most important independent predictors of AP mortality were persistent organ failure and infectious pancreatic necrosis complicated by multidrug-resistant organisms [Citation33]. The extent of necrosis has been correlated with organ failure and mortality, this may be associated with further complications of the pancreas and extrapancreatic necrosis, which could predispose the patient to infections, pseudoaneurysms, and intestinal fistulas. Sepsis is also an independent risk factor for acute kidney injury [Citation34]. In addition, unsaturated fatty acids could worsen systemic inflammation and organ failure. The substances released by the necrotizing pancreas are involved in the pathogenesis of AKI [Citation35]. The main and secondary mediators of systemic inflammatory response play a key role in the pathogenesis of AP and significantly participate in the development of AP-AKI and other organ dysfunction [Citation2]. TNF-α, released during the happens of AP, interacts with ductal pancreatic cells, glomeruli and renal tubules, accompanied by obvious local inflammatory reaction cells, continuous tissue ischemia, interstitial edema and cell necrosis.
Currently, the APACHE II, BISAP, and Ranson scores are clinically used to predict the prognosis of AP patients in the ICU [Citation36–38]. there is no relevant research to calculate the predictive value of the above score for early AKI in patients with acute pancreatitis in the ICU. The nomogram showed a good degree of differentiation and calibration in the training and validation cohorts. The AUC of the APACHE II, BISAP, and Ranson scores in the training and validation cohorts were lower than that of the nomogram. In addition to validating the APACHE II score in the validation cohort (This may be related to the sample size of the validation set, and further sample size verification is required in the future), the AUC between the nomogram and the scores showed statistically significant differences. Compared with the above scoring system, the variables of the nomogram can be simply obtained and are easy to calculate. Further, the nomogram can be used to stratify the risk of AKI in AP patients admitted to the ICU, thus guiding physicians to offer targeted management.
However, this study had some limitations. First, this study was retrospective, and single-centered. Therefore, future studies should be conducted prospectively in multiple centres to improve clinical utility and external validation, respectively. In addition, the nomogram did not include novel biomarkers or imaging results, which could hinder the performance of the model. The model did not include other factors, such as causes of acute pancreatitis, classification of acute pancreatitis, intra-abdominal hypertension and abdominal compartment syndrome that could potentially affect the development of AKI in AP. Third, the sample size of this study is not large, and only internal verification is used to evaluate the accuracy and effectiveness of the model. Future studies should employ large sample sizes, and include more variables to validate our findings.
Ethics approval and consent to participate
All procedures performed in the present study were in accordance with the principles outlined in the 1964 Helsinki Declaration and its later amendments. The collection of patient information and creation of the research resource was reviewed by the Institutional Review Board at the Beth Israel Deaconess Medical Center, which granted a waiver of informed consent and approved the data-sharing initiative. Because the MIMIC III database has received ethical approval from the Massachusetts Institute of Technology (Cambridge, MA) and Beth Israel Deaconess Medical Center (Boston, MA) and its data is publicly available and all patient data are deidentied.
Author contributions
XJD contributed to the collection of the data and creation of models, performed the study, contributed to its design, the analysis of the data and revised the manuscript. SMW drafted the article, revised the manuscript for important intellectual content, oversight the research activity planning and execution, and varied research outputs. YC contributed to the formal analysis of the data and oversight the research activity planning and execution; QZ conceived the study and revised the manuscript. All authors contributed substantially to the manuscript. All authors approved the final version submitted for publication.
Acknowledgements
The authors are absolutely grateful for the publicly available MIMIC database. The authors also wish to sincerely thank my friend Fei Zhao, for his guidance on peer review comments. The authors thank the Home for Researchers editorial team (www.home-for-researchers.com) for the language editing service. The authors would like to express our special gratitude to the preprint published by Research Square, which was published on researchsquare.com website on 2022/08/14 and can be accessed through the internet link (https://www.researchsquare.com/article/rs-1961177/v1) or DOI (10.21203/rs.3.rs-1961177/v1).
Disclosure statement
No potential conflict of interest was reported by the author(s).
Data availability statement
The data that support the findings of this study are openly available on the MIMIC-IV website at https://physionet.org/content/mimiciv/1.0/.
Additional information
Funding
References
- Banks PA, Bollen TL, Dervenis C, et al. Classification of acute pancreatitis–2012: revision of the atlanta classification and definitions by international consensus. Gut. 2013;62(1):102–111.
- Scurt FG, Bose K, Canbay A, et al. Acute kidney injury following acute pancreatitis (AP-AKI): definition, pathophysiology, diagnosis and therapy. Z Gastroenterol. 2020;58(12):1241–1266.
- Hamada S, Masamune A, Shimosegawa T. Management of acute pancreatitis in Japan: analysis of nationwide epidemiological survey. WJG. 2016;22(28):6335–6344.
- Lin HY, Lai JI, Lai YC, et al. Acute renal failure in severe pancreatitis: a population-based study. Ups J Med Sci. 2011;116(2):155–159.
- Huang HL, Nie X, Cai B, et al. Procalcitonin levels predict acute kidney injury and prognosis in acute pancreatitis: a prospective study. PLoS One. 2013;8(12):e82250.
- Yang D, Zhao L, Kang J, et al. Development and validation of a predictive model for acute kidney injury in patients with moderately severe and severe acute pancreatitis. Clin Exp Nephrol. 2022;26(8):770–787.
- Qu C, Gao L, Yu XQ, et al. Machine learning models of acute kidney injury prediction in acute pancreatitis patients. Gastroenterol Res Pract. 2020;2020:1–8.
- Lu L, Feng Y, Liu YH, et al. The systemic immune-inflammation index may be a novel and strong marker for the accurate early prediction of acute kidney injury in severe acute pancreatitis patients. J Invest Surg. 2021;35(5):962–966.
- Johnson AEW, Bulgarelli L, Shen L, et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci Data. 2023;10(1):1.
- Johnson AE, Stone DJ, Celi LA, et al. The MIMIC code repository: enabling reproducibility in critical care research. J Am Med Inf Assoc. 2018;25(1):32–39.
- Kidney disease: improving global outcomes (KDIGO) acute kidney injury work group. KDIGO clinical practice guideline for acute kidney injury. Kidney Inter, Suppl. 2012;2(1):1–138.
- Buuren S, Groothuis-Oudshoorn K. Mice: multivariate imputation by chained equations inR. J. Stat. Soft. 2011;45(3):1–67.
- Devani K, Charilaou P, Radadiya D, et al. Acute pancreatitis: trends in outcomes and the role of acute kidney injury in mortality– a propensity-matched analysis. Pancreatology. 2018;18(8):870–877.
- Wu WT, Li YJ, Feng AZ, et al. Data mining in clinical big data: the frequently used databases, steps, and methodological models. Mil Med Res. 2021;8(1):44.
- Lankisch PG, Burchard-Reckert S, Petersen M, et al. Etiology and age have only a limited influence on the course of acute pancreatitis. Pancreas. 1996;13(4):344–349.
- Halonen KI, Leppaniemi AK, Puolakkainen PA, et al. Severe acute pancreatitis: prognostic factors in 270 consecutive patients. Pancreas. 2000;21(3):266–271.
- Grams ME, Matsushita K, Sang Y, et al. Explaining the racial difference in AKI incidence. J Am Soc Nephrol. 2014;25(8):1834–1841.
- NaveenKumar SK, Thushara RM, Sundaram MS, et al. Unconjugated bilirubin exerts Pro-Apoptotic effect on platelets via p38-MAPK activation. Sci Rep. 2015;5(1):15045.
- Betjes MG, Bajema I. The pathology of jaundice-related renal insufficiency: cholemic nephrosis revisited. J Nephrol. 2006;19(2):229–233.
- Xia S, Ni Y, Zhou Q, et al. Emodin attenuates severe acute pancreatitis via antioxidant and anti-inflammatory activity. Inflammation. 2019;42(6):2129–2138.
- Bopanna S, Nayak B, Prakash S, et al. Increased oxidative stress and deficient antioxidant levels may be involved in the pathogenesis of idiopathic recurrent acute pancreatitis. Pancreatology. 2017;17(4):529–533.
- Mohammadi Kebar S, Hosseini Nia S, Maleki N, et al. The incidence rate, risk factors and clinical outcome of acute kidney injury in critical patients. Iran J Public Health. 2018;47(11):1717–1724.
- Cuthbertson CM, Christophi C. Disturbances of the microcirculation in acute pancreatitis. Br J Surg. 2006;93(5):518–530.
- Chu LP, Yu YF, Guo LC, et al. Predictive value of complement and coagulation indicators in sepsis related acute kidney injury. Zhonghua Nei Ke Za Zhi. 2020;59(11):854–859.
- Pan L, Mo M, Huang A, et al. Coagulation parameters may predict clinical outcomes in patients with septic acute kidney injury. CN. 2021;96(5):253–262.
- Liu C, Zhou X, Ling L, et al. Prediction of mortality and organ failure based on coagulation and fibrinolysis markers in patients with acute pancreatitis: a retrospective study. Medicine. 2019;98(21):e15648.
- Shi N, Sun GD, Ji YY, et al. Effects of acute kidney injury on acute pancreatitis patients’ survival rate in intensive care unit: a retrospective study. WJG. 2021;27(38):6453–6464.
- Annane D, Renault A, Brun-Buisson C, et al.; CRICS-TRIGGERSEP Network. Hydrocortisone plus fludrocortisone for adults with septic shock. N Engl J Med. 2018;378(9):809–818.
- Mistry M, Muirhead EE, Yamaguchi Y, et al. Renal function in rats with angiotensin II-salt-induced hypertension: effect of thromboxane synthesis inhibition and receptor blockade. J Hypertens. 1990;8(1):75–83.
- Fischer AJ, Andreottola F, Lenz P, et al. Acute pancreatitis in intensive care medicine: which risk score is useful?. Med Klin Intensivmed Notfmed. 2017;112(8):717–723.
- Beger HG, Rau BM. Severe acute pancreatitis: clinical course and management. WJG. 2007;13(38):5043–5051.
- Rau BM, Bothe A, Kron M, et al. Role of early multisystem organ failure as major risk factor for pancreatic infections and death in severe acute pancreatitis. Clin Gastroenterol Hepatol. 2006;4(8):1053–1061.
- Jain S, Mahapatra SJ, Gupta S, et al. Infected pancreatic necrosis due to multidrug-resistant organisms and persistent organ failure predict mortality in acute pancreatitis. Clin Transl Gastroenterol. 2018;9(10):190.
- Zhang L, Wang Z, Zhou Z, et al. Developing an ensemble machine learning model for early prediction of sepsis-associated acute kidney injury. iScience. 2022;25(9):104932.
- Nassar TI, Qunibi WY. AKI associated with acute pancreatitis. CJASN. 2019;14(7):1106–1115.
- Knaus WA, Draper EA, Wagner DP, et al. APACHE II: a severity of disease classification system. Crit Care Med. 1985;13(10):818–829.
- Wu BU, Johannes RS, Sun X, et al. The early prediction of mortality in acute pancreatitis: a large population-based study. Gut. 2008;57(12):1698–1703.
- Ranson JH, Rifkind KM, Roses DF, et al. Objective early identification of severe acute pancreatitis. Am J Gastroenterol. 1974;61(6):443–451.
Appendix 1.
Data extraction variables and deletion ratios of patients with acute pancreatitis in MIMIC-IV 1.0 database
Appendix 2.
Baseline characteristics of the populations studied in two groups, the training cohort and the validation cohort