Dataset Information

Prediction of intensive care unit admission (>24h) after surgery in elective noncardiac surgical patients using machine learning algorithms

ABSTRACT:

Background

To develop a highly discriminative machine learning model for the prediction of intensive care unit admission (>24h) using the easily available preoperative information from electronic health records. An accurate prediction model for ICU admission after surgery is of great importance for surgical risk assessment and appropriate utilization of ICU resources.

Method

Data were collected retrospectively from a large hospital, comprising 135,442 adult patients who underwent surgery except for cardiac surgery between 1 January 2014, and 31 July 2018 in China. Multiple existing predictive machine learning algorithms were explored to construct the prediction model, including logistic regression, random forest, adaptive boosting, and gradient boosting machine. Four secondary analyses were conducted to improve the interpretability of the results.

Results

A total of 2702 (2.0%) patients were admitted to the intensive care unit postoperatively. The gradient boosting machine model attained the highest area under the receiver operating characteristic curve of 0.90. The machine learning models predicted intensive care unit admission better than the American Society of Anesthesiologists Physical Status (area under the receiver operating characteristic curve: 0.68). The gradient boosting machine recognized several features as highly significant predictors for postoperatively intensive care unit admission. By applying subgroup analysis and secondary analysis, we found that patients with operations on the digestive, respiratory, and vascular systems had higher probabilities for intensive care unit admission.

Conclusion

Compared with conventional American Society of Anesthesiologists Physical Status and logistic regression model, the gradient boosting machine could improve the performance in the prediction of intensive care unit admission. Machine learning models could be used to improve the discrimination and identify the need for intensive care unit admission after surgery in elective noncardiac surgical patients, which could help manage the surgical risk.

SUBMITTER: Lan L

PROVIDER: S-EPMC9326842 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:ObjectiveTo compare the performance of machine learning models against the traditionally derived Combined Assessment of Risk Encountered in Surgery (CARES) model and the American Society of Anaesthesiologists-Physical Status (ASA-PS) in the prediction of 30-day postsurgical mortality and need for intensive care unit (ICU) stay >24 hours.BackgroundPrediction of surgical risk preoperatively is important for clinical shared decision-making and planning of health resources such as ICU beds. The current growth of electronic medical records coupled with machine learning presents an opportunity to improve the performance of established risk models.MethodsAll patients aged 18 years and above who underwent noncardiac and nonneurological surgery at Singapore General Hospital (SGH) between 1 January 2012 and 31 October 2016 were included. Patient demographics, comorbidities, preoperative laboratory results, and surgery details were obtained from their electronic medical records. Seventy percent of the observations were randomly selected for training, leaving 30% for testing. Baseline models were CARES and ASA-PS. Candidate models were trained using random forest, adaptive boosting, gradient boosting, and support vector machine. Models were evaluated on area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC).ResultsA total of 90,785 patients were included, of whom 539 (0.6%) died within 30 days and 1264 (1.4%) required ICU admission >24 hours postoperatively. Baseline models achieved high AUROCs despite poor sensitivities by predicting all negative in a predominantly negative dataset. Gradient boosting was the best performing model with AUPRCs of 0.23 and 0.38 for mortality and ICU admission outcomes respectively.ConclusionsMachine learning can be used to improve surgical risk prediction compared to traditional risk calculators. AUPRC should be used to evaluate model predictive performance instead of AUROC when the dataset is imbalanced.

Project description:Assessment of need for intravascular volume resuscitation remains challenging for anesthesiologists. Dynamic waveform indices, including systolic and pulse pressure variation, are demonstrated as reliable measures of fluid responsiveness for mechanically ventilated patients. Despite widespread use, real-world reference distributions for systolic and pulse pressure variation values have not been established for euvolemic intraoperative patients. The authors sought to establish systolic and pulse pressure variation reference distributions and assess the impact of modifying factors.The authors evaluated adult patients undergoing general anesthetics for elective noncardiac surgery. Median systolic and pulse pressure variations during a 50-min postinduction period were noted for each case. Modifying factors including body mass index, age, ventilator settings, positioning, and hemodynamic management were studied via univariate and multivariable analyses. For systolic pressure variation values, effects of data entry method (manually entered vs. automated recorded) were similarly studied.Among 1,791 cases, per-case median systolic and pulse pressure variation values formed nonparametric distributions. For each distribution, median values, interquartile ranges, and reference intervals (2.5th to 97.5th percentile) were, respectively, noted: these included manually entered systolic pressure variation (6.0, 5.0 to 7.0, and 3.0 to 11.0 mmHg), automated systolic pressure variation (4.7, 3.9 to 6.0, and 2.2 to 10.4 mmHg), and automated pulse pressure variation (7.0, 5.0 to 9.0, and 2.0 to 16.0%). Nonsupine positioning and preoperative ? blocker were independently associated with altered systolic and pulse pressure variations, whereas ventilator tidal volume more than 8?ml/kg ideal body weight and peak inspiratory pressure more than 16?cm H2O demonstrated independent associations for systolic pressure variation only.This study establishes real-world systolic and pulse pressure variation reference distributions absent in the current literature. Through a consideration of reference distributions and modifying factors, the authors' study provides further evidence for assessing intraoperative volume status and fluid management therapies.

Project description:BackgroundAortic cross-clamp utilized during cardiac surgery facilitates motionless and bloodless surgical field. However, the duration of clamp time has an inverse effect on early post-operative recovery period. In this study, we sought to examine the predictors of aortic clamp duration and intensive care unit length of stay.ResultsSix hundred and nine adult patients presented for elective cardiac surgery between December 2019 and December 2020 were enrolled. The age of patients ranged from 18 to 82 years (mean 55.62 years, SD ± 12.3 years). Male/female ratio is 4.6:1. Most patients (87.2%) were planned for coronary artery bypass grafting (CABG) and 78 patients (12.8%) for single heart valve procedure. Operative time (OT) ranged from 120 to 402 min and averaged 259.4 min (SD ± 45.9 min). ACC time ranged from 15 to 159 min and averaged 50.56 min (SD ± 19.4 min). Factors associated with significantly longer ACCT were: smoking (OR = 1.89 (95% CI 1.3-2.74), p value = 0.01), respiratory disease (OR = 0.48 (95% CI 0.24-0.96), p value = 0.039), obesity (OR = 1.76 (95% CI 1.18-2.63), p value = 0.005) and AVR (OR = 2.11 (95% CI 1.17-3.83), p value = 0.013). Low cardiac output syndrome (LCOS) was observed in 19.2% of patients. Longer than average ACCT was associated with increased use of inotropes (p value < 0.001), intra-aortic balloon pump (p value < 0.001) and first 24 h post-operative blood loss (p value < 0.001). The average intensive care unit length of stay (ICULOS) was 1.64 days (SD ± 1.1 days). Patients' ACCT converged positively and significantly on ICULOS (Beta coefficient = 1.013 (95% CI 1.01-1.015), p value < 0.001).ConclusionAortic cross-clamping is a crucial method in cardiac surgery to achieve motionless field; however, prolongation of this method had an incremental risks for the development of low cardiac output syndrome, increased first 24 h post-operative blood loss and longer stay in the intensive care unit.