Use of Machine Learning to Develop and Evaluate Models Using Preoperative and Intraoperative Data to Identify Risks of Postoperative Complications

Machine-learning-based models assist the prediction of pulmonary embolism in autoimmune diseases: A retrospective, multicenter study

BACKGROUND

Pulmonary embolism (PE) is a severe and acute cardiovascular syndrome with high mortality among patients with autoimmune inflammatory rheumatic diseases (AIIRDs). Accurate prediction and timely intervention play a pivotal role in enhancing survival rates. However, there is a notable scarcity of practical early prediction and risk assessment systems of PE in patients with AIIRD.

METHODS

In the training cohort, 60 AIIRD with PE cases and 180 age-, gender-, and disease-matched AIIRD non-PE cases were identified from 7254 AIIRD cases in Tongji Hospital from 2014 to 2022. Univariable logistic regression (LR) and least absolute shrinkage and selection operator (LASSO) were used to select the clinical features for further training with machine learning (ML) methods, including random forest (RF), support vector machines (SVM), neural network (NN), logistic regression (LR), gradient boosted decision tree (GBDT), classification and regression trees (CART), and C5.0 models. The performances of these models were subsequently validated using a multicenter validation cohort.

RESULTS

In the training cohort, 24 and 13 clinical features were selected by univariable LR and LASSO strategies, respectively. The five ML models (RF, SVM, NN, LR, and GBDT) showed promising performances, with an area under the receiver operating characteristic (ROC) curve (AUC) of 0.962-1.000 in the training cohort and 0.969-0.999 in the validation cohort. CART and C5.0 models achieved AUCs of 0.850 and 0.932, respectively, in the training cohort. Using D-dimer as a pre-screening index, the refined C5.0 model achieved an AUC exceeding 0.948 in the training cohort and an AUC above 0.925 in the validation cohort. These results markedly outperformed the use of D-dimer levels alone.

CONCLUSION

ML-based models are proven to be precise for predicting the onset of PE in patients with AIIRD exhibiting clinical suspicion of PE.

TRIAL REGISTRATION

Chictr.org.cn : ChiCTR2200059599.

Development and Validation of an Explainable Machine Learning Model for Predicting Myocardial Injury After Noncardiac Surgery in Two Centers in China: Retrospective Study

Background

Myocardial injury after noncardiac surgery (MINS) is an easily overlooked complication but closely related to postoperative cardiovascular adverse outcomes; therefore, the early diagnosis and prediction are particularly important.

Objective

We aimed to develop and validate an explainable machine learning (ML) model for predicting MINS among older patients undergoing noncardiac surgery.

Methods

The retrospective cohort study included older patients who had noncardiac surgery from 1 northern center and 1 southern center in China. The data sets from center 1 were divided into a training set and an internal validation set. The data set from center 2 was used as an external validation set. Before modeling, the least absolute shrinkage and selection operator and recursive feature elimination methods were used to reduce dimensions of data and select key features from all variables. Prediction models were developed based on the extracted features using several ML algorithms, including category boosting, random forest, logistic regression, naïve Bayes, light gradient boosting machine, extreme gradient boosting, support vector machine, and decision tree. Prediction performance was assessed by the area under the receiver operating characteristic (AUROC) curve as the main evaluation metric to select the best algorithms. The model performance was verified by internal and external validation data sets with the best algorithm and compared to the Revised Cardiac Risk Index. The Shapley Additive Explanations (SHAP) method was applied to calculate values for each feature, representing the contribution to the predicted risk of complication, and generate personalized explanations.

Results

A total of 19,463 eligible patients were included; among those, 12,464 patients in center 1 were included as the training set; 4754 patients in center 1 were included as the internal validation set; and 2245 in center 2 were included as the external validation set. The best-performing model for prediction was the CatBoost algorithm, achieving the highest AUROC of 0.805 (95% CI 0.778-0.831) in the training set, validating with an AUROC of 0.780 in the internal validation set and 0.70 in external validation set. Additionally, CatBoost demonstrated superior performance compared to the Revised Cardiac Risk Index (AUROC 0.636; P<.001). The SHAP values indicated the ranking of the level of importance of each variable, with preoperative serum creatinine concentration, red blood cell distribution width, and age accounting for the top three. The results from the SHAP method can predict events with positive values or nonevents with negative values, providing an explicit explanation of individualized risk predictions.

Conclusions

The ML models can provide a personalized and fairly accurate risk prediction of MINS, and the explainable perspective can help identify potentially modifiable sources of risk at the patient level.

Social vulnerability and surgery outcomes: a cross-sectional analysis

BACKGROUND

Post-operative complications present a challenge to the healthcare system due to the high unpredictability of their incidence. Socioeconomic conditions have been established as social determinants of health. However, their contribution relating to postoperative complications is still unclear as it can be heterogeneous based on community, type of surgical services, and sex and gender. Uncovering these relations can enable improved public health policy to reduce such complications.

METHODS

In this study, we conducted a large population cross-sectional analysis of social vulnerability and the odds of various post-surgical complications. We collected electronic health records data from over 50,000 surgeries that happened between 2012 and 2018 at a quaternary health center in St. Louis, Missouri, United States and the corresponding zip code of the patients. We built statistical logistic regression models of postsurgical complications with the social vulnerability index of the tract consisting of the zip codes of the patient as the independent variable along with sex and race interaction.

RESULTS

Our sample from the St. Louis area exhibited high variance in social vulnerability with notable rapid increase in vulnerability from the south west to the north of the Mississippi river indicating high levels of inequality. Our sample had more females than males, and females had slightly higher social vulnerability index. Postoperative complication incidence ranged from 0.75% to 41% with lower incidence rate among females. We found that social vulnerability was associated with abnormal heart rhythm with socioeconomic status and housing status being the main association factors. We also found associations of the interaction of social vulnerability and female sex with an increase in odds of heart attack and surgical wound infection. Those associations disappeared when controlling for general health and comorbidities.

CONCLUSIONS

Our results indicate that social vulnerability measures such as socioeconomic status and housing conditions could affect postsurgical outcomes through preoperative health. This suggests that the domains of preventive medicine and public health should place social vulnerability as a priority to achieve better health outcomes of surgical interventions.

Construction and Validation of a Predictive Model for Coronary Artery Disease Using Extreme Gradient Boosting

Purpose

Early recognition of coronary artery disease (CAD) could delay its progress and significantly reduce mortality. Sensitive, specific, cost-efficient and non-invasive indicators for assessing individual CAD risk in community population screening are urgently needed.

Patients and Methods

3112 patients with CAD and 3182 controls were recruited from three clinical centers in China, and differences in baseline and clinical characteristics were compared. For the discovery cohort, the least absolute shrinkage and selection operator (LASSO) regression was used to identify significant features and four machine learning algorithms (logistic regression, support vector machine (SVM), random forest (RF) and extreme gradient boosting (XGBoost)) were applied to construct models for CAD risk assessment, the receiver operating characteristics (ROC) curve and precision-recall (PR) curve were conducted to evaluate their predictive accuracy. The optimal model was interpreted by Shapley additive explanations (SHAP) analysis and assessed by the ROC curve, calibration curve, and decision curve analysis (DCA) and validated by two external cohorts.

Results

Using LASSO filtration, all included variables were considered to be statistically significant. Four machine learning models were constructed based on these features and the results of ROC and PR curve implied that the XGBoost model exhibited the highest predictive performance, which yielded a high area of ROC curve (AUC) of 0.988 (95% CI: 0.986-0.991) to distinguish CAD patients from controls with a sensitivity of 94.6% and a specificity of 94.6%. The calibration curve showed that the predicted results were in good agreement with actual observations, and DCA exhibited a better net benefit across a wide range of threshold probabilities. External validation of the model also exhibited favorable discriminatory performance, with an AUC, sensitivity, and specificity of 0.953 (95% CI: 0.945-0.960), 89.9%, and 87.1% in the validation cohort, and 0.935 (95% CI: 0.915-0.955), 82.0%, and 90.3% in the replication cohort.

Conclusion

Our model is highly informative for clinical practice and will be conducive to primary prevention and tailoring the precise management for CAD patients.

From the mathematical model to the patient: The scientific and human aspects of artificial intelligence in gastrointestinal surgery

Recent medical literature shows that the application of artificial intelligence (AI) models in gastrointestinal pathology is an exponentially growing field, with promising models that show very high performances. Regarding inflammatory bowel disease (IBD), recent reviews demonstrate promising diagnostic and prognostic AI models. However, studies are generally at high risk of bias (especially in AI models that are image-based). The creation of specific AI models that improve diagnostic performance and allow the establishment of a general prognostic forecast in IBD is of great interest, as it may allow the stratification of patients into subgroups and, in turn, allow the creation of different diagnostic and therapeutic protocols for these patients. Regarding surgical models, predictive models of postoperative complications have shown great potential in large-scale studies. In this work, the authors present the development of a predictive algorithm for early post-surgical complications in Crohn's disease based on a Random Forest model with exceptional predictive ability for complications within the cohort. The present work, based on logical and reasoned, clinical, and applicable aspects, lays a solid foundation for future prospective work to further develop post-surgical prognostic tools for IBD. The next step is to develop in a prospective and multicenter way, a collaborative path to optimize this line of research and make it applicable to our patients.

Machine learning with clinical and intraoperative biosignal data for predicting postoperative delirium after cardiac surgery

Early identification of patients at high risk of delirium is crucial for its prevention. Our study aimed to develop machine learning models to predict delirium after cardiac surgery using intraoperative biosignals and clinical data. We introduced a novel approach to extract relevant features from continuously measured intraoperative biosignals. These features reflect the patient's overall or baseline status, the extent of unfavorable conditions encountered intraoperatively, and beat-to-beat variability within the data. We developed a soft voting ensemble machine learning model using retrospective data from 1,912 patients. The model was then prospectively validated with data from 202 additional patients, achieving a high performance with an area under the receiver operating characteristic curve of 0.887 and an accuracy of 0.881. According to the SHapley Additive exPlanation method, several intraoperative biosignal features had high feature importance, suggesting that intraoperative patient management plays a crucial role in preventing delirium after cardiac surgery.

Evaluation of the clinical value of 10 estimating glomerular filtration rate equations and construction of a prediction model for kidney damage in adults from central China

Objectives

This study aimed to evaluate 10 estimating glomerular filtration rate (eGFR) equations in central China population and construct a diagnostic prediction model for assessing the kidney damage severity.

Methods

The concordance of 10 eGFR equations was investigated in healthy individuals from central China, and their clinical effectiveness in diagnosing kidney injury was evaluated. Subsequently, relevant clinical indicators were selected to develop a clinical prediction model for kidney damage.

Results

The overall concordance between CKD-EPIASR-Scr and CKD-EPI2021-Scr was the highest (weightedκ = 0.964) in healthy population. The CG formula, CKD-EPIASR-Scr and CKD-EPI2021-Scr performed better than others in terms of concordance with referenced GFR (rGFR), but had poor ability to distinguish between rGFR < 90 or < 60 mL/min·1.73 m2. This finding was basically consistent across subgroups. Finally, two logistic regression prediction models were constructed based on rGFR < 90 or 60 mL/min·1.73 m2. The area under the curve of receiver operating characteristic values of two prediction models were 0.811 vs 0.846 in training set and 0.812 vs 0.800 in testing set.

Conclusion

The concordance of CKD-EPIASR-Scr and CKD-EPI2021-Scr was the highest in the central China population. The Cockcroft-Gault formula, CKD-EPIASR-Scr, and CKD-EPI2021-Scr more accurately reflected true kidney function, while performed poorly in the staging diagnosis of CKD. The diagnostic prediction models showed the good clinical application performance in identifying mild or moderate kidney injury. These findings lay a solid foundation for future research on renal function assessment and predictive equations.

Machine learning based predictive modeling of readmissions following extracorporeal membrane oxygenation hospitalizations

Background

Despite increasing utilization and survival benefit over the last decade, extracorporeal membrane oxygenation (ECMO) remains resource-intensive with significant complications and rehospitalization risk. We thus utilized machine learning (ML) to develop prediction models for 90-day nonelective readmission following ECMO.

Methods

All adult patients receiving ECMO who survived index hospitalization were tabulated from the 2016-2020 Nationwide Readmissions Database. Extreme Gradient Boosting (XGBoost) models were developed to identify features associated with readmission following ECMO. Area under the receiver operating characteristic (AUROC), mean Average Precision (mAP), and the Brier score were calculated to estimate model performance relative to logistic regression (LR). Shapley Additive Explanation summary (SHAP) plots evaluated the relative impact of each factor on the model. An additional sensitivity analysis solely included patient comorbidities and indication for ECMO as potential model covariates.

Results

Of ∼22,947 patients, 4495 (19.6 %) were readmitted nonelectively within 90 days. The XGBoost model exhibited superior discrimination (AUROC 0.64 vs 0.49), classification accuracy (mAP 0.30 vs 0.20) and calibration (Brier score 0.154 vs 0.165, all P < 0.001) in predicting readmission compared to LR. SHAP plots identified duration of index hospitalization, undergoing heart/lung transplantation, and Medicare insurance to be associated with increased odds of readmission. Upon sub-analysis, XGBoost demonstrated superior disclination compared to LR (AUROC 0.61 vs 0.60, P < 0.05). Chronic liver disease and frailty were linked with increased odds of nonelective readmission.

Conclusions

ML outperformed LR in predicting readmission following ECMO. Future work is needed to identify other factors linked with readmission and further optimize post-ECMO care among this cohort.

Utilising intraoperative respiratory dynamic features for developing and validating an explainable machine learning model for postoperative pulmonary complications

BACKGROUND

Timely detection of modifiable risk factors for postoperative pulmonary complications (PPCs) could inform ventilation strategies that attenuate lung injury. We sought to develop, validate, and internally test machine learning models that use intraoperative respiratory features to predict PPCs.

METHODS

We analysed perioperative data from a cohort comprising patients aged 65 yr and older at an academic medical centre from 2019 to 2023. Two linear and four nonlinear learning models were developed and compared with the current gold-standard risk assessment tool ARISCAT (Assess Respiratory Risk in Surgical Patients in Catalonia Tool). The Shapley additive explanation of artificial intelligence was utilised to interpret feature importance and interactions.

RESULTS

Perioperative data were obtained from 10 284 patients who underwent 10 484 operations (mean age [range] 71 [65-98] yr; 42% female). An optimised XGBoost model that used preoperative variables and intraoperative respiratory variables had area under the receiver operating characteristic curves (AUROCs) of 0.878 (0.866-0.891) and 0.881 (0.879-0.883) in the validation and prospective cohorts, respectively. These models outperformed ARISCAT (AUROC: 0.496-0.533). The intraoperative dynamic features of respiratory dynamic system compliance, mechanical power, and driving pressure were identified as key modifiable contributors to PPCs. A simplified model based on XGBoost including 20 variables generated an AUROC of 0.864 (0.852-0.875) in an internal testing cohort. This has been developed into a web-based tool for further external validation (https://aorm.wchscu.cn/).

CONCLUSIONS

These findings suggest that real-time identification of surgical patients' risk of postoperative pulmonary complications could help personalise intraoperative ventilatory strategies and reduce postoperative pulmonary complications.

Development of a Probabilistic Boolean network (PBN) to model intraoperative blood pressure management

BACKGROUND

Blood pressure is a vital sign for organ perfusion that anesthesiologists measure and modulate during surgery. However, current decision-making processes rely heavily on clinicians' experience, which can lead to variability in treatment across surgeries. With the advent of machine learning, we can now create models to predict the outcomes of interventions and guide perioperative decision-making. The first step in this process involves translating the clinical decision-making process into a framework understood by an algorithm. Probabilistic Boolean networks (PBNs) provide an information-rich approach to this problem. A PBN trends toward a steady state, and its decisions are easily understood via its Boolean predictor functions. We hypothesize that a PBN can be developed that corrects hemodynamic instability in patients by selecting clinical interventions to maintain blood pressure within a given range.

METHODS

Data on patients over the age of 65 undergoing surgery with general anesthesia from 2018 to 2020 were drawn from the UF Health PRECEDE data set with IRB approval (IRB201700747). Parameters examined included heart rate, blood pressure, and frequency of medications given 15 min after anesthetic induction and 15 min before awakening. The medication frequency data were truncated into a 66/33 split for the training and validation set used in the PBN. The model was coded using Python 3 and evaluated by comparing the frequency of medications chosen by the program to the values in the testing set via linear regression analysis.

RESULTS

The network developed successfully models a hemodynamically unstable patient and corrects the imbalance by administering medications. This is evidenced by the model achieving a stable, steady state matrix in all iterations. However, the model's ability to emulate clinical drug selection was variable. It was successful with its use of vasodilator selection but struggled with the appropriate selection of vasopressors.

CONCLUSIONS

The PBN has demonstrated the ability to choose appropriate interventions based on a patient's current vitals. Additional work must be done to have the network emulate the frequency at which drugs are selected from in clinical practice. In its current state, the model provides an understanding of how a PBN behaves in the context of correcting hemodynamic instability and can aid in developing more robust models in the future.

A narrative review of the application of machine learning in venous thromboembolism

OBJECTIVE

To summarize the current research progress of machine learning and venous thromboembolism.

METHODS

The literature on risk factors, diagnosis, prevention and prognosis of machine learning and venous thromboembolism in recent years was reviewed.

RESULTS

Machine learning is the future of biomedical research, personalized medicine, and computer-aided diagnosis, and will significantly promote the development of biomedical research and healthcare. However, many medical professionals are not familiar with it. In this review, we will introduce several commonly used machine learning algorithms in medicine, discuss the application of machine learning in venous thromboembolism, and reveal the challenges and opportunities of machine learning in medicine.

CONCLUSION

The incidence of venous thromboembolism is high, the diagnostic measures are diverse, and it is necessary to classify and treat machine learning, and machine learning as a research tool, it is more necessary to strengthen the special research of venous thromboembolism and machine learning.

Effect of Machine Learning on Anaesthesiology Clinician Prediction of Postoperative Complications: The Perioperative ORACLE Randomised Clinical Trial

Background

Anaesthesiology clinicians can implement risk mitigation strategies if they know which patients are at greatest risk for postoperative complications. Although machine learning models predicting complications exist, their impact on clinician risk assessment is unknown.

Methods

This single-centre randomised clinical trial enrolled patients age ≥18 undergoing surgery with anaesthesiology services. Anaesthesiology clinicians providing remote intraoperative telemedicine support reviewed electronic health records with (assisted group) or without (unassisted group) also reviewing machine learning predictions. Clinicians predicted the likelihood of postoperative 30-day all-cause mortality and postoperative acute kidney injury within 7 days. Area under the receiver operating characteristic curve (AUROC) for the clinician predictions was determined.

Results

Among 5,071 patient cases reviewed by 89 clinicians, the observed incidence was 2% for postoperative death and 11% for acute kidney injury. Clinician predictions agreed with the models more strongly in the assisted versus unassisted group (weighted kappa 0.75 versus 0.62 for death [difference 0.13, 95%CI 0.10-0.17] and 0.79 versus 0.54 for kidney injury [difference 0.25, 95%CI 0.21-0.29]). Clinicians predicted death with AUROC of 0.793 in the assisted group and 0.780 in the unassisted group (difference 0.013, 95%CI -0.070 to 0.097). Clinicians predicted kidney injury with AUROC of 0.734 in the assisted group and 0.688 in the unassisted group (difference 0.046, 95%CI -0.003 to 0.091).

Conclusions

Although there was evidence that the models influenced clinician predictions, clinician performance was not statistically significantly different with and without machine learning assistance. Further work is needed to clarify the role of machine learning in real-time perioperative risk stratification.

Trial Registration

ClinicalTrials.gov NCT05042804.

Effect of Telemedicine Support for Intraoperative Anaesthesia Care on Postoperative Outcomes: The TECTONICS Randomised Clinical Trial

Background

Novel applications of telemedicine can improve care quality and patient outcomes. Telemedicine for intraoperative decision support has not been rigorously studied.

Methods

This single centre randomised clinical trial ( clinicaltrials.gov NCT03923699 ) of unselected adult surgical patients was conducted between July 1, 2019 and January 31, 2023. Patients received usual care or decision support from a telemedicine service, the Anesthesiology Control Tower (ACT). The ACT provided real-time recommendations to intraoperative anaesthesia clinicians based on case reviews, machine-learning forecasting, and physiologic alerts. ORs were randomised 1:1. Co-primary outcomes of 30-day all-cause mortality, respiratory failure, acute kidney injury (AKI), and delirium were analysed as intention-to-treat.

Results

The trial completed planned enrolment with 71927 surgeries (35956 ACT; 35971 usual care). After multiple testing correction, there was no significant effect of the ACT vs. usual care on 30-day mortality [641/35956 (1.8%) vs 638/35971 (1.8%), risk difference 0.0% (95% CI -0.2% to 0.3%), p=0.96], respiratory failure [1089/34613 (3.1%) vs 1112/34619 (3.2%), risk difference -0.1% (95% CI -0.4% to 0.3%), p=0.96], AKI [2357/33897 (7%) vs 2391/33795 (7.1%), risk difference -0.1% (-0.6% to 0.4%), p=0.96], or delirium [1283/3928 (32.7%) vs 1279/3989 (32.1%), risk difference 0.6% (-2.0% to 3.2%), p=0.96]. There were no significant differences in secondary outcomes or in sensitivity analyses.

Conclusions

In this large RCT of a novel application of telemedicine-based remote monitoring and decision support using real-time alerts and case reviews, we found no significant differences in postoperative outcomes. Large-scale intraoperative telemedicine is feasible, and we suggest future avenues where it may be impactful.

Machine Learning-Based Predictive Models for Patients with Venous Thromboembolism: A Systematic Review

BACKGROUND

 Venous thromboembolism (VTE) is a chronic disorder with a significant health and economic burden. Several VTE-specific clinical prediction models (CPMs) have been used to assist physicians in decision-making but have several limitations. This systematic review explores if machine learning (ML) can enhance CPMs by analyzing extensive patient data derived from electronic health records. We aimed to explore ML-CPMs' applications in VTE for risk stratification, outcome prediction, diagnosis, and treatment.

METHODS

 Three databases were searched: PubMed, Google Scholar, and IEEE electronic library. Inclusion criteria focused on studies using structured data, excluding non-English publications, studies on non-humans, and certain data types such as natural language processing and image processing. Studies involving pregnant women, cancer patients, and children were also excluded. After excluding irrelevant studies, a total of 77 studies were included.

RESULTS

 Most studies report that ML-CPMs outperformed traditional CPMs in terms of receiver operating area under the curve in the four clinical domains that were explored. However, the majority of the studies were retrospective, monocentric, and lacked detailed model architecture description and external validation, which are essential for quality audit. This review identified research gaps and highlighted challenges related to standardized reporting, reproducibility, and model comparison.

CONCLUSION

 ML-CPMs show promise in improving risk assessment and individualized treatment recommendations in VTE. Apparently, there is an urgent need for standardized reporting and methodology for ML models, external validation, prospective and real-world data studies, as well as interventional studies to evaluate the impact of artificial intelligence in VTE.

Development of interpretable machine learning models for prediction of acute kidney injury after noncardiac surgery: a retrospective cohort study

BACKGROUND

Early identification of patients at high-risk of postoperative acute kidney injury (AKI) can facilitate the development of preventive approaches. This study aimed to develop prediction models for postoperative AKI in noncardiac surgery using machine learning algorithms. The authors also evaluated the predictive performance of models that included only preoperative variables or only important predictors.

MATERIALS AND METHODS

Adult patients undergoing noncardiac surgery were retrospectively included in the study (76 457 patients in the discovery cohort and 11 910 patients in the validation cohort). AKI was determined using the KDIGO criteria. The prediction model was developed using 87 variables (56 preoperative variables and 31 intraoperative variables). A variety of machine learning algorithms were employed to develop the model, including logistic regression, random forest, extreme gradient boosting, and gradient boosting decision trees. The performance of different models was compared using the area under the receiver operating characteristic curve (AUROC). Shapley Additive Explanations (SHAP) analysis was employed for model interpretation.

RESULTS

The patients in the discovery cohort had a median age of 52 years (IQR: 42-61 years), and 1179 patients (1.5%) developed AKI after surgery. The gradient boosting decision trees algorithm showed the best predictive performance using all available variables, or only preoperative variables. The AUROCs were 0.849 (95% CI: 0.835-0.863) and 0.828 (95% CI: 0.813-0.843), respectively. The SHAP analysis showed that age, surgical duration, preoperative serum creatinine, and gamma-glutamyltransferase, as well as American Society of Anesthesiologists physical status III were the most important five features. When gradually reducing the features, the AUROCs decreased from 0.852 (including the top 40 features) to 0.839 (including the top 10 features). In the validation cohort, the authors observed a similar pattern regarding the models' predictive performance.

CONCLUSIONS

The machine learning models the authors developed had satisfactory predictive performance for identifying high-risk postoperative AKI patients. Furthermore, the authors found that model performance was only slightly affected when only preoperative variables or only the most important predictive features were included.

Multiparametric MRI model to predict molecular subtypes of breast cancer using Shapley additive explanations interpretability analysis

PURPOSE

The purpose of this study was to assess the predictive performance of multiparametric magnetic resonance imaging (MRI) for molecular subtypes and interpret features using SHapley Additive exPlanations (SHAP) analysis.

MATERIAL AND METHODS

Patients with breast cancer who underwent pre-treatment MRI (including ultrafast dynamic contrast-enhanced MRI, magnetic resonance spectroscopy, diffusion kurtosis imaging and intravoxel incoherent motion) were recruited between February 2019 and January 2022. Thirteen semantic and thirteen multiparametric features were collected and the key features were selected to develop machine-learning models for predicting molecular subtypes of breast cancers (luminal A, luminal B, triple-negative and HER2-enriched) by using stepwise logistic regression. Semantic model and multiparametric model were built and compared based on five machine-learning classifiers. Model decision-making was interpreted using SHAP analysis.

RESULTS

A total of 188 women (mean age, 53 ± 11 [standard deviation] years; age range: 25-75 years) were enrolled and further divided into training cohort (131 women) and validation cohort (57 women). XGBoost demonstrated good predictive performance among five machine-learning classifiers. Within the validation cohort, the areas under the receiver operating characteristic curves (AUCs) for the semantic models ranged from 0.693 (95% confidence interval [CI]: 0.478-0.839) for HER2-enriched subtype to 0.764 (95% CI: 0.681-0.908) for luminal A subtype, inferior to multiparametric models that yielded AUCs ranging from 0.771 (95% CI: 0.630-0.888) for HER2-enriched subtype to 0.857 (95% CI: 0.717-0.957) for triple-negative subtype. The AUCs between the semantic and the multiparametric models did not show significant differences (P range: 0.217-0.640). SHAP analysis revealed that lower iAUC, higher kurtosis, lower D*, and lower kurtosis were distinctive features for luminal A, luminal B, triple-negative breast cancer, and HER2-enriched subtypes, respectively.

CONCLUSION

Multiparametric MRI is superior to semantic models to effectively predict the molecular subtypes of breast cancer.

Experts vs. machine - comparison of machine learning to expert-informed prediction of outcome after major liver surgery

BACKGROUND

Machine learning (ML) has been successfully implemented for classification tasks (e.g., cancer diagnosis). ML performance for more challenging predictions is largely unexplored. This study's objective was to compare machine learning vs. expert-informed predictions for surgical outcome in patients undergoing major liver surgery.

METHODS

Single tertiary center data on preoperative parameters and postoperative complications for elective hepatic surgery patients were included (2008-2021). Expert-informed prediction models were established on 14 parameters identified by two expert liver surgeons to impact on postoperative outcome. ML models used all available preoperative patient variables (n = 62). Model performance was compared for predicting 3-month postoperative overall morbidity. Temporal validation and additional analysis in major liver resection patients were conducted.

RESULTS

889 patients included. Expert-informed models showed low average bias (2-5 CCI points) with high over/underprediction. ML models performed similarly: average prediction 5-10 points higher than observed CCI values with high variability (95% CI -30 to 50). No performance improvement for major liver surgery patients.

CONCLUSION

No clinical relevance in the application of ML for predicting postoperative overall morbidity was found. Despite being a novel hype, ML has the potential for application in clinical practice. However, at this stage it does not replace established approaches of prediction modelling.

Developing and validating machine learning-based prediction models for frailty occurrence in those with chronic obstructive pulmonary disease

Background

Frailty is a medical syndrome caused by multiple factors, characterized by decreased strength, endurance, and diminished physiological function, resulting in increased susceptibility to dependence and/or death. Patients with chronic obstructive pulmonary disease (COPD) tend to be more vulnerable to frailty due to their physical and psychological burdens. Therefore, the aim of this study was to develop a reliable and accurate vulnerability risk prediction model for frailty in patients with COPD in order to improve the identification and prediction of patient frailty. The specific objectives of this study were to determine the prevalence of frailty in patients with COPD and develop a prediction model and evaluate its predictive power.

Methods

Clinical information was analyzed using data from the 2018 China Health and Retirement Longitudinal Study (CHARLS) database, and 34 indicators, including behavioral factors, health status, mental health parameters, and various sociodemographic variables, were examined in the study. The adaptive synthetic sampling technique was used for unbalanced data. Three methods, ridge regressor, extreme gradient boosting (XGBoost) classifier, and random forest (RF) regressor, were used to filter predictors. Seven machine learning (ML) techniques including logistic regression (LR), support vector machines (SVM), multilayer perceptron, light gradient-boosting machine, XGBoost, RF, and K-nearest neighbors were used to analyze and determine the optimal model. For customized risk assessment, an online predictive risk modeling website was created, along with Shapley additive explanation (SHAP) interpretations.

Results

Depression, smoking, gender, social activities, dyslipidemia, asthma, and residence type (urban vs. rural) were predictors for the development of frailty in patients with COPD. In the test set, the XGBoost model had an area under the curve of 0.942 (95% confidence interval: 0.925-0.959), an accuracy of 0.915, a sensitivity of 0.873, and a specificity of 0.911, indicating that it was the best model.

Conclusions

The ML predictive model developed in this study is a useful and easy-to-use instrument for assessing the vulnerability risk of patients with COPD and may aid clinical physicians in screening high-risk patients.

Predicting postoperative rehemorrhage in hypertensive intracerebral hemorrhage using noncontrast CT radiomics and clinical data with an interpretable machine learning approach

In hypertensive intracerebral hemorrhage (HICH) patients, while emergency surgeries effectively reduce intracranial pressure and hematoma volume, their significant risk of causing postoperative rehemorrhage necessitates early detection and management to improve patient prognosis. This study sought to develop and validate machine learning (ML) models leveraging clinical data and noncontrast CT radiomics to pinpoint patients at risk of postoperative rehemorrhage, equipping clinicians with an early detection tool for prompt intervention. The study conducted a retrospective analysis on 609 HICH patients, dividing them into training and external verification cohorts. These patients were categorized into groups with and without postoperative rehemorrhage. Radiomics features from noncontrast CT images were extracted, standardized, and employed to create several ML models. These models underwent internal validation using both radiomics and clinical data, with the best model's feature significance assessed via the Shapley additive explanations (SHAP) method, then externally validated. In the study of 609 patients, postoperative rehemorrhage rates were similar in the training (18.8%, 80/426) and external verification (17.5%, 32/183) cohorts. Six significant noncontrast CT radiomics features were identified, with the support vector machine (SVM) model outperforming others in both internal and external validations. SHAP analysis highlighted five critical predictors of postoperative rehemorrhage risk, encompassing three radiomics features from noncontrast CT and two clinical data indicators. This study highlights the effectiveness of an SVM model combining radiomics features from noncontrast CT and clinical parameters in predicting postoperative rehemorrhage among HICH patients. This approach enables timely and effective interventions, thereby improving patient outcomes.

Development and performance assessment of novel machine learning models for predicting postoperative pneumonia in aneurysmal subarachnoid hemorrhage patients: external validation in MIMIC-IV

Background

Postoperative pneumonia (POP) is one of the primary complications after aneurysmal subarachnoid hemorrhage (aSAH) and is associated with postoperative mortality, extended hospital stay, and increased medical fee. Early identification of pneumonia and more aggressive treatment can improve patient outcomes. We aimed to develop a model to predict POP in aSAH patients using machine learning (ML) methods.

Methods

This internal cohort study included 706 patients with aSAH undergoing intracranial aneurysm embolization or aneurysm clipping. The cohort was randomly split into a train set (80%) and a testing set (20%). Perioperative information was collected from participants to establish 6 machine learning models for predicting POP after surgical treatment. The area under the receiver operating characteristic curve (AUC), precision-recall curve were used to assess the accuracy, discriminative power, and clinical validity of the predictions. The final model was validated using an external validation set of 97 samples from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database.

Results

In this study, 15.01% of patients in the training set and 12.06% in the testing set with POP after underwent surgery. Multivariate logistic regression analysis showed that mechanical ventilation time (MVT), Glasgow Coma Scale (GCS), Smoking history, albumin level, neutrophil-to-albumin Ratio (NAR), c-reactive protein (CRP)-to-albumin ratio (CAR) were independent predictors of POP. The logistic regression (LR) model presented significantly better predictive performance (AUC: 0.91) than other models and also performed well in the external validation set (AUC: 0.89).

Conclusion

A machine learning model for predicting POP in aSAH patients was successfully developed using a machine learning algorithm based on six perioperative variables, which could guide high-risk POP patients to take appropriate preventive measures.

User-Centered Design of a Machine Learning Dashboard for Prediction of Postoperative Complications

BACKGROUND

Machine learning models can help anesthesiology clinicians assess patients and make clinical and operational decisions, but well-designed human-computer interfaces are necessary for machine learning model predictions to result in clinician actions that help patients. Therefore, the goal of this study was to apply a user-centered design framework to create a user interface for displaying machine learning model predictions of postoperative complications to anesthesiology clinicians.

METHODS

Twenty-five anesthesiology clinicians (attending anesthesiologists, resident physicians, and certified registered nurse anesthetists) participated in a 3-phase study that included (phase 1) semistructured focus group interviews and a card sorting activity to characterize user workflows and needs; (phase 2) simulated patient evaluation incorporating a low-fidelity static prototype display interface followed by a semistructured interview; and (phase 3) simulated patient evaluation with concurrent think-aloud incorporating a high-fidelity prototype display interface in the electronic health record. In each phase, data analysis included open coding of session transcripts and thematic analysis.

RESULTS

During the needs assessment phase (phase 1), participants voiced that (a) identifying preventable risk related to modifiable risk factors is more important than nonpreventable risk, (b) comprehensive patient evaluation follows a systematic approach that relies heavily on the electronic health record, and (c) an easy-to-use display interface should have a simple layout that uses color and graphs to minimize time and energy spent reading it. When performing simulations using the low-fidelity prototype (phase 2), participants reported that (a) the machine learning predictions helped them to evaluate patient risk, (b) additional information about how to act on the risk estimate would be useful, and (c) correctable problems related to textual content existed. When performing simulations using the high-fidelity prototype (phase 3), usability problems predominantly related to the presentation of information and functionality. Despite the usability problems, participants rated the system highly on the System Usability Scale (mean score, 82.5; standard deviation, 10.5).

CONCLUSIONS

Incorporating user needs and preferences into the design of a machine learning dashboard results in a display interface that clinicians rate as highly usable. Because the system demonstrates usability, evaluation of the effects of implementation on both process and clinical outcomes is warranted.

Preoperative Prediction of Postoperative Infections Using Machine Learning and Electronic Health Record Data

OBJECTIVE

To estimate preoperative risk of postoperative infections using structured electronic health record (EHR) data.

BACKGROUND

Surveillance and reporting of postoperative infections is primarily done through costly, labor-intensive manual chart reviews on a small sample of patients. Automated methods using statistical models applied to postoperative EHR data have shown promise to augment manual review as they can cover all operations in a timely manner. However, there are no specific models for risk-adjusting infectious complication rates using EHR data.

METHODS

Preoperative EHR data from 30,639 patients (2013-2019) were linked to the American College of Surgeons National Surgical Quality Improvement Program preoperative data and postoperative infection outcomes data from 5 hospitals in the University of Colorado Health System. EHR data included diagnoses, procedures, operative variables, patient characteristics, and medications. Lasso and the knockoff filter were used to perform controlled variable selection. Outcomes included surgical site infection, urinary tract infection, sepsis/septic shock, and pneumonia up to 30 days postoperatively.

RESULTS

Among >15,000 candidate predictors, 7 were chosen for the surgical site infection model and 6 for each of the urinary tract infection, sepsis, and pneumonia models. Important variables included preoperative presence of the specific outcome, wound classification, comorbidities, and American Society of Anesthesiologists physical status classification. The area under the receiver operating characteristic curve for each model ranged from 0.73 to 0.89.

CONCLUSIONS

Parsimonious preoperative models for predicting postoperative infection risk using EHR data were developed and showed comparable performance to existing American College of Surgeons National Surgical Quality Improvement Program risk models that use manual chart review. These models can be used to estimate risk-adjusted postoperative infection rates applied to large volumes of EHR data in a timely manner.

Artificial Intelligence: From Buzzword to Useful Tool in Clinical Pharmacology

The advent of artificial intelligence (AI) in clinical pharmacology and drug development is akin to the dawning of a new era. Previously dismissed as merely technological hype, these approaches have emerged as promising tools in different domains, including health care, demonstrating their potential to empower clinical pharmacology decision making, revolutionize the drug development landscape, and advance patient care. Although challenges remain, the remarkable progress already made signals that the leap from hype to reality is well underway, and AI promises to offer clinical pharmacology new tools and possibilities for optimizing patient care is gradually coming to fruition. This review dives into the burgeoning world of AI and machine learning (ML), showcasing different applications of AI in clinical pharmacology and the impact of successful AI/ML implementation on drug development and/or regulatory decisions. This review also highlights recommendations for areas of opportunity in clinical pharmacology, including data analysis (e.g., handling large data sets, screening to identify important covariates, and optimizing patient population) and efficiencies (e.g., automation, translation, literature curation, and training). Realizing the benefits of AI in drug development and understanding its value will lead to the successful integration of AI tools in our clinical pharmacology and pharmacometrics armamentarium.

Predicting short-term major postoperative complications in intestinal resection for Crohn's disease: A machine learning-based study

BACKGROUND

Due to the complexity and numerous comorbidities associated with Crohn's disease (CD), the incidence of postoperative complications is high, significantly impacting the recovery and prognosis of patients. Consequently, additional studies are required to precisely predict short-term major complications following intestinal resection (IR), aiding surgical decision-making and optimizing patient care.

AIM

To construct novel models based on machine learning (ML) to predict short-term major postoperative complications in patients with CD following IR.

METHODS

A retrospective analysis was performed on clinical data derived from a patient cohort that underwent IR for CD from January 2017 to December 2022. The study participants were randomly allocated to either a training cohort or a validation cohort. The logistic regression and random forest (RF) were applied to construct models in the training cohort, with model discrimination evaluated using the area under the curves (AUC). The validation cohort assessed the performance of the constructed models.

RESULTS

Out of the 259 patients encompassed in the study, 5.0% encountered major postoperative complications (Clavien-Dindo ≥ III) within 30 d following IR for CD. The AUC for the logistic model was 0.916, significantly lower than the AUC of 0.965 for the RF model. The logistic model incorporated a preoperative CD activity index (CDAI) of ≥ 220, a diminished preoperative serum albumin level, conversion to laparotomy surgery, and an extended operation time. A nomogram for the logistic model was plotted. Except for the surgical approach, the other three variables ranked among the top four important variables in the novel ML model.

CONCLUSION

Both the nomogram and RF exhibited good performance in predicting short-term major postoperative complications in patients with CD, with the RF model showing more superiority. A preoperative CDAI of ≥ 220, a diminished preoperative serum albumin level, and an extended operation time might be the most crucial variables. The findings of this study can assist clinicians in identifying patients at a higher risk for complications and offering personalized perioperative management to enhance patient outcomes.

Anesthesia decision analysis using a cloud-based big data platform

Big data technologies have proliferated since the dawn of the cloud-computing era. Traditional data storage, extraction, transformation, and analysis technologies have thus become unsuitable for the large volume, diversity, high processing speed, and low value density of big data in medical strategies, which require the development of novel big data application technologies. In this regard, we investigated the most recent big data platform breakthroughs in anesthesiology and designed an anesthesia decision model based on a cloud system for storing and analyzing massive amounts of data from anesthetic records. The presented Anesthesia Decision Analysis Platform performs distributed computing on medical records via several programming tools, and provides services such as keyword search, data filtering, and basic statistics to reduce inaccurate and subjective judgments by decision-makers. Importantly, it can potentially to improve anesthetic strategy and create individualized anesthesia decisions, lowering the likelihood of perioperative complications.

Development and validation of 'Patient Optimizer' (POP) algorithms for predicting surgical risk with machine learning

BACKGROUND

Pre-operative risk assessment can help clinicians prepare patients for surgery, reducing the risk of perioperative complications, length of hospital stay, readmission and mortality. Further, it can facilitate collaborative decision-making and operational planning.

OBJECTIVE

To develop effective pre-operative risk assessment algorithms (referred to as Patient Optimizer or POP) using Machine Learning (ML) that predict the development of post-operative complications and provide pilot data to inform the design of a larger prospective study.

METHODS

After institutional ethics approval, we developed a base model that encapsulates the standard manual approach of combining patient-risk and procedure-risk. In an automated process, additional variables were included and tested with 10-fold cross-validation, and the best performing features were selected. The models were evaluated and confidence intervals calculated using bootstrapping. Clinical expertise was used to restrict the cardinality of categorical variables (e.g. pathology results) by including the most clinically relevant values. The models were created with logistic regression (LR) and extreme gradient-boosted trees using XGBoost (Chen and Guestrin, 2016). We evaluated performance using the area under the receiver operating characteristic curve (AUROC) and the area under the precision-recall curve (AUPRC). Data was obtained from a metropolitan university teaching hospital from January 2015 to July 2020. Data collection was restricted to adult patients undergoing elective surgery.

RESULTS

A total of 11,475 adult admissions were included. The performance of XGBoost and LR was very similar across endpoints and metrics. For predicting the risk of any post-operative complication, kidney failure and length-of-stay (LOS), POP with XGBoost achieved an AUROC (95%CI) of 0.755 (0.744, 0.767), 0.869 (0.846, 0.891) and 0.841 (0.833, 0.847) respectively and AUPRC of 0.651 (0.632, 0.669), 0.336 (0.282, 0.390) and 0.741 (0.729, 0.753) respectively. For 30-day readmission and in-patient mortality, POP with XGBoost achieved an AUROC (95%CI) of 0.610 (0.587, 0.635) and 0.866 (0.777, 0.943) respectively and AUPRC of 0.116 (0.104, 0.132) and 0.031 (0.015, 0.072) respectively.

CONCLUSION

The POP algorithms effectively predicted any post-operative complication, kidney failure and LOS in the sample population. A larger study is justified to improve the algorithm to better predict complications and length of hospital stay. A larger dataset may also improve the prediction of additional specific complications, readmission and mortality.

Gut microbiome features and metabolites in non-alcoholic fatty liver disease among community-dwelling middle-aged and older adults

BACKGROUND

The specific microbiota and associated metabolites linked to non-alcoholic fatty liver disease (NAFLD) are still controversial. Thus, we aimed to understand how the core gut microbiota and metabolites impact NAFLD.

METHODS

The data for the discovery cohort were collected from the Guangzhou Nutrition and Health Study (GNHS) follow-up conducted between 2014 and 2018. We collected 272 metadata points from 1546 individuals. The metadata were input into four interpretable machine learning models to identify important gut microbiota associated with NAFLD. These models were subsequently applied to two validation cohorts [the internal validation cohort (n = 377), and the prospective validation cohort (n = 749)] to assess generalizability. We constructed an individual microbiome risk score (MRS) based on the identified gut microbiota and conducted animal faecal microbiome transplantation experiment using faecal samples from individuals with different levels of MRS to determine the relationship between MRS and NAFLD. Additionally, we conducted targeted metabolomic sequencing of faecal samples to analyse potential metabolites.

RESULTS

Among the four machine learning models used, the lightGBM algorithm achieved the best performance. A total of 12 taxa-related features of the microbiota were selected by the lightGBM algorithm and further used to calculate the MRS. Increased MRS was positively associated with the presence of NAFLD, with odds ratio (OR) of 1.86 (1.72, 2.02) per 1-unit increase in MRS. An elevated abundance of the faecal microbiota (f__veillonellaceae) was associated with increased NAFLD risk, whereas f__rikenellaceae, f__barnesiellaceae, and s__adolescentis were associated with a decreased presence of NAFLD. Higher levels of specific gut microbiota-derived metabolites of bile acids (taurocholic acid) might be positively associated with both a higher MRS and NAFLD risk. FMT in mice further confirmed a causal association between a higher MRS and the development of NAFLD.

CONCLUSIONS

We confirmed that an alteration in the composition of the core gut microbiota might be biologically relevant to NAFLD development. Our work demonstrated the role of the microbiota in the development of NAFLD.

Prediction of tumor lysis syndrome in childhood acute lymphoblastic leukemia based on machine learning models: a retrospective study

Background

Tumor lysis syndrome (TLS) often occurs early after induction chemotherapy for acute lymphoblastic leukemia (ALL) and can rapidly progress. This study aimed to construct a machine learning model to predict the risk of TLS using clinical indicators at the time of ALL diagnosis.

Methods

This observational cohort study was conducted at the National Clinical Research Center for Child Health and Disease. Data were collected from pediatric ALL patients diagnosed between December 2008 and December 2021. Four machine learning models were constructed using the Least Absolute Shrinkage and Selection Operator (LASSO) to select key clinical indicators for model construction.

Results

The study included 2,243 pediatric ALL patients, and the occurrence of TLS was 8.87%. A total of 33 indicators with missing values ≤30% were collected, and 12 risk factors were selected through LASSO regression analysis. The CatBoost model with the best performance after feature screening was selected to predict the TLS of ALL patients. The CatBoost model had an AUC of 0.832 and an accuracy of 0.758. The risk factors most associated with TLS were the absence of potassium, phosphorus, aspartate transaminase (AST), white blood cell count (WBC), and urea levels.

Conclusion

We developed the first TLS prediction model for pediatric ALL to assist clinicians in risk stratification at diagnosis and in developing personalized treatment protocols. This study is registered on the China Clinical Trials Registry platform (ChiCTR2200060616).

Clinical trial registration

https://www.chictr.org.cn/, identifier ChiCTR2200060616.

Artificial Intelligence in Operating Room Management

This systematic review examines the recent use of artificial intelligence, particularly machine learning, in the management of operating rooms. A total of 22 selected studies from February 2019 to September 2023 are analyzed. The review emphasizes the significant impact of AI on predicting surgical case durations, optimizing post-anesthesia care unit resource allocation, and detecting surgical case cancellations. Machine learning algorithms such as XGBoost, random forest, and neural networks have demonstrated their effectiveness in improving prediction accuracy and resource utilization. However, challenges such as data access and privacy concerns are acknowledged. The review highlights the evolving nature of artificial intelligence in perioperative medicine research and the need for continued innovation to harness artificial intelligence's transformative potential for healthcare administrators, practitioners, and patients. Ultimately, artificial intelligence integration in operative room management promises to enhance healthcare efficiency and patient outcomes.

Postoperative delirium prediction after cardiac surgery using machine learning models

OBJECTIVE

Postoperative delirium (POD) is a common postoperative complication in elderly patients, especially those undergoing cardiac surgery, which seriously affects the short- and long-term prognosis of patients. Early identification of risk factors for the development of POD can help improve the perioperative management of surgical patients. In the present study, five machine learning models were developed to predict patients at high risk of delirium after cardiac surgery and their performance was compared.

METHODS

A total of 367 patients who underwent cardiac surgery were retrospectively included in this study. Using single-factor analysis, 21 risk factors for POD were selected for inclusion in machine learning. The dataset was divided using 10-fold cross-validation for model training and testing. Five machine learning models (random forest (RF), support vector machine (SVM), radial based kernel neural network (RBFNN), K-nearest neighbour (KNN), and Kernel ridge regression (KRR)) were compared using area under the receiver operating characteristic curve (AUC-ROC), accuracy (ACC), sensitivity (SN), specificity (SPE), and Matthews coefficient (MCC).

RESULTS

Among 367 patients, 105 patients developed POD, the incidence of delirium was 28.6 %. Among the five ML models, RF had the best performance in ACC (87.99 %), SN (69.27 %), SPE (95.38 %), MCC (70.00 %) and AUC (0.9202), which was far superior to the other four models.

CONCLUSION

Delirium is common in patients after cardiac surgery. This analysis confirms the importance of the computational ML models in predicting the occurrence of delirium after cardiac surgery, especially the outstanding performance of the RF model, which has practical clinical applications for early identification of patients at risk of developing POD.

Remote Monitoring and Artificial Intelligence: Outlook for 2050

Remote monitoring and artificial intelligence will become common and intertwined in anesthesiology by 2050. In the intraoperative period, technology will lead to the development of integrated monitoring systems that will integrate multiple data streams and allow anesthesiologists to track patients more effectively. This will free up anesthesiologists to focus on more complex tasks, such as managing risk and making value-based decisions. This will also enable the continued integration of remote monitoring and control towers having profound effects on coverage and practice models. In the PACU and ICU, the technology will lead to the development of early warning systems that can identify patients who are at risk of complications, enabling early interventions and more proactive care. The integration of augmented reality will allow for better integration of diverse types of data and better decision-making. Postoperatively, the proliferation of wearable devices that can monitor patient vital signs and track their progress will allow patients to be discharged from the hospital sooner and receive care at home. This will require increased use of telemedicine, which will allow patients to consult with doctors remotely. All of these advances will require changes to legal and regulatory frameworks that will enable new workflows that are different from those familiar to today's providers.

Prediction models for postoperative delirium in elderly patients with machine-learning algorithms and SHapley Additive exPlanations

Postoperative delirium (POD) is a common and severe complication in elderly patients with hip fractures. Identifying high-risk patients with POD can help improve the outcome of patients with hip fractures. We conducted a retrospective study on elderly patients (≥65 years of age) who underwent orthopedic surgery with hip fracture between January 2014 and August 2019. Conventional logistic regression and five machine-learning algorithms were used to construct prediction models of POD. A nomogram for POD prediction was built with the logistic regression method. The area under the receiver operating characteristic curve (AUC-ROC), accuracy, sensitivity, and precision were calculated to evaluate different models. Feature importance of individuals was interpreted using Shapley Additive Explanations (SHAP). About 797 patients were enrolled in the study, with the incidence of POD at 9.28% (74/797). The age, renal insufficiency, chronic obstructive pulmonary disease (COPD), use of antipsychotics, lactate dehydrogenase (LDH), and C-reactive protein are used to build a nomogram for POD with an AUC of 0.71. The AUCs of five machine-learning models are 0.81 (Random Forest), 0.80 (GBM), 0.68 (AdaBoost), 0.77 (XGBoost), and 0.70 (SVM). The sensitivities of the six models range from 68.8% (logistic regression and SVM) to 91.9% (Random Forest). The precisions of the six machine-learning models range from 18.3% (logistic regression) to 67.8% (SVM). Six prediction models of POD in patients with hip fractures were constructed using logistic regression and five machine-learning algorithms. The application of machine-learning algorithms could provide convenient POD risk stratification to benefit elderly hip fracture patients.

Machine learning prediction model of major adverse outcomes after pediatric congenital heart surgery-a retrospective cohort study

BACKGROUND

Major adverse postoperative outcomes (APOs) can greatly affect mortality, hospital stay, care management and planning, and quality of life. This study aimed to evaluate the performance of five machine learning (ML) algorithms for predicting four major APOs after pediatric congenital heart surgery and their clinically meaningful model interpretations.

METHODS

Between August 2014 and December 2021, 23,000 consecutive pediatric patients receiving congenital heart surgery were enrolled. Based on the split date of 1 January 2019, we selected 13,927 participants for the training cohort, and 9,073 participants for the testing cohort. Four predefined major APOs including low cardiac output syndrome (LCOS), pneumonia, renal failure, and deep venous thrombosis (DVT) were investigated. 39 clinical and laboratory features were inputted in five ML models: light gradient boosting machine (LightGBM), logistic regression (LR), support vector machine, random forest, and CatBoost. The performance and interpretations of ML models were evaluated using the area under the receiver operating characteristic curve (AUC) and Shapley Additive Explanations (SHAP).

RESULTS

In the training cohort, CatBoost algorithms outperformed others with the mean AUCs of 0.908 for LCOS and 0.957 for renal failure, while LightGBM and LR achieved the best mean AUCs of 0.886 for pneumonia and 0.942 for DVT, respectively. In the testing cohort, the best-performing ML model for each major APOs with the following mean AUCs: LCOS (LightGBM), 0.893 (95% confidence interval (CI), 0.884-0.895); pneumonia (LR), 0.929 (95% CI, 0.926-0.931); renal failure (LightGBM), 0.963 (95% CI, 0.947-0.979), and DVT (LightGBM), 0.970 (95% CI, 0.953-0.982). The performance of ML models using only clinical variables was slightly lower than those using combined data, with the mean AUCs of 0.873 for LCOS, 0.894 for pneumonia, 0.953 for renal failure, and 0.933 for DVT. The SHAP showed that mechanical ventilation time was the most important contributor of four major APOs.

CONCLUSIONS

In pediatric congenital heart surgery, the established ML model can accurately predict the risk of four major APOs, providing reliable interpretations for high-risk contributor identification and informed clinical decisions making.

Prediction of postoperative infectious complications in elderly patients with colorectal cancer: a study based on improved machine learning

BACKGROUND

Infectious complications after colorectal cancer (CRC) surgery increase perioperative mortality and are significantly associated with poor prognosis. We aimed to develop a model for predicting infectious complications after colorectal cancer surgery in elderly patients based on improved machine learning (ML) using inflammatory and nutritional indicators.

METHODS

The data of 512 elderly patients with colorectal cancer in the Third Affiliated Hospital of Anhui Medical University from March 2018 to April 2022 were retrospectively collected and randomly divided into a training set and validation set. The optimal cutoff values of NLR (3.80), PLR (238.50), PNI (48.48), LCR (0.52), and LMR (2.46) were determined by receiver operating characteristic (ROC) curve; Six conventional machine learning models were constructed using patient data in the training set: Linear Regression, Random Forest, Support Vector Machine (SVM), BP Neural Network (BP), Light Gradient Boosting Machine (LGBM), Extreme Gradient Boosting (XGBoost) and an improved moderately greedy XGBoost (MGA-XGBoost) model. The performance of the seven models was evaluated by area under the receiver operator characteristic curve, accuracy (ACC), precision, recall, and F1-score of the validation set.

RESULTS

Five hundred twelve cases were included in this study; 125 cases (24%) had postoperative infectious complications. Postoperative infectious complications were notably associated with 10 items features: American Society of Anesthesiologists scores (ASA), operation time, diabetes, presence of stomy, tumor location, NLR, PLR, PNI, LCR, and LMR. MGA-XGBoost reached the highest AUC (0.862) on the validation set, which was the best model for predicting postoperative infectious complications in elderly patients with colorectal cancer. Among the importance of the internal characteristics of the model, LCR accounted for the highest proportion.

CONCLUSIONS

This study demonstrates for the first time that the MGA-XGBoost model with 10 risk factors might predict postoperative infectious complications in elderly CRC patients.

Machine learning models to predict success of endoscopic sleeve gastroplasty using total and excess weight loss percent achievement: a multicentre study

BACKGROUND

The large amount of heterogeneous data collected in surgical/endoscopic practice calls for data-driven approaches as machine learning (ML) models. The aim of this study was to develop ML models to predict endoscopic sleeve gastroplasty (ESG) efficacy at 12 months defined by total weight loss (TWL) % and excess weight loss (EWL) % achievement. Multicentre data were used to enhance generalizability: evaluate consistency among different center of ESG practice and assess reproducibility of the models and possible clinical application. Models were designed to be dynamic and integrate follow-up clinical data into more accurate predictions, possibly assisting management and decision-making.

METHODS

ML models were developed using data of 404 ESG procedures performed at 12 centers across Europe. Collected data included clinical and demographic variables at the time of ESG and at follow-up. Multicentre/external and single center/internal and temporal validation were performed. Training and evaluation of the models were performed on Python's scikit-learn library. Performance of models was quantified as receiver operator curve (ROC-AUC), sensitivity, specificity, and calibration plots.

RESULTS

Multicenter external validation: ML models using preoperative data show poor performance. Best performances were reached by linear regression (LR) and support vector machine models for TWL% and EWL%, respectively, (ROC-AUC: TWL% 0.87, EWL% 0.86) with the addition of 6-month follow-up data. Single-center internal validation: Preoperative data only ML models show suboptimal performance. Early, i.e., 3-month follow-up data addition lead to ROC-AUC of 0.79 (random forest classifiers model) and 0.81 (LR models) for TWL% and EWL% achievement prediction, respectively. Single-center temporal validation shows similar results.

CONCLUSIONS

Although preoperative data only may not be sufficient for accurate postoperative predictions, the ability of ML models to adapt and evolve with the patients changes could assist in providing an effective and personalized postoperative care. ML models predictive capacity improvement with follow-up data is encouraging and may become a valuable support in patient management and decision-making.

Interpretable Machine Learning Models Using Peripheral Immune Cells to Predict 90-Day Readmission or Mortality in Acute Heart Failure Patients

BACKGROUND

Acute heart failure (AHF) carries a grave prognosis, marked by high readmission and mortality rates within 90 days post-discharge. This underscores the urgent need for enhanced care transitions, early monitoring, and precise interventions for at-risk individuals during this critical period.

OBJECTIVE

Our study aims to develop and validate an interpretable machine learning (ML) model that integrates peripheral immune cell data with conventional clinical markers. Our goal is to accurately predict 90-day readmission or mortality in patients AHF.

METHODS

In our study, we conducted a retrospective analysis on 1210 AHF patients, segregating them into training and external validation cohorts. Patients were categorized based on their 90-day outcomes post-discharge into groups of 'with readmission/mortality' and 'without readmission/mortality'. We developed various ML models using data from peripheral immune cells, traditional clinical indicators, or both, which were then internally validated. The feature importance of the most promising model was examined through the Shapley Additive Explanations (SHAP) method, culminating in external validation.

RESULTS

In our cohort of 1210 patients, 28.4% (344) faced readmission or mortality within 90 days post-discharge. Our study pinpointed 10 significant indicators-spanning peripheral immune cells and traditional clinical metrics-that predict these outcomes, with the support vector machine (SVM) model showing superior performance. SHAP analysis further distilled these predictors to five key determinants, including three clinical indicators and two immune cell types, essential for assessing 90-day readmission or mortality risks.

CONCLUSION

Our analysis identified the SVM model, which merges traditional clinical indicators and peripheral immune cells, as the most effective for predicting 90-day readmission or mortality in AHF patients. This innovative approach promises to refine risk assessment and enable more targeted interventions for at-risk individuals through continuous improvement.

Prediction of Complications and Prognostication in Perioperative Medicine: A Systematic Review and PROBAST Assessment of Machine Learning Tools

BACKGROUND

The utilization of artificial intelligence and machine learning as diagnostic and predictive tools in perioperative medicine holds great promise. Indeed, many studies have been performed in recent years to explore the potential. The purpose of this systematic review is to assess the current state of machine learning in perioperative medicine, its utility in prediction of complications and prognostication, and limitations related to bias and validation.

METHODS

A multidisciplinary team of clinicians and engineers conducted a systematic review using the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) protocol. Multiple databases were searched, including Scopus, Cumulative Index to Nursing and Allied Health Literature (CINAHL), the Cochrane Library, PubMed, Medline, Embase, and Web of Science. The systematic review focused on study design, type of machine learning model used, validation techniques applied, and reported model performance on prediction of complications and prognostication. This review further classified outcomes and machine learning applications using an ad hoc classification system. The Prediction model Risk Of Bias Assessment Tool (PROBAST) was used to assess risk of bias and applicability of the studies.

RESULTS

A total of 103 studies were identified. The models reported in the literature were primarily based on single-center validations (75%), with only 13% being externally validated across multiple centers. Most of the mortality models demonstrated a limited ability to discriminate and classify effectively. The PROBAST assessment indicated a high risk of systematic errors in predicted outcomes and artificial intelligence or machine learning applications.

CONCLUSIONS

The findings indicate that the development of this field is still in its early stages. This systematic review indicates that application of machine learning in perioperative medicine is still at an early stage. While many studies suggest potential utility, several key challenges must be first overcome before their introduction into clinical practice.

EDITOR’S PERSPECTIVE

Artificial intelligence in the prediction of venous thromboembolism: A systematic review and pooled analysis

BACKGROUND

Accurate diagnostic and prognostic predictions of venous thromboembolism (VTE) are crucial for VTE management. Artificial intelligence (AI) enables autonomous identification of the most predictive patterns from large complex data. Although evidence regarding its performance in VTE prediction is emerging, a comprehensive analysis of performance is lacking.

AIMS

To systematically review the performance of AI in the diagnosis and prediction of VTE and compare it to clinical risk assessment models (RAMs) or logistic regression models.

METHODS

A systematic literature search was performed using PubMed, MEDLINE, EMBASE, and Web of Science from inception to April 20, 2021. Search terms included "artificial intelligence" and "venous thromboembolism." Eligible criteria were original studies evaluating AI in the prediction of VTE in adults and reporting one of the following outcomes: sensitivity, specificity, positive predictive value, negative predictive value, or area under receiver operating curve (AUC). Risks of bias were assessed using the PROBAST tool. Unpaired t-test was performed to compare the mean AUC from AI versus conventional methods (RAMs or logistic regression models).

RESULTS

A total of 20 studies were included. Number of participants ranged from 31 to 111 888. The AI-based models included artificial neural network (six studies), support vector machines (four studies), Bayesian methods (one study), super learner ensemble (one study), genetic programming (one study), unspecified machine learning models (two studies), and multiple machine learning models (five studies). Twelve studies (60%) had both training and testing cohorts. Among 14 studies (70%) where AUCs were reported, the mean AUC for AI versus conventional methods were 0.79 (95% CI: 0.74-0.85) versus 0.61 (95% CI: 0.54-0.68), respectively (p < .001). However, the good to excellent discriminative performance of AI methods is unlikely to be replicated when used in clinical practice, because most studies had high risk of bias due to missing data handling and outcome determination.

CONCLUSION

The use of AI appears to improve the accuracy of diagnostic and prognostic prediction of VTE over conventional risk models; however, there was a high risk of bias observed across studies. Future studies should focus on transparent reporting, external validation, and clinical application of these models.

Development of postoperative delirium prediction models in patients undergoing cardiovascular surgery using machine learning algorithms

Associations between delirium and postoperative adverse events in cardiovascular surgery have been reported and the preoperative identification of high-risk patients of delirium is needed to implement focused interventions. We aimed to develop and validate machine learning models to predict post-cardiovascular surgery delirium. Patients aged ≥ 40 years who underwent cardiovascular surgery at a single hospital were prospectively enrolled. Preoperative and intraoperative factors were assessed. Each patient was evaluated for postoperative delirium 7 days after surgery. We developed machine learning models using the Bernoulli naive Bayes, Support vector machine, Random forest, Extra-trees, and XGBoost algorithms. Stratified fivefold cross-validation was performed for each developed model. Of the 87 patients, 24 (27.6%) developed postoperative delirium. Age, use of psychotropic drugs, cognitive function (Mini-Cog < 4), index of activities of daily living (Barthel Index < 100), history of stroke or cerebral hemorrhage, and eGFR (estimated glomerular filtration rate) < 60 were selected to develop delirium prediction models. The Extra-trees model had the best area under the receiver operating characteristic curve (0.76 [standard deviation 0.11]; sensitivity: 0.63; specificity: 0.78). XGBoost showed the highest sensitivity (AUROC, 0.75 [0.07]; sensitivity: 0.67; specificity: 0.79). Machine learning algorithms could predict post-cardiovascular delirium using preoperative data.Trial registration: UMIN-CTR (ID; UMIN000049390).

Interpretable machine-learning model for real-time, clustered risk factor analysis of sepsis and septic death in critical care

BACKGROUND AND OBJECTIVE

Interpretable and real-time prediction of sepsis and risk factor analysis could enable timely treatment by clinicians and improve patient outcomes. To develop an interpretable machine-learning model for the prediction and risk factor analysis of sepsis and septic death.

METHODS

This is a retrospective observational cohort study based on the Medical Information Mart for Intensive Care (MIMIC-IV) dataset; 69,619 patients from the database were screened. The two outcomes include patients diagnosed with sepsis and the death of septic patients. Clinical variables from ICU admission to outcomes were analyzed: demographic data, vital signs, Glasgow Coma Scale scores, laboratory test results, and results for arterial blood gasses (ABGs). Model performance was compared using the area under the receiver operating characteristic curve (AUROC). Model interpretations were based on the Shapley additive explanations (SHAP), and the clustered analysis was based on the combination of K-means and dimensionality reduction algorithms of t-SNE and PCA.

RESULTS

For the analysis of sepsis and septic death, 47,185 and 2480 patients were enrolled, respectively. The XGBoost model achieved a predictive value of area under the curve (AUC): 0.745 [0.731-0.759] for sepsis prediction and 0.8 [0.77, 0.828] for septic death prediction. The real-time prediction model was trained to predict by day and visualize the individual or combined risk factor effects on the outcomes based on SHAP values. Clustered analysis separated the two phenotypes with distinct risk factors among patients with septic death.

CONCLUSION

The proposed real-time, clustered prediction model for sepsis and septic death exhibited superior performance in predicting the outcomes and visualizing the risk factors in a real-time and interpretable manner to distinguish and mitigate patient risks, thus promising immense potential in effective clinical decision making and comprehensive understanding of complex diseases such as sepsis.

Temporal Generalizability of Machine Learning Models for Predicting Postoperative Delirium Using Electronic Health Record Data: Model Development and Validation Study

BACKGROUND

Although machine learning models demonstrate significant potential in predicting postoperative delirium, the advantages of their implementation in real-world settings remain unclear and require a comparison with conventional models in practical applications.

OBJECTIVE

The objective of this study was to validate the temporal generalizability of decision tree ensemble and sparse linear regression models for predicting delirium after surgery compared with that of the traditional logistic regression model.

METHODS

The health record data of patients hospitalized at an advanced emergency and critical care medical center in Kumamoto, Japan, were collected electronically. We developed a decision tree ensemble model using extreme gradient boosting (XGBoost) and a sparse linear regression model using least absolute shrinkage and selection operator (LASSO) regression. To evaluate the predictive performance of the model, we used the area under the receiver operating characteristic curve (AUROC) and the Matthews correlation coefficient (MCC) to measure discrimination and the slope and intercept of the regression between predicted and observed probabilities to measure calibration. The Brier score was evaluated as an overall performance metric. We included 11,863 consecutive patients who underwent surgery with general anesthesia between December 2017 and February 2022. The patients were divided into a derivation cohort before the COVID-19 pandemic and a validation cohort during the COVID-19 pandemic. Postoperative delirium was diagnosed according to the confusion assessment method.

RESULTS

A total of 6497 patients (68.5, SD 14.4 years, women n=2627, 40.4%) were included in the derivation cohort, and 5366 patients (67.8, SD 14.6 years, women n=2105, 39.2%) were included in the validation cohort. Regarding discrimination, the XGBoost model (AUROC 0.87-0.90 and MCC 0.34-0.44) did not significantly outperform the LASSO model (AUROC 0.86-0.89 and MCC 0.34-0.41). The logistic regression model (AUROC 0.84-0.88, MCC 0.33-0.40, slope 1.01-1.19, intercept -0.16 to 0.06, and Brier score 0.06-0.07), with 8 predictors (age, intensive care unit, neurosurgery, emergency admission, anesthesia time, BMI, blood loss during surgery, and use of an ambulance) achieved good predictive performance.

CONCLUSIONS

The XGBoost model did not significantly outperform the LASSO model in predicting postoperative delirium. Furthermore, a parsimonious logistic model with a few important predictors achieved comparable performance to machine learning models in predicting postoperative delirium.

Predictive Modeling and Integrated Risk Assessment of Postoperative Mortality and Pneumonia in Traumatic Brain Injury Patients through Clustering and Machine Learning: Retrospective Study

This study harnessed machine learning to forecast postoperative mortality (POM) and postoperative pneumonia (PPN) among surgical traumatic brain injury (TBI) patients. Our analysis centered on the following key variables: Glasgow Coma Scale (GCS), midline brain shift (MSB), and time from injury to emergency room arrival (TIE). Additionally, we introduced innovative clustered variables to enhance predictive accuracy and risk assessment. Exploring data from 617 patients spanning 2012 to 2022, we observed that 22.9% encountered postoperative mortality, while 30.0% faced postoperative pneumonia (PPN). Sensitivity for POM and PPN prediction, before incorporating clustering, was in the ranges of 0.43-0.82 (POM) and 0.54-0.76 (PPN). Following clustering, sensitivity values were 0.47-0.76 (POM) and 0.61-0.77 (PPN). Accuracy was in the ranges of 0.67-0.76 (POM) and 0.70-0.81 (PPN) prior to clustering and 0.42-0.73 (POM) and 0.55-0.73 (PPN) after clustering. Clusters characterized by low GCS, small MSB, and short TIE exhibited a 3.2-fold higher POM risk compared to clusters with high GCS, small MSB, and short TIE. In summary, leveraging clustered variables offers a novel avenue for predicting POM and PPN in TBI patients. Assessing the amalgamated impact of GCS, MSB, and TIE characteristics provides valuable insights for clinical decision making.

Prediction of lactate concentrations after cardiac surgery using machine learning and deep learning approaches

Background

Although conventional prediction models for surgical patients often ignore intraoperative time-series data, deep learning approaches are well-suited to incorporate time-varying and non-linear data with complex interactions. Blood lactate concentration is one important clinical marker that can reflect the adequacy of systemic perfusion during cardiac surgery. During cardiac surgery and cardiopulmonary bypass, minute-level data is available on key parameters that affect perfusion. The goal of this study was to use machine learning and deep learning approaches to predict maximum blood lactate concentrations after cardiac surgery. We hypothesized that models using minute-level intraoperative data as inputs would have the best predictive performance.

Methods

Adults who underwent cardiac surgery with cardiopulmonary bypass were eligible. The primary outcome was maximum lactate concentration within 24 h postoperatively. We considered three classes of predictive models, using the performance metric of mean absolute error across testing folds: (1) static models using baseline preoperative variables, (2) augmentation of the static models with intraoperative statistics, and (3) a dynamic approach that integrates preoperative variables with intraoperative time series data.

Results

2,187 patients were included. For three models that only used baseline characteristics (linear regression, random forest, artificial neural network) to predict maximum postoperative lactate concentration, the prediction error ranged from a median of 2.52 mmol/L (IQR 2.46, 2.56) to 2.58 mmol/L (IQR 2.54, 2.60). The inclusion of intraoperative summary statistics (including intraoperative lactate concentration) improved model performance, with the prediction error ranging from a median of 2.09 mmol/L (IQR 2.04, 2.14) to 2.12 mmol/L (IQR 2.06, 2.16). For two modelling approaches (recurrent neural network, transformer) that can utilize intraoperative time-series data, the lowest prediction error was obtained with a range of median 1.96 mmol/L (IQR 1.87, 2.05) to 1.97 mmol/L (IQR 1.92, 2.05). Intraoperative lactate concentration was the most important predictive feature based on Shapley additive values. Anemia and weight were also important predictors, but there was heterogeneity in the importance of other features.

Conclusion

Postoperative lactate concentrations can be predicted using baseline and intraoperative data with moderate accuracy. These results reflect the value of intraoperative data in the prediction of clinically relevant outcomes to guide perioperative management.

An Intraoperative Telemedicine Program to Improve Perioperative Quality Measures: The ACTFAST-3 Randomized Clinical Trial

Importance

Telemedicine for clinical decision support has been adopted in many health care settings, but its utility in improving intraoperative care has not been assessed.

Objective

To pilot the implementation of a real-time intraoperative telemedicine decision support program and evaluate whether it reduces postoperative hypothermia and hyperglycemia as well as other quality of care measures.

Design, Setting, and Participants

This single-center pilot randomized clinical trial (Anesthesiology Control Tower-Feedback Alerts to Supplement Treatments [ACTFAST-3]) was conducted from April 3, 2017, to June 30, 2019, at a large academic medical center in the US. A total of 26 254 adult surgical patients were randomized to receive either usual intraoperative care (control group; n = 12 980) or usual care augmented by telemedicine decision support (intervention group; n = 13 274). Data were initially analyzed from April 22 to May 19, 2021, with updates in November 2022 and February 2023.

Intervention

Patients received either usual care (medical direction from the anesthesia care team) or intraoperative anesthesia care monitored and augmented by decision support from the Anesthesiology Control Tower (ACT), a real-time, live telemedicine intervention. The ACT incorporated remote monitoring of operating rooms by a team of anesthesia clinicians with customized analysis software. The ACT reviewed alerts and electronic health record data to inform recommendations to operating room clinicians.

Main Outcomes and Measures

The primary outcomes were avoidance of postoperative hypothermia (defined as the proportion of patients with a final recorded intraoperative core temperature >36 °C) and hyperglycemia (defined as the proportion of patients with diabetes who had a blood glucose level ≤180 mg/dL on arrival to the postanesthesia recovery area). Secondary outcomes included intraoperative hypotension, temperature monitoring, timely antibiotic redosing, intraoperative glucose evaluation and management, neuromuscular blockade documentation, ventilator management, and volatile anesthetic overuse.

Results

Among 26 254 participants, 13 393 (51.0%) were female and 20 169 (76.8%) were White, with a median (IQR) age of 60 (47-69) years. There was no treatment effect on avoidance of hyperglycemia (7445 of 8676 patients [85.8%] in the intervention group vs 7559 of 8815 [85.8%] in the control group; rate ratio [RR], 1.00; 95% CI, 0.99-1.01) or hypothermia (7602 of 11 447 patients [66.4%] in the intervention group vs 7783 of 11 672 [66.7.%] in the control group; RR, 1.00; 95% CI, 0.97-1.02). Intraoperative glucose measurement was more common among patients with diabetes in the intervention group (RR, 1.07; 95% CI, 1.01-1.15), but other secondary outcomes were not significantly different.

Conclusions and Relevance

In this randomized clinical trial, anesthesia care quality measures did not differ between groups, with high confidence in the findings. These results suggest that the intervention did not affect the targeted care practices. Further streamlining of clinical decision support and workflows may help the intraoperative telemedicine program achieve improvement in targeted clinical measures.

Trial Registration

ClinicalTrials.gov Identifier: NCT02830126.

Artificial intelligence and "the Art of Kintsugi" in Anesthesiology: ten influential papers for clinical users

Artificial intelligence refers to the simulation of human intelligence in machines that are programmed to think like humans and mimic their actions. In the present review we chose ten influential papers from the last five years and through Kintsugi, shed the light on recent evolution of artificial intelligence in anesthesiology. A comprehensive search in in Medline, Embase, Web of Science and Scopus databases was conducted. Each author searched the databases independently and created a list of six articles that influenced their clinical practice during this period, with a focus on their area of competence. During a subsequent step, each researcher presented his own list and most cited papers were selected to create the final collection of ten articles. In recent years purely methodological works with a cryptic technology (black-box) represented by the intact and static vessel, translated to a "modern artificial intelligence" in clinical practice and comprehensibility (glass-box). The purposes of this review are to explore the ten most cited papers about artificial intelligence in anesthesiology and to understand how and when it should be integrated in clinical practice.

Harnessing Machine Learning for Prediction of Postoperative Pulmonary Complications: Retrospective Cohort Design

Postoperative pulmonary complications (PPCs) are significant causes of postoperative morbidity and mortality. This study presents the utilization of machine learning for predicting PPCs and aims to identify the important features of the prediction models. This study used a retrospective cohort design and collected data from two hospitals. The dataset included perioperative variables such as patient characteristics, preexisting diseases, and intraoperative factors. Various algorithms, including logistic regression, random forest, light-gradient boosting machines, extreme-gradient boosting machines, and multilayer perceptrons, have been employed for model development and evaluation. This study enrolled 111,212 adult patients, with an overall incidence rate of 8.6% for developing PPCs. The area under the receiver-operating characteristic curve (AUROC) of the models was 0.699-0.767, and the f1 score was 0.446-0.526. In the prediction models, except for multilayer perceptron, the 10 most important features were obtained. In feature-reduced models, including 10 important features, the AUROC was 0.627-0.749, and the f1 score was 0.365-0.485. The number of packed red cells, urine, and rocuronium doses were similar in the three models. In conclusion, machine learning provides valuable insights into PPC prediction, significant features for prediction, and the feasibility of models that reduce the number of features.

XGBoost-SHAP-based interpretable diagnostic framework for alzheimer's disease

BACKGROUND

Due to the class imbalance issue faced when Alzheimer's disease (AD) develops from normal cognition (NC) to mild cognitive impairment (MCI), present clinical practice is met with challenges regarding the auxiliary diagnosis of AD using machine learning (ML). This leads to low diagnosis performance. We aimed to construct an interpretable framework, extreme gradient boosting-Shapley additive explanations (XGBoost-SHAP), to handle the imbalance among different AD progression statuses at the algorithmic level. We also sought to achieve multiclassification of NC, MCI, and AD.

METHODS

We obtained patient data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database, including clinical information, neuropsychological test results, neuroimaging-derived biomarkers, and APOE-ε4 gene statuses. First, three feature selection algorithms were applied, and they were then included in the XGBoost algorithm. Due to the imbalance among the three classes, we changed the sample weight distribution to achieve multiclassification of NC, MCI, and AD. Then, the SHAP method was linked to XGBoost to form an interpretable framework. This framework utilized attribution ideas that quantified the impacts of model predictions into numerical values and analysed them based on their directions and sizes. Subsequently, the top 10 features (optimal subset) were used to simplify the clinical decision-making process, and their performance was compared with that of a random forest (RF), Bagging, AdaBoost, and a naive Bayes (NB) classifier. Finally, the National Alzheimer's Coordinating Center (NACC) dataset was employed to assess the impact path consistency of the features within the optimal subset.

RESULTS

Compared to the RF, Bagging, AdaBoost, NB and XGBoost (unweighted), the interpretable framework had higher classification performance with accuracy improvements of 0.74%, 0.74%, 1.46%, 13.18%, and 0.83%, respectively. The framework achieved high sensitivity (81.21%/74.85%), specificity (92.18%/89.86%), accuracy (87.57%/80.52%), area under the receiver operating characteristic curve (AUC) (0.91/0.88), positive clinical utility index (0.71/0.56), and negative clinical utility index (0.75/0.68) on the ADNI and NACC datasets, respectively. In the ADNI dataset, the top 10 features were found to have varying associations with the risk of AD onset based on their SHAP values. Specifically, the higher SHAP values of CDRSB, ADAS13, ADAS11, ventricle volume, ADASQ4, and FAQ were associated with higher risks of AD onset. Conversely, the higher SHAP values of LDELTOTAL, mPACCdigit, RAVLT_immediate, and MMSE were associated with lower risks of AD onset. Similar results were found for the NACC dataset.

CONCLUSIONS

The proposed interpretable framework contributes to achieving excellent performance in imbalanced AD multiclassification tasks and provides scientific guidance (optimal subset) for clinical decision-making, thereby facilitating disease management and offering new research ideas for optimizing AD prevention and treatment programs.

Improved Re-estimation of Perioperative Cardiac Risk Using the Surgical Apgar Score: A Retrospective Cohort Study

OBJECTIVE

To assess whether the Surgical Apgar Score (SAS) improves re-estimation of perioperative cardiac risk.

BACKGROUND

The SAS is a novel risk index that integrates three relevant and easily measurable intraoperative parameters (blood loss, heart rate, mean arterial pressure) to predict outcomes. The incremental prognostic value of the SAS when used in combination with standard preoperative risk indices is unclear.

METHODS

We conducted a retrospective cohort study of adults (18 years and older) who underwent elective noncardiac surgery at a quaternary care hospital in Canada (2009-2014). The primary outcome was postoperative acute myocardial injury. The SAS (range 0-10) was calculated based on intraoperative estimated blood loss, lowest mean arterial pressure, and lowest heart rate documented in electronic medical records. Incremental prognostic value of the SAS when combined with the Revised Cardiac Risk Index was assessed based on discrimination (c-statistic), reclassification (integrated discrimination improvement, net reclassification index), and clinical utility (decision curve analysis).

RESULTS

The cohort included 16,835 patients, of whom 607 (3.6%) patients had acute postoperative myocardial injury. Addition of the SAS to the Revised Cardiac Risk Index improved risk estimation based on the integrated discrimination improvement [2.0%; 95% confidence interval (CI): 1.5%-2.4%], continuous net reclassification index (54%; 95% CI: 46%-62%), and c-index, which increased from 0.68 (95% CI: 0.65-0.70) to 0.75 (95% CI: 0.73-0.77). On decision curve analysis, addition of the SAS to the Revised Cardiac Risk Index resulted in a higher net benefit at all decision thresholds.

CONCLUSIONS

When combined with a validated preoperative risk index, the SAS improved the accuracy of cardiac risk assessment for noncardiac surgery. Further research is needed to delineate how intraoperative data can better guide postoperative decision-making.