A novel risk-predicted nomogram for sepsis associated-acute kidney injury among critically ill patients

Development and external validation of a nomogram for the early prediction of acute kidney injury in septic patients: a multicenter retrospective clinical study

Background and purpose: Acute kidney injury (AKI) is a common serious complication in sepsis patients with a high mortality rate. This study aimed to develop and validate a predictive model for sepsis associated acute kidney injury (SA-AKI). Methods: In our study, we retrospectively constructed a development cohort comprising 733 septic patients admitted to eight Grade-A tertiary hospitals in Shanghai from January 2021 to October 2022. Additionally, we established an external validation cohort consisting of 336 septic patients admitted to our hospital from January 2017 to December 2019. Risk predictors were selected by LASSO regression, and a corresponding nomogram was constructed. We evaluated the model's discrimination, precision and clinical benefit through receiver operating characteristic (ROC) curves, calibration plots, decision curve analysis (DCA) and clinical impact curves (CIC) in both internal and external validation. Results: AKI incidence was 53.2% in the development cohort and 48.2% in the external validation cohort. The model included five independent indicators: chronic kidney disease stages 1 to 3, blood urea nitrogen, procalcitonin, D-dimer and creatine kinase isoenzyme. The AUC of the model in the development and validation cohorts was 0.914 (95% CI, 0.894-0.934) and 0.923 (95% CI, 0.895-0.952), respectively. The calibration plot, DCA, and CIC demonstrated the model's favorable clinical applicability. Conclusion: We developed and validated a robust nomogram model, which might identify patients at risk of SA-AKI and promising for clinical applications.

Predictive models of sepsis-associated acute kidney injury based on machine learning: a scoping review

BACKGROUND

With the development of artificial intelligence, the application of machine learning to develop predictive models for sepsis-associated acute kidney injury has made potential breakthroughs in early identification, grading, diagnosis, and prognosis determination.

METHODS

Here, we conducted a systematic search of the PubMed, Cochrane Library, Embase (Ovid), Web of Science, and Scopus databases on April 28, 2023, and screened relevant literature. Then, we comprehensively extracted relevant data related to machine learning algorithms, predictors, and predicted objectives. We subsequently performed a critical evaluation of research quality, data aggregation, and analyses.

RESULTS

We screened 25 studies on predictive models for sepsis-associated acute kidney injury from a total of originally identified 2898 studies. The most commonly used machine learning algorithm is traditional logistic regression, followed by eXtreme gradient boosting. We categorized these predictive models into early identification models (60%), prognostic prediction models (32%), and subtype identification models (8%) according to their predictive purpose. The five most commonly used predictors were serum creatinine levels, lactate levels, age, blood urea nitrogen concentration, and diabetes mellitus. In addition, a single data source, insufficient assessment of clinical utility, lack of model bias assessment, and hyperparameter adjustment may be the main reasons for the low quality of the current research.

CONCLUSIONS

However, studies on the nondeath prognostic outcomes, the long-term clinical outcomes, and the subtype identification models are insufficient. Additionally, the poor quality of the research and the insufficient practicality of the model are problems that need to be addressed urgently.

Clinical Characteristics and Risk Factors of Sepsis in Patients with Liver Abscess

Aims/Background Liver abscess (LA) is a serious medical condition that predisposes patients to sepsis. However, predicting sepsis in LA patients has rarely been explored. This study employed univariate and multivariate logistic regression analyses to identify independent risk factors for sepsis, which would provide guidance for clinical diagnosis and treatment. Methods A total of 122 patients with LA treated in Peking University People's Hospital from 1 January 2016 to 31 October 2022 were recruited. Among the cases, 35 patients had sepsis (sepsis group) while the remaining 87 did not have sepsis (non-sepsis group). Clinical data were collected for all enrolled cases. Univariate analysis was performed to identify potential predictors, which were tested in multivariable logistic analysis to pinpoint the independent risk factors for sepsis in LA patients; these findings were utilized to develop a prediction model. Receiver operating characteristic (ROC) curve was used to evaluate the diagnostic efficacy of the prediction model. Informed consent to participate was obtained from the patients or their relatives. Results The incidence of shivering in the sepsis group was significantly higher than that in the non-sepsis group (p < 0.05). Through the univariate analysis, it was found that the reduction in platelet count and prothrombin time activity and the elevation of glycosylated hemoglobin (HbAlc) and procalcitonin (PCT) were more significant in the sepsis group than in the non-sepsis group (p < 0.05). Multivariate logistic regression analysis revealed that PCT and HbAlc were independent risk predictors of sepsis in LA patients within the derivation cohort (p < 0.05). Conclusion Elevated levels of HbAlc and PCT were independent risk factors for sepsis associated with LA. Patients with LA exhibiting elevated PCT levels demonstrated a 21% increased susceptibility to sepsis, and those with elevated HbAlc levels showed a 38% heightened risk for sepsis.

Nomogram predicting early urinary incontinence after radical prostatectomy

PURPOSE

One of the most frequent side effects of radical prostatectomy (RP) is urinary incontinence. The primary cause of urine incontinence is usually thought to be impaired urethral sphincter function; nevertheless, the pathophysiology and recovery process of urine incontinence remains unclear. This study aimed to identify potential risk variables, build a risk prediction tool that considers preoperative urodynamic findings, and direct doctors to take necessary action to reduce the likelihood of developing early urinary incontinence.

METHODS

We retrospectively screened patients who underwent radical prostatectomy between January 1, 2020 and December 31, 2023 at the First People 's Hospital of Nantong, China. According to nomogram results, patients who developed incontinence within three months were classified as having early incontinence. The training group's general characteristics were first screened using univariate logistic analysis, and the LASSO method was applied for the best prediction. Multivariate logistic regression analysis was carried out to determine independent risk factors for early postoperative urine incontinence in the training group and to create nomograms that predict the likelihood of developing early urinary incontinence. The model was internally validated by computing the performance of the validation cohort. The nomogram discrimination, correction, and clinical usefulness were assessed using the c-index, receiver operating characteristic curve, correction plot, and clinical decision curve.

RESULTS

The study involved 142 patients in all. Multivariate logistic regression analysis following RP found seven independent risk variables for early urinary incontinence. A nomogram was constructed based on these independent risk factors. The training and validation groups' c-indices showed that the model had high accuracy and stability. The calibration curve demonstrates that the corrective effect of the training and verification groups is perfect, and the area under the receiver operating characteristic curve indicates great identification capacity. Using a nomogram, the clinical net benefit was maximised within a probability threshold of 0.01-1, according to decision curve analysis (DCA).

CONCLUSION

The nomogram model created in this study can offer a clear, personalised analysis of the risk of early urine incontinence following RP. It is highly discriminatory and accurate, and it can help create efficient preventative measures and identify high-risk populations.

A clinical risk score for predicting acute kidney injury in sepsis patients receiving normal saline in Northern Thailand: a retrospective cohort study

BACKGROUND

Normal saline is commonly used for resuscitation in sepsis patients but has a high chloride content, potentially increasing the risk of acute kidney injury (AKI). This study evaluated risk factors and developed a predictive risk score for AKI in sepsis patients treated with normal saline.

METHODS

This retrospective cohort study utilized the medical and electronic health records of sepsis patients who received normal saline between January 2018 and May 2020. Predictors of AKI used to construct the predictive risk score were identified through multivariate logistic regression models, with discrimination and calibration assessed using the area under the receiver operating characteristic curve (AUROC) and the expected-to-observed (E/O) ratio. Internal validation was conducted using bootstrapping techniques.

RESULTS

AKI was reported in 211 of 735 patients (28.7%). Eight potential risk factors, including norepinephrine, the Acute Physiology and Chronic Health Evaluation II score, serum chloride, respiratory failure with invasive mechanical ventilation, nephrotoxic antimicrobial drug use, history of angiotensin-converting enzyme inhibitors/angiotensin receptor blockers use, history of liver disease, and serum creatinine were used to create the NACl RENAL-Cr score. The model demonstrated good discrimination and calibration (AUROC, 0.79; E/O, 1). The optimal cutoff was 2.5 points, with corresponding sensitivity, specificity, positive predictive value, and negative predictive value scores of 71.6%, 72.5%, 51.2%, and 86.4%, respectively.

CONCLUSIONS

The NACl RENAL-Cr score, consisting of eight critical variables, was used to predict AKI in sepsis patients who received normal saline. This tool can assist healthcare professionals when deciding on sepsis treatment and AKI monitoring.

Machine learning for prediction of acute kidney injury in patients diagnosed with sepsis in critical care

BACKGROUND AND OBJECTIVE

Acute Kidney Injury (AKI) is a common and severe complication in patients diagnosed with sepsis. It is associated with higher mortality rates, prolonged hospital stays, increased utilization of medical resources, and financial burden on patients' families. This study aimed to establish and validate predictive models using machine learning algorithms to accurately predict the occurrence of AKI in patients diagnosed with sepsis.

METHODS

This retrospective study utilized real observational data from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. It included patients aged 18 to 90 years diagnosed with sepsis who were admitted to the ICU for the first time and had hospital stays exceeding 48 hours. Predictive models, employing various machine learning algorithms including Light Gradient Boosting Machine (LightGBM), EXtreme Gradient Boosting (XGBoost), Random Forest (RF), Decision Tree (DT), Artificial Neural Network (ANN), Support Vector Machine (SVM), and Logistic Regression (LR), were developed. The dataset was randomly divided into training and test sets at a ratio of 4:1.

RESULTS

A total of 10,575 sepsis patients were included in the analysis, of whom 8,575 (81.1%) developed AKI during hospitalization. A selection of 47 variables was utilized for model construction. The models derived from LightGBM, XGBoost, RF, DT, ANN, SVM, and LR achieved AUCs of 0.801, 0.773, 0.772, 0.737, 0.720, 0.765, and 0.776, respectively. Among these models, LightGBM demonstrated the most superior predictive performance.

CONCLUSIONS

These machine learning models offer valuable predictive capabilities for identifying AKI in patients diagnosed with sepsis. The LightGBM model, with its superior predictive capability, could aid clinicians in early identification of high-risk patients.

Development and validation of a machine learning model to predict the use of renal replacement therapy in 14,374 patients with COVID-19

PURPOSE

To develop a model to predict the use of renal replacement therapy (RRT) in COVID-19 patients.

MATERIALS AND METHODS

Retrospective analysis of multicenter cohort of intensive care unit (ICU) admissions of Brazil involving COVID-19 critically adult patients, requiring ventilatory support, admitted to 126 Brazilian ICUs, from February 2020 to December 2021 (development) and January to May 2022 (validation). No interventions were performed.

RESULTS

Eight machine learning models' classifications were evaluated. Models were developed using an 80/20 testing/train split ratio and cross-validation. Thirteen candidate predictors were selected using the Recursive Feature Elimination (RFE) algorithm. Discrimination and calibration were assessed. Temporal validation was performed using data from 2022. Of 14,374 COVID-19 patients with initial respiratory support, 1924 (13%) required RRT. RRT patients were older (65 [53-75] vs. 55 [42-68]), had more comorbidities (Charlson's Comorbidity Index 1.0 [0.00-2.00] vs 0.0 [0.00-1.00]), had higher severity (SAPS-3 median: 61 [51-74] vs 48 [41-58]), and had higher in-hospital mortality (71% vs 22%) compared to non-RRT. Risk factors for RRT, such as Creatinine, Glasgow Coma Scale, Urea, Invasive Mechanical Ventilation, Age, Chronic Kidney Disease, Platelets count, Vasopressors, Noninvasive Ventilation, Hypertension, Diabetes, modified frailty index (mFI) and Gender, were identified. The best discrimination and calibration were found in the Random Forest (AUC [95%CI]: 0.78 [0.75-0.81] and Brier's Score: 0.09 [95%CI: 0.08-0.10]). The final model (Random Forest) showed comparable performance in the temporal validation (AUC [95%CI]: 0.79 [0.75-0.84] and Brier's Score, 0.08 [95%CI: 0.08-0.1]).

CONCLUSIONS

An early ML model using easily available clinical and laboratory data accurately predicted the use of RRT in critically ill patients with COVID-19. Our study demonstrates that using ML techniques is feasible to provide early prediction of use of RRT in COVID-19 patients.

A nomogram for predicting lymph nodes metastasis at the inferior mesenteric artery in rectal cancer: a retrospective case-control study

According to past and current literature, metastasis of the lymph nodes at the inferior mesenteric artery (IMA-LN), also known as 253LN of colorectal cancer has been seldom investigated. To date, there are still controversies on whether the 253LN need to be routinely cleaned. Using specific criteria, 347 patients who underwent radical resection for rectal cancer between April 2019 and July 2022 were selected for the study. Logistic regression was used to determine the likelihood that a patient may suffer 253LN metastasis, and a nomogram for 253LN metastasis subsequently developed. The c-index and calibration curve were used to evaluate precision and discrimination in the nomogram, and the appropriateness of the final nomogram for the clinical setting determined using decision curve analysis (DCA). 253LN metastases appeared in the pathological specimens of 29 (8.4%) of the selected patients. Logistic regression showed that preoperative parameters including serum carcinoembryonic antigen (CEA) value ( > 5 ng / ml, OR = 2.894, P = 0.023), distance from anal margin (> 9 cm, OR = 2.406, P = 0.045) and degree of differentiation (poor, OR = 9.712, P < 0.001) were significantly associated with 253LN metastasis. A nomogram to predict 253LN metastasis in rectal cancer was developed and showed considerable discrimination and good precision (c-index = 0.750). Furthermore, DCA confirmed that the nomogram has some feasibility for the clinical environment. Clinicopathological and radiological patient data can be pivotal for making surgical decisions relating to 253LN metastasis. A nomogram was developed using this data, providing an objective method that can significantly improve prognoses in colorectal cancer.

Classification and Regression Tree Predictive Model for Acute Kidney Injury in Traumatic Brain Injury Patients

Background

Acute kidney injury (AKI) is prevalent in hospitalized patients with traumatic brain injury (TBI), and increases the risk of poor outcomes. We designed this study to develop a visual and convenient decision-tree-based model for predicting AKI in TBI patients.

Methods

A total of 376 patients admitted to the emergency department of the West China Hospital for TBI between January 2015 and June 2019 were included. Demographic information, vital signs on admission, laboratory test results, radiological signs, surgical options, and medications were recorded as variables. AKI was confirmed since the second day after admission, based on the Kidney Disease Improving Global Outcomes criteria. We constructed two predictive models for AKI using least absolute shrinkage and selection operator (LASSO) regression and classification and regression tree (CART), respectively. Receiver operating characteristic (ROC) curves of these two predictive models were drawn, and the area under the ROC curve (AUC) was calculated to compare their predictive accuracy.

Results

The incidence of AKI on the second day after admission was 10.4% among patients with TBI. Lasso regression identified five potent predictive factors for AKI: glucose, serum creatinine, cystatin C, serum uric acid, and fresh frozen plasma transfusions. The CART analysis showed that glucose, serum uric acid, and cystatin C ranked among the top three in terms of the feature importance of the decision tree model. The AUC value of the decision-tree predictive model was 0.892, which was higher than the 0.854 of the LASSO regression model, although the difference was not statistically significant.

Conclusion

The decision tree model is valuable for predicting AKI among patients with TBI. This tree-based flowchart is convenient for physicians to identify patients with TBI who are at high risk of AKI and prompts them to develop suitable therapeutic strategies.

Development and validation of a nomogram for predicting sepsis in patients with pyogenic liver abscess

Pyogenic liver abscess (PLA) is a severe condition that significantly increases the risk of sepsis. However, there is a notable dearth of research regarding the prediction of sepsis in PLA patients. The objective of this study was to develop and validate a prognostic nomogram for predicting sepsis in PLA patients. A total of 206 PLA patients were enrolled in our study, out of which 60 individuals (29.1%) met the Sepsis-3 criteria. Independent risk factors for sepsis were identified through univariate and multivariate logistic regression analyses. Subsequently, a nomogram was developed based on age, positive blood culture, procalcitonin, alanine aminotransferase, blood urea nitrogen, and D-dimer. The nomogram demonstrated excellent calibration and discrimination, as evidenced by the area under the receiver operating characteristic curve (AUC) values of 0.946 (95% confidence interval [CI], 0.912-0.979) and 0.980 (95%CI 0.951-1.000) in the derivation and validation cohorts, respectively. Furthermore, decision-curve analysis confirmed the clinical utility of the nomogram. This study provides valuable insights for the prevention of sepsis in PLA patients and underscores the potential application of the prognostic nomogram in clinical practice.

Nomogram for predicting fulminant necrotizing enterocolitis

BACKGROUND

Fulminant necrotizing enterocolitis (FNEC) is the most serious subtype of NEC and has a high mortality rate and a high incidence of sequelae. Onset prediction can help in the establishment of a customized treatment strategy. This study aimed to develop and evaluate a predictive nomogram for FNEC.

METHODS

We conducted a retrospective observation to study the clinical data of neonates diagnosed with NEC (Bell stage ≥ IIB). Neonates were divided into the FNEC and NEC groups. A multivariate logistic regression model was used to construct the nomogram model. The performance of the nomogram was assessed using area under the curve, calibration analysis, and decision curve analysis.

RESULTS

A total of 206 neonate cases were included, among which 40 (19.4%) fulfilled the definition of FNEC. The identified predictors were assisted ventilation after NEC onset; shock at NEC onset; feeding volumes before NEC onset; neutrophil counts on the day of NEC onset; and neutrophil, lymphocyte, and monocyte counts on day 1 after NEC onset. The nomogram exhibited good discrimination, with an area under the receiver operating characteristic curve of 0.884 (95% CI 0.825-0.943). The predictive model was well calibrated. Decision curve analysis confirmed the clinical usefulness of this nomogram.

CONCLUSION

A nomogram with a potentially effective application was developed to facilitate the individualized prediction of FNEC, with the hope of providing further direction for the early diagnosis of FNEC and timing of intervention.

Development and validation of a prognostic nomogram for Takotsubo syndrome patients in the intensive care units: a retrospective cohort study

Patients with Takotsubo syndrome (TTS) admitted to the intensive care unit (ICU) always confront a higher risk of in-hospital death than those hospitalized in the cardiology unit. The prognosis of the latter was analyzed by a large number of studies. However, there was no utility model to predict the risk of in-hospital death for patients with TTS in the ICU. This study aimed to establish a model predicting in-hospital death in patients with TTS admitted to ICU. We retrospectively included ICU patients with TTS from the MIMIC-IV database. The outcome of the nomogram was in-hospital death. Least Absolute Shrinkage Selection Operator (LASSO) analysis selected predictors preliminarily. The model was developed by multivariable logistic regression analysis. Calibration, decision curve analysis (DCA), and receiver operating characteristic (ROC) measured the performance of the nomogram on the accuracy, clinical utility, and discrimination, respectively. Eventually, 368 ICU patients with TTS were enrolled in this research. The in-hospital mortality was 13.04%. LASSO regression and multivariate logistic regression analysis verified risk factors significantly associated with in-hospital mortality. They were potassium, prothrombin time (PT), age, myocardial infarction, white cell count (WBC), hematocrit, anion gap, and sequential organ failure assessment (SOFA) score. This nomogram excellently discriminated against patients with a risk of in-hospital death. The area under curve (AUC) was 0.779 (95%CI: 0.732-0.826) in training set and 0.775 (95%CI: 0.711-0.839) in test set. The calibration plot and DCA showed good clinical benefits for this nomogram. We developed a nomogram that predicts the probability of in-hospital death for ICU patients with TTS. This nomogram was able to discriminate patients with a high risk of in-hospital death and performed clinical utility.

Interaction of hemoglobin, transfusion, and acute kidney injury in patients undergoing cardiopulmonary bypass: a group-based trajectory analysis

Anemia is a risk factor for acute kidney injury (AKI) following cardiopulmonary bypass (CPB). Whether red blood cell (RBC) transfusion-enhanced hemoglobin levels contribute to low AKI rates remains unclear. We investigated the interaction between hemoglobin, RBC transfusion, and AKI after CPB. Hemoglobin trajectories within 72 h were analyzed using group-based trajectory analysis. Multivariable logistic analysis and inverse probability-weighted regression were adopted to evaluate the associations between hemoglobin and AKI in RBC and non-RBC transfusion subgroups. We analyzed 6226 patients’ data. In the transfusion subgroup, three hemoglobin trajectories were identified. The AKI incidence was lowest in the trajectory with the lowest hemoglobin level (trajectory 1, less transfusion), and it was comparable in trajectories 2 and 3 (20.7% vs. 32.7% vs. 29.4%, p < 0.001, respectively). In four logistic models, the odds ratio for AKI with trajectory 1 as the reference ranged from 1.44 to 1.85 for trajectory 2 (p < 0.001) and 1.45 to 1.66 for trajectory 3 (p < 0.050). The average treatment effect on AKI was 5.6% (p = 0.009) for trajectory 2 and 7.5% (p = 0.041) for trajectory 3, with trajectory 1 as the reference. In the non-RBC transfusion subgroup, three approximately linear hemoglobin trajectories (9, 10, and 12 g/dL) were observed; however, both the crude and adjusted AKI incidence were similar within the three trajectories. In patients undergoing CPB, hemoglobin level >9 g/dL was not associated with decreased AKI incidence in the subgroup without RBC transfusion. However, in patients with RBC transfusion, maintaining hemoglobin level >9 g/dL by RBC transfusion was associated with increased AKI incidence.

A nomogram to predict hyperkalemia in patients with hemodialysis: a retrospective cohort study

BACKGROUND

Hyperkalemia increases the risk of mortality and cardiovascular-related hospitalizations in patients with hemodialysis. Predictors of hyperkalemia are yet to be identified. We aimed at developing a nomogram able to predict hyperkalemia in patients with hemodialysis.

METHODS

We retrospectively screened patients with end-stage renal disease (ESRD) who had regularly received hemodialysis between Jan 1, 2017, and Aug 31, 2021, at Lishui municipal central hospital in China. The outcome for the nomogram was hyperkalemia, defined as serum potassium [K⁺] ≥ 5.5 mmol/L. Data were collected from hemodialysis management system. Least Absolute Shrinkage Selection Operator (LASSO) analysis selected predictors preliminarily. A prediction model was constructed by multivariate logistic regression and presented as a nomogram. The performance of nomogram was measured by the receiver operating characteristic (ROC) curve, calibration diagram, and decision curve analysis (DCA). This model was validated internally by calculating the performance on a validation cohort.

RESULTS

A total of 401 patients were enrolled in this study. 159 (39.65%) patients were hyperkalemia. All participants were divided into development (n = 256) and validation (n = 145) cohorts randomly. Predictors in this nomogram were the number of hemodialysis session, blood urea nitrogen (BUN), serum sodium, serum calcium, serum phosphorus, and diabetes. The ROC curve of the training set was 0.82 (95%CI 0.77, 0.88). Similar ROC curve was achieved at validation set 0.81 (0.74, 0.88). The calibration curve demonstrated that the prediction outcome was correlated with the observed outcome.

CONCLUSION

This nomogram helps clinicians in predicting the risk of PEW and managing serum potassium in the patients with hemodialysis.

Development of a risk prediction nomogram for sarcopenia in hemodialysis patients

Background

Sarcopenia is associated with various adverse outcomes in hemodialysis patients. However, current tools for assessing and diagnosing sarcopenia have limited applicability. In this study, we aimed to develop a simple and reliable nomogram to predict the risk of sarcopenia in hemodialysis patients that could assist physicians identify high-risk patients early.

Methods

A total of 615 patients undergoing hemodialysis at the First Affiliated Hospital College of Medicine Zhejiang University between March to June 2021 were included. They were randomly divided into either the development cohort (n = 369) or the validation cohort (n = 246). Multivariable logistic regression analysis was used to screen statistically significant variables for constructing the risk prediction nomogram for Sarcopenia. The line plots were drawn to evaluate the effectiveness of the nomogram in three aspects, namely differentiation, calibration, and clinical net benefit, and were further validated by the Bootstrap method.

Results

The study finally included five clinical factors to construct the nomogram, including age, C-reactive protein, serum phosphorus, body mass index, and mid-upper arm muscle circumference, and constructed a nomogram. The area under the ROC curve of the line chart model was 0.869, with a sensitivity and specificity of 77% sensitivity and 83%, the Youden index was 0.60, and the internal verification C-statistic was 0.783.

Conclusions

This study developed and validated a nomogram model to predict the risk of sarcopenia in hemodialysis patients, which can be used for early identification and timely intervention in high-risk groups.

Machine learning for the prediction of acute kidney injury in patients with sepsis

BACKGROUND

Acute kidney injury (AKI) is the most common and serious complication of sepsis, accompanied by high mortality and disease burden. The early prediction of AKI is critical for timely intervention and ultimately improves prognosis. This study aims to establish and validate predictive models based on novel machine learning (ML) algorithms for AKI in critically ill patients with sepsis.

METHODS

Data of patients with sepsis were extracted from the Medical Information Mart for Intensive Care III (MIMIC- III) database. Feature selection was performed using a Boruta algorithm. ML algorithms such as logistic regression (LR), k-nearest neighbors (KNN), support vector machine (SVM), decision tree, random forest, Extreme Gradient Boosting (XGBoost), and artificial neural network (ANN) were applied for model construction by utilizing tenfold cross-validation. The performances of these models were assessed in terms of discrimination, calibration, and clinical application. Moreover, the discrimination of ML-based models was compared with those of Sequential Organ Failure Assessment (SOFA) and the customized Simplified Acute Physiology Score (SAPS) II model.

RESULTS

A total of 3176 critically ill patients with sepsis were included for analysis, of which 2397 cases (75.5%) developed AKI during hospitalization. A total of 36 variables were selected for model construction. The models of LR, KNN, SVM, decision tree, random forest, ANN, XGBoost, SOFA and SAPS II score were established and obtained area under the receiver operating characteristic curves of 0.7365, 0.6637, 0.7353, 0.7492, 0.7787, 0.7547, 0.821, 0.6457 and 0.7015, respectively. The XGBoost model had the best predictive performance in terms of discrimination, calibration, and clinical application among all models.

CONCLUSION

The ML models can be reliable tools for predicting AKI in septic patients. The XGBoost model has the best predictive performance, which can be used to assist clinicians in identifying high-risk patients and implementing early interventions to reduce mortality.

Artificial Intelligence for Risk Prediction of Rehospitalization with Acute Kidney Injury in Sepsis Survivors

Sepsis survivors have a higher risk of long-term complications. Acute kidney injury (AKI) may still be common among sepsis survivors after discharge from sepsis. Therefore, our study utilized an artificial-intelligence-based machine learning approach to predict future risks of rehospitalization with AKI between 1 January 2008 and 31 December 2018. We included a total of 23,761 patients aged ≥ 20 years who were admitted due to sepsis and survived to discharge. We adopted a machine learning method by using models based on logistic regression, random forest, extra tree classifier, gradient boosting decision tree (GBDT), extreme gradient boosting, and light gradient boosting machine (LGBM). The LGBM model exhibited the highest area under the receiver operating characteristic curves (AUCs) of 0.816 to predict rehospitalization with AKI in sepsis survivors and followed by the GBDT model with AUCs of 0.813. The top five most important features in the LGBM model were C-reactive protein, white blood cell counts, use of inotropes, blood urea nitrogen and use of diuretics. We established machine learning models for the prediction of the risk of rehospitalization with AKI in sepsis survivors, and the machine learning model may set the stage for the broader use of clinical features in healthcare.