Prediction models for acute kidney injury in patients with gastrointestinal cancers: a real-world study based on Bayesian networks

Acute kidney injury after radical gastrectomy: incidence, risk factors, and impact on prognosis

Background

Acute kidney injury (AKI) is a serious adverse event often overlooked following major abdominal surgery. While radical gastrectomy stands as the primary curative method for treating gastric cancer patients, little information exists regarding AKI post-surgery. Hence, this study aimed to ascertain the incidence rate, risk factors, and consequences of AKI among patients undergoing radical gastrectomy.

Methods

This was a population-based, retrospective cohort study. The incidence of AKI was calculated. Multivariate logistic regression was used to identify independent predictors of AKI. Survival curves were plotted by using the Kaplan-Meier method and differences in survival rates between groups were analyzed by using the log-rank test.

Results

Of the 2,875 patients enrolled in this study, 61 (2.1%) developed postoperative AKI, with AKI Network 1, 2, and 3 in 50 (82.0%), 6 (9.8%), and 5 (8.2%), respectively. Of these, 49 patients had fully recovered by discharge. Risk factors for AKI after radical gastrectomy were preoperative hypertension (odds ratio [OR], 1.877; 95% CI, 1.064-3.311; P = 0.030), intraoperative blood loss (OR, 1.001; 95% CI, 1.000-1.002; P = 0.023), operation time (OR, 1.303; 95% CI, 1.030-1.649; P = 0.027), and postoperative intensive care unit (ICU) admission (OR, 4.303; 95% CI, 2.301-8.045; P < 0.001). The probability of postoperative complications, mortality during hospitalization, and length of stay in patients with AKI after surgery were significantly higher than those in patients without AKI. There was no statistical difference in overall survival (OS) rates between patients with AKI and without AKI (1-year, 3-year, 5-year overall survival rates of patients with AKI and without AKI were 93.3% vs 92.0%, 70.9% vs 73.6%, and 57.1% vs 67.1%, respectively, P = 0.137).

Conclusions

AKI following radical gastrectomy is relatively rare and typically self-limited. AKI is linked with preoperative hypertension, intraoperative blood loss, operation time, and postoperative ICU admission. While AKI raises the likelihood of postoperative complications, it does not affect OS.

Diabetes mellitus early warning and factor analysis using ensemble Bayesian networks with SMOTE-ENN and Boruta

Diabetes mellitus (DM) has become the third chronic non-infectious disease affecting patients after tumor, cardiovascular and cerebrovascular diseases, becoming one of the major public health issues worldwide. Detection of early warning risk factors for DM is key to the prevention of DM, which has been the focus of some previous studies. Therefore, from the perspective of residents' self-management and prevention, this study constructed Bayesian networks (BNs) combining feature screening and multiple resampling techniques for DM monitoring data with a class imbalance in Shanxi Province, China, to detect risk factors in chronic disease monitoring programs and predict the risk of DM. First, univariate analysis and Boruta feature selection algorithm were employed to conduct the preliminary screening of all included risk factors. Then, three resampling techniques, SMOTE, Borderline-SMOTE (BL-SMOTE) and SMOTE-ENN, were adopted to deal with data imbalance. Finally, BNs developed by three algorithms (Tabu, Hill-climbing and MMHC) were constructed using the processed data to find the warning factors that strongly correlate with DM. The results showed that the accuracy of DM classification is significantly improved by the BNs constructed by processed data. In particular, the BNs combined with the SMOTE-ENN resampling improved the most, and the BNs constructed by the Tabu algorithm obtained the best classification performance compared with the hill-climbing and MMHC algorithms. The best-performing joint Boruta-SMOTE-ENN-Tabu model showed that the risk factors of DM included family history, age, central obesity, hyperlipidemia, salt reduction, occupation, heart rate, and BMI.

Bayesian networks and imaging-derived phenotypes highlight the role of fat deposition in COVID-19 hospitalisation risk

Objective: Obesity is a significant risk factor for adverse outcomes following coronavirus infection (COVID-19). However, BMI fails to capture differences in the body fat distribution, the critical driver of metabolic health. Conventional statistical methodologies lack functionality to investigate the causality between fat distribution and disease outcomes. Methods: We applied Bayesian network (BN) modelling to explore the mechanistic link between body fat deposition and hospitalisation risk in 459 participants with COVID-19 (395 non-hospitalised and 64 hospitalised). MRI-derived measures of visceral adipose tissue (VAT), subcutaneous adipose tissue (SAT), and liver fat were included. Conditional probability queries were performed to estimate the probability of hospitalisation after fixing the value of specific network variables. Results: The probability of hospitalisation was 18% higher in people living with obesity than those with normal weight, with elevated VAT being the primary determinant of obesity-related risk. Across all BMI categories, elevated VAT and liver fat (>10%) were associated with a 39% mean increase in the probability of hospitalisation. Among those with normal weight, reducing liver fat content from >10% to <5% reduced hospitalisation risk by 29%. Conclusion: Body fat distribution is a critical determinant of COVID-19 hospitalisation risk. BN modelling and probabilistic inferences assist our understanding of the mechanistic associations between imaging-derived phenotypes and COVID-19 hospitalisation risk.

Characterization of Risk Prediction Models for Acute Kidney Injury: A Systematic Review and Meta-analysis

Importance

Despite the expansion of published prediction models for acute kidney injury (AKI), there is little evidence of uptake of these models beyond their local derivation nor data on their association with patient outcomes.

Objective

To systematically review published AKI prediction models across all clinical subsettings.

Data Sources

MEDLINE via PubMed (January 1946 to April 2021) and Embase (January 1947 to April 2021) were searched using medical subject headings and text words related to AKI and prediction models.

Study Selection

All studies that developed a prediction model for AKI, defined as a statistical model with at least 2 predictive variables to estimate future occurrence of AKI, were eligible for inclusion. There was no limitation on study populations or methodological designs.

Data Extraction and Synthesis

Two authors independently searched the literature, screened the studies, and extracted and analyzed the data following the Preferred Reporting Items for Systematic Review and Meta-analyses guideline. The data were pooled using a random-effects model, with subgroups defined by 4 clinical settings. Between-study heterogeneity was explored using multiple methods, and funnel plot analysis was used to identify publication bias.

Main Outcomes and Measures

C statistic was used to measure the discrimination of prediction models.

Results

Of the 6955 studies initially identified through literature searching, 150 studies, with 14.4 million participants, met the inclusion criteria. The study characteristics differed widely in design, population, AKI definition, and model performance assessments. The overall pooled C statistic was 0.80 (95% CI, 0.79-0.81), with pooled C statistics in different clinical subsettings ranging from 0.78 (95% CI, 0.75-0.80) to 0.82 (95% CI, 0.78-0.86). Between-study heterogeneity was high overall and in the different clinical settings (eg, contrast medium-associated AKI: I2 = 99.9%; P < .001), and multiple methods did not identify any clear sources. A high proportion of models had a high risk of bias (126 [84.4%]) according to the Prediction Model Risk Of Bias Assessment Tool.

Conclusions and Relevance

In this study, the discrimination of the published AKI prediction models was good, reflected by high C statistics; however, the wide variation in the clinical settings, populations, and predictive variables likely drives the highly heterogenous findings that limit clinical utility. Standardized procedures for development and validation of prediction models are urgently needed.

Machine learning for acute kidney injury: Changing the traditional disease prediction mode

Acute kidney injury (AKI) is a serious clinical comorbidity with clear short-term and long-term prognostic implications for inpatients. The diversity of risk factors for AKI has been recognized in previous studies, and a series of predictive models have been developed using traditional statistical methods in conjunction with its preventability, but they have failed to meet the expectations in limited clinical applications, the rapid spread of electronic health records and artificial intelligence machine learning technology has brought new hope for the construction of AKI prediction models. In this article, we systematically review the definition and classification of machine learning methods, modeling ideas and evaluation methods, and the characteristics and current status of modeling studies. According to the modeling objectives, we subdivided them into critical care medical setting models, all medical environment models, special surgery models, special disease models, and special nephrotoxin exposure models. As the first review article to comprehensively summarize and analyze machine learning prediction models for AKI, we aim to objectively describe the advantages and disadvantages of machine learning approaches to modeling, and help other researchers more quickly and intuitively understand the current status of modeling research, inspire ideas and learn from experience, so as to guide and stimulate more research and more in-depth exploration in the future, which will ultimately provide greater help to improve the overall status of AKI diagnosis and treatment.

Poor glycaemic control and ectopic fat deposition mediates the increased risk of non-alcoholic steatohepatitis in high-risk populations with type 2 diabetes: Insights from Bayesian-network modelling

BACKGROUND

An estimated 55.5% and 37.3% of people globally with type 2 diabetes (T2D) will have concomitant non-alcoholic fatty liver disease (NAFLD) and the more severe fibroinflammatory stage, non-alcoholic steatohepatitis (NASH). NAFLD and NASH prevalence is projected to increase exponentially over the next 20 years. Bayesian Networks (BNs) offer a powerful tool for modelling uncertainty and visualising complex systems to provide important mechanistic insight.

METHODS

We applied BN modelling and probabilistic reasoning to explore the probability of NASH in two extensively phenotyped clinical cohorts: 1) 211 participants with T2D pooled from the MODIFY study & UK Biobank (UKBB) online resource; and 2) 135 participants without T2D from the UKBB. MRI-derived measures of visceral (VAT), subcutaneous (SAT), skeletal muscle (SMI), liver fat (MRI-PDFF), liver fibroinflammatory change (liver cT1) and pancreatic fat (MRI-PDFF) were combined with plasma biomarkers for network construction. NASH was defined according to liver PDFF >5.6% and liver cT1 >800ms. Conditional probability queries were performed to estimate the probability of NASH after fixing the value of specific network variables.

RESULTS

In the T2D cohort we observed a stepwise increase in the probability of NASH with each obesity classification (normal weight: 13%, overweight: 23%, obese: 36%, severe obesity: 62%). In the T2D and non-T2D cohorts, elevated (vs. normal) VAT conferred a 20% and 1% increase in the probability of NASH, respectively, while elevated SAT caused a 7% increase in NASH risk within the T2D cohort only. In those with T2D, reducing HbA1c from the 'high' to 'low' value reduced the probability of NASH by 22%.

CONCLUSION

Using BNs and probabilistic reasoning to study the probability of NASH, we highlighted the relative contribution of obesity, ectopic fat (VAT and liver) and glycaemic status to increased NASH risk, namely in people with T2D. Such modelling can provide insights into the efficacy and magnitude of public health and pharmacological interventions to reduce the societal burden of NASH.

Machine learning models for predicting acute kidney injury: a systematic review and critical appraisal

BACKGROUND

The number of studies applying machine learning (ML) to predict acute kidney injury (AKI) has grown steadily over the past decade. We assess and critically appraise the state of the art in ML models for AKI prediction, considering performance, methodological soundness, and applicability.

METHODS

We searched PubMed and ArXiv, extracted data, and critically appraised studies based on the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD), Checklist for Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies (CHARMS), and Prediction Model Risk of Bias Assessment Tool (PROBAST) guidelines.

RESULTS

Forty-six studies from 3166 titles were included. Thirty-eight studies developed a model, five developed and externally validated one, and three studies externally validated one. Flexible ML methods were used more often than deep learning, although the latter was common with temporal variables and text as predictors. Predictive performance showed an area under receiver operating curves ranging from 0.49 to 0.99. Our critical appraisal identified a high risk of bias in 39 studies. Some studies lacked internal validation, whereas external validation and interpretability of results were rarely considered. Fifteen studies focused on AKI prediction in the intensive care setting, and the US-derived Medical Information Mart for Intensive Care (MIMIC) data set was commonly used. Reproducibility was limited as data and code were usually unavailable.

CONCLUSIONS

Flexible ML methods are popular for the prediction of AKI, although more complex models based on deep learning are emerging. Our critical appraisal identified a high risk of bias in most models: Studies should use calibration measures and external validation more often, improve model interpretability, and share data and code to improve reproducibility.

Predicting mortality in acute kidney injury patients undergoing continuous renal replacement therapy using a visualization model: A retrospective study

Background: Patients with severe acute kidney injury (AKI) require continuous renal replacement therapy (CRRT) when hemodynamically unstable. We aimed to identify prognostic factors and develop a nomogram that could predict mortality in patients with AKI undergoing CRRT. Methods: Data were extracted from the Dryad Digital Repository. We enrolled 1,002 participants and grouped them randomly into training (n = 670) and verification (n = 332) datasets based on a 2:1 proportion. Based on Cox proportional modeling of the training set, we created a web-based dynamic nomogram to estimate all-cause mortality. Results: The model incorporated phosphate, Charlson comorbidity index, body mass index, mean arterial pressure, levels of creatinine and albumin, and sequential organ failure assessment scores as independent predictive indicators. Model calibration and discrimination were satisfactory. In the training dataset, the area under the curves (AUCs) for estimating the 28-, 56-, and 84-day all-cause mortality were 0.779, 0.780, and 0.787, respectively. The model exhibited excellent calibration and discrimination in the validation dataset, with AUC values of 0.791, 0.778, and 0.806 for estimating 28-, 56-, and 84-day all-cause mortality, respectively. The calibration curves exhibited the consistency of the model between the two cohorts. To visualize the results, we created a web-based calculator. Conclusion: We created a web-based calculator for assessing fatality risk in patients with AKI receiving CRRT, which may help rationalize clinical decision-making and personalized therapy.

Prognostic factors for renal function deterioration during palliative first-line chemotherapy for metastatic colorectal cancer: a retrospective study

PURPOSE

First-line choice of therapy is critical as it affects treatment decisions in later lines in patients with metastatic colorectal cancer (mCRC). We assessed changes in renal function for 1 year among patients diagnosed with mCRC who received first-line chemotherapy. We aimed to analyze the prognostic factors and effect of each chemotherapy regimen on the renal function of the patients.

METHODS

We retrospectively investigated patients with mCRC who were treated with a standard triplet regimen (FOLFOX/FOLFIRI with bevacizumab/cetuximab) in the first-line setting at Korea University Anam Hospital from 2015 to 2020. We checked renal function at 3-month intervals for 12 months. We calculated changes in eGFR (△eGFR, estimated glomerular filtration rate) and compared them with clinical factors such as age, sex, chronic disease, body mass index (BMI), disease status, baseline proteinuria, and first-line chemotherapy regimen.

RESULTS

Among 472 patients with mCRC, the median eGFR at baseline was 90.9 mL/min/1.73 m²; it was significantly lower (80.1 mL/min/1.73 m², p < 0.001) at 12 months after chemotherapy initiation. Particularly, the eGFR of patients treated with FOLFIRI + bevacizumab was 74.9 mL/min/1.73 m². The 1-year incidence rate of acute kidney injury (AKI) was 9.1%, with the lowest occurrence in patients receiving FOLFOX/cetuximab (2.1%) and the highest in those receiving FOLFIRI + bevacizumab (19.2%). Renal dysfunction was more frequent with FOLFIRI + bevacizumab as compared to the other regimens. Additionally, old age, low BMI, and proteinuria at baseline were also associated with a decreased eGFR.

CONCLUSIONS

These findings can serve as important factors when selecting the first-line chemotherapy regimen for patients with mCRC.

Development and Validation of a Personalized Model With Transfer Learning for Acute Kidney Injury Risk Estimation Using Electronic Health Records

IMPORTANCE

Acute kidney injury (AKI) is a heterogeneous syndrome prevalent among hospitalized patients. Personalized risk estimation and risk factor identification may allow effective intervention and improved outcomes.

OBJECTIVE

To develop and validate personalized AKI risk estimation models using electronic health records (EHRs), examine whether personalized models were beneficial in comparison with global and subgroup models, and assess the heterogeneity of risk factors and their outcomes in different subpopulations.

DESIGN, SETTING, AND PARTICIPANTS

This diagnostic study analyzed EHR data from 1 tertiary care hospital and used machine learning and logistic regression to develop and validate global, subgroup, and personalized risk estimation models. Transfer learning was implemented to enhance the personalized model. Predictor outcomes across subpopulations were analyzed, and metaregression was used to explore predictor interactions. Adults who were hospitalized for 2 or more days from November 1, 2007, to December 31, 2016, were included in the analysis. Patients with moderate or severe kidney dysfunction at admission were excluded. Data were analyzed between August 28, 2019, and May 8, 2022.

EXPOSURES

Clinical and laboratory variables in the EHR.

MAIN OUTCOMES AND MEASURES

The main outcome was AKI of any severity, and AKI was defined using the Kidney Disease: Improving Global Outcomes serum creatinine criteria. Performance of the models was measured with area under the receiver operating characteristic curve (AUROC), area under the precision-recall curve, and calibration.

RESULTS

The study cohort comprised 76 957 inpatient encounters. Patients had a mean (SD) age of 55.5 (17.4) years and included 42 159 men (54.8%). The personalized model with transfer learning outperformed the global model for AKI estimation in terms of AUROC among general inpatients (0.78 [95% CI, 0.77-0.79] vs 0.76 [95% CI, 0.75-0.76]; P < .001) and across the high-risk subgroups (0.79 [95% CI, 0.78-0.80] vs 0.75 [95% CI, 0.74-0.77]; P < .001) and low-risk subgroups (0.74 [95% CI, 0.73-0.75] vs 0.71 [95% CI, 0.70-0.72]; P < .001). The AUROC improvement reached 0.13 for the high-risk subgroups, such as those undergoing liver transplant and cardiac surgery. Moreover, the personalized model with transfer learning performed better than or comparably with the best published models in well-studied AKI subgroups. Predictor outcomes varied significantly between patients, and interaction analysis uncovered modifiers of the predictor outcomes.

CONCLUSIONS AND RELEVANCE

Results of this study demonstrated that a personalized modeling with transfer learning is an improved AKI risk estimation approach that can be used across diverse patient subgroups. Risk factor heterogeneity and interactions at the individual level highlighted the need for agile, personalized care.

Machine learning for the prediction of acute kidney injury in critical care patients with acute cerebrovascular disease

Purpose

Acute kidney injury (AKI) is a common complication and associated with a poor clinical outcome. In this study, we developed and validated a model for predicting the risk of AKI through machine learning methods in critical care patients with acute cerebrovascular disease.

Methods

This study was a retrospective study based on two different cohorts. Five machine learning methods were used to develop AKI risk prediction models. We used six popular metrics (AUROC, F2-Score, accuracy, sensitivity, specificity and precision) to evaluate the performance of these models.

Results

We identified 2935 patients in the MIMIC-III database and 499 patients in our local database to develop and validate the AKI risk prediction model. The incidence of AKI in these two different cohorts was 18.3% and 61.7%, respectively. Analysis showed that several laboratory parameters (serum creatinine, hemoglobin, white blood cell count, bicarbonate, blood urea nitrogen, sodium, albumin, and platelet count), age, and length of hospital stay, were the top ten important factors associated with AKI. The analysis demonstrated that the XGBoost had higher AUROC (0.880, 95%CI: 0.831–0.929), indicating that the XGBoost model was better at predicting AKI risk in patients with acute cerebrovascular disease than other models.

Conclusions

This study developed machine learning methods to identify critically ill patients with acute cerebrovascular disease who are at a high risk of developing AKI. This result suggested that machine learning techniques had the potential to improve the prediction of AKI risk models in critical care.

Management of acute kidney injury in gastrointestinal tumor: An overview

Gastrointestinal tumors remain a global health problem. Acute kidney injury (AKI) is a common complication during the treatment of gastrointestinal tumors. AKI can cause a decrease in the remission rate and an increase in mortality. In this review, we analyzed the causes and risk factors for AKI in gastrointestinal tumor patients. The possible mechanisms of AKI were divided into three groups: pretreatment, intrafraction and post-treatment causes. Treatment and prevention measures were proposed according to various factors to provide guidance to clinicians and oncologists that can reduce the incidence of AKI and improve the quality of life and survival rate of gastrointestinal tumor patients.

Development and Validation of a Prediction Model for Survival in Diabetic Patients With Acute Kidney Injury

OBJECTIVE

We aimed to analyze the risk factors affecting all-cause mortality in diabetic patients with acute kidney injury (AKI) and to develop and validate a nomogram for predicting the 90-day survival rate of patients.

METHODS

Clinical data of diabetic patients with AKI who were diagnosed at The First Affiliated Hospital of Guangxi Medical University from April 30, 2011, to April 30, 2021, were collected. A total of 1,042 patients were randomly divided into a development cohort and a validation cohort at a ratio of 7:3. The primary study endpoint was all-cause death within 90 days of AKI diagnosis. Clinical parameters and demographic characteristics were analyzed using Cox regression to develop a prediction model for survival in diabetic patients with AKI, and a nomogram was then constructed. The concordance index (C-index), receiver operating characteristic curve, and calibration plot were used to evaluate the prediction model.

RESULTS

The development cohort enrolled 730 patients with a median follow-up time of 87 (40-98) days, and 86 patients (11.8%) died during follow-up. The 90-day survival rate was 88.2% (644/730), and the recovery rate for renal function in survivors was 32.9% (212/644). Multivariate analysis showed that advanced age (HR = 1.064, 95% CI = 1.043-1.085), lower pulse pressure (HR = 0.964, 95% CI = 0.951-0.977), stage 3 AKI (HR = 4.803, 95% CI = 1.678-13.750), lower 25-hydroxyvitamin D3 (HR = 0.944, 95% CI = 0.930-0.960), and multiple organ dysfunction syndrome (HR = 2.056, 95% CI = 1.287-3.286) were independent risk factors affecting the all-cause death of diabetic patients with AKI (all p < 0.01). The C-indices of the prediction cohort and the validation cohort were 0.880 (95% CI = 0.839-0.921) and 0.798 (95% CI = 0.720-0.876), respectively. The calibration plot of the model showed excellent consistency between the prediction probability and the actual probability.

CONCLUSION

We developed a new prediction model that has been internally verified to have good discrimination, calibration, and clinical value for predicting the 90-day survival rate of diabetic patients with AKI.