Prediction of Prolonged Intensive Care Unit Length of Stay Following Cardiac Surgery

Predictors of Prolonged Intensive Care Unit Stay and In-Hospital Mortality Following Cardiac Surgery: An Integrated Analysis from the PROCARD-ATI Study

Background: Despite advances in surgical techniques and perioperative management, reliable intraoperative predictors of adverse postoperative outcomes in cardiac surgery remain elusive. This study aimed to identify perioperative factors associated with prolonged intensive care unit (ICU) stay and in-hospital mortality while defining actionable thresholds. Methods: A retrospective analysis was conducted on 130 adult cardiac surgery patients (with a median age of 61 years, 66.2% men) from October 2022 to November 2024. Data on preoperative risk factors, intraoperative variables (cardiopulmonary bypass time-CPBT, aortic cross-clamp time-AXCT), and postoperative outcomes (ICU length of stay, in-hospital mortality) were extracted from electronic medical records. Results: Prolonged ICU stay (≥7 days) occurred in 38.5% of patients, and in-hospital mortality was 10%. AXCT was the sole independent predictor of prolonged ICU stay (OR = 1.046, 95% CI = 1.014-1.080, p = 0.005), with a 110-min cut-off (sensitivity 71%, specificity 61%, AUC = 0.729). A Kaplan-Meier analysis showed significantly longer ICU stays above this threshold (p = 0.006). For in-hospital mortality, prolonged CPBT (OR = 1.030, 95% CI = 1.003-1.057, p = 0.030), emergency surgery (OR = 0.043, 95% CI = 0.002-0.863, p = 0.040), and higher AXCT (OR = 0.965, 95% CI = 0.934-0.997, p = 0.034) were the independent predictors. A receiver operating characteristic analysis identified 140 min for AXCT (sensitivity 67%, specificity 70%, AUC = 0.707) and 227 min for CPBT (sensitivity 83%, specificity 69%, AUC = 0.824) as the optimal cut-offs. A combined model (emergency surgery yes/no, AXCT > 140 min, CPBT > 227 min) yielded excellent discrimination (AUC = 0.846). Conclusions: These findings suggest perioperative benchmarks that may guide surgical teams in refining operative strategies, reducing ICU resource utilization, and improving survival following cardiac surgery.

Predicting admission to and length of stay in intensive care units after general anesthesia: Time-dependent role of pre- and intraoperative data for clinical decision-making

BACKGROUND

Accurate prediction of intensive care unit (ICU) admission and length of stay (LOS) after major surgery is essential for optimizing patient outcomes and healthcare resources. Factors such as age, BMI, comorbidities, and perioperative complications significantly influence ICU admissions and LOS. Machine learning methods have been increasingly utilized to predict these outcomes, but their clinical utility beyond traditional metrics remains underexplored.

METHODS

This study examined a sub-cohort of 6043 patients who underwent general anesthesia at Seoul National University Hospital from August 2016 to June 2017. Various prediction models, including logistic regression and random forest, were developed for ICU admission and different LOS thresholds, e.g., a LOS of more than a week. Clinical utility was evaluated using decision curve analysis (DCA) across predefined risk preferences.

RESULTS

Among patients studied, 19.8 % were admitted to the ICU, with 1.4 % staying longer than a week. Prediction models demonstrated high discrimination (AUROC 0.93 to 0.96) and good calibration for ICU admission and short LOS. DCA revealed that intraoperative data provided the greatest decision-related benefit for predicting ICU admission, while preoperative data became more important for predicting longer LOS.

CONCLUSION

Intraoperative data are crucial for immediate postoperative decisions, while preoperative data are essential for extended LOS predictions. These findings highlight the need for a comprehensive risk assessment approach in perioperative care, utilizing both preoperative and intraoperative information to enhance clinical decision-making and resource allocation.

Machine Learning Model for Risk Prediction of Prolonged Intensive Care Unit in Patients Receiving Intra-aortic Balloon Pump Therapy during Coronary Artery Bypass Graft Surgery

This study aimed to construct machine learning models and predict prolonged intensive care units (ICU) stay in patients receiving perioperative intra-aortic balloon pump (IABP) therapy during cardiac surgery. 236 patients were divided into the normal (≤ 14 days) and prolonged (> 14 days) ICU groups based on the 75th percentile of ICU duration across the entire cohort. Seven machine learning models were trained and validated. The Shapley Additive explanations (SHAP) method was employed to illustrate the effects of the features. 94 patients (39.83%) experienced prolonged ICU stay. The XGBoost model outperformed other models in predictive performance, as evidenced by its highest area under the receiver operating characteristic curve (training: 0.92; validation: 0.73). The SHAP analysis identified tracheotomy, albumin, Sv1, and cardiac troponin T as the top four risk variables. The XGBoost model predicted risk variables for prolonged ICU stay in patients, possibly contributing to improving perioperative management and reducing ICU duration.

Frequency and Predictors of Pneumonia After Isolated Coronary Artery Bypass Grafting (CABG): A Single-Center Study

Background: CABG is a commonly performed procedure to improve survival and quality of life in patients with coronary artery disease. Despite advances in surgical techniques and perioperative care, postoperative pneumonia remains a serious complication contributing to increased morbidity, mortality and healthcare costs. This study aims to evaluate the incidence of postoperative pneumonia (POP) and identify its risk factors in patients undergoing isolated CABG. Methods: This retrospective study analyzed 430 patients who underwent CABG between 2019 and 2024. Patient demographics, clinical characteristics, surgical details and laboratory data were collected. Statistical analysis included univariate and multivariate logistic regression to identify significant predictors of pneumonia. Results: The incidence of POP after CABG was 10% (43/430). In patients with POP, diabetes mellitus (p = 0.03) and chronic kidney disease (p = 0.048) prevalence was higher, cardiopulmonary bypass (CPB) (p = 0.01) and cross-clamp time (p = 0.003) was longer, LDH levels (p = 0.017) were higher, hemoglobin (p = 0.012) and albumin (p = 0.015) levels were lower, and lymphocyte % (p = 0.04) was lower; prevalence of COPD and length of stay (LOS) in hospital tended to be higher (both p < 0.06). Multivariate binary logistic regression identified COPD (OR 4.383, 95% CI: 1.106-17.363, p = 0.035), CPB time (OR 1.013, 95% CI: 1.001-1.025, p = 0.030) and LOS (OR 1.052, 95% CI: 1.004-1.103, p = 0.035) as independent predictors of POP. Conclusions: Postoperative pneumonia is a common complication after CABG and is strongly associated with preoperative COPD, CPB time and length of stay in hospital. These findings underline the importance of preoperative risk assessment and optimization. Early identification of high-risk patients may allow targeted strategies such as enhanced respiratory support and prophylactic antibiotics to reduce the incidence of pneumonia and improve clinical outcomes.

Multi-institutional model to predict intensive care unit length of stay after cardiac surgery

OBJECTIVE

Intensive care unit length of stay (ICU LOS) accounts for a large percentage of inpatient cost after cardiac surgery. The Society of Thoracic Surgeons risk calculator predicts total LOS but does not discriminate between ICU and non-ICU time. We sought to develop a predictive model of prolonged ICU LOS.

METHODS

Adult patients undergoing Society of Thoracic Surgeons index operations within a regional collaborative (2014-2021) were included. Prolonged ICU LOS was defined as ICU care for ≥72 hours postoperatively. A logistic regression model was used to develop a prediction model for the prolonged ICU LOS with prespecified risk factors identified from our previous single-center study. Internal prediction model validation was determined by bootstrapping resampling method. The prediction model performance was assessed by measures of discrimination and calibration.

RESULTS

We identified 37,519 patients that met inclusion criteria with 11,801 (31.5%) patients experiencing prolonged ICU stay. From the logistic regression model, there were significant associations between prolonged ICU LOS and all pre-specified factors except sleep apnea (all P < .05). Model for End-Stage Liver Disease, preoperative intra-aortic balloon pump use, and procedure types were the most significant predictors of prolonged ICU LOS (all P < .0001). Our prediction model had not only a good discrimination power (bootstrapped-corrected C-index = 0.71) but also excellent calibration (bootstrapped-corrected mean absolute error = 0.005).

CONCLUSIONS

Prolonged ICU stay after cardiac surgery can be predicted with good predictive accuracy using preoperative data and may aid in patient counseling and resource allocation. Through use of a state-wide database, the application of this model may extend to other practices.

A Risk Prediction Model for Prolonged Length of Stay After Minimally Invasive Valve Surgery

OBJECTIVE

Minimally invasive surgery determines shorter postoperative hospital length of stay (LOS) even in cardiac surgery. Potential preoperative factors affecting LOS are still not known in minimally invasive heart valve surgery (MIVS). We aimed to identify preoperative variables influencing prolonged LOS in MIVS.

METHODS

We reviewed 189 patients who underwent MIVS via minithoracotomy at our institution. Prolonged LOS was defined as more than 7 postoperative days. Poisson and logistic regression were used to screen the predictors.

RESULTS

The mean postoperative LOS was 9.13 days, and 64 patients (33.9%) experienced a prolonged LOS. These patients were older, more frequently in New York Heart Association (NYHA) class III or IV, showed worse left ventricular ejection function (LVEF), and had a higher incidence of reoperation and chronic kidney disease (CKD). At univariate analysis, the most significant preoperative factors affecting prolonged LOS were age (odds ratio [OR] = 1.04), NYHA class III or IV (OR = 3.03), reduced LVEF (OR = 3.22), CKD (OR = 2.7), and redo surgery (OR = 3.6). After adjustment, the most significant preoperative factors predicting prolonged LOS were age (OR = 1.03, 95% CI: 1.01 to 1.06, P = 0.02) and redo surgery (OR = 3.33, 95% CI: 1.29 to 8.9, P = 0.01).

CONCLUSIONS

The most important factors affecting prolonged LOS after MIVS were represented by age and redo surgery, although other preoperative characteristics such as reduced LVEF, NYHA class III or IV, and CKD play a significant role in delaying recovery after MIVS. Further larger studies are needed to better identify potential preoperative predictors of prolonged LOS after MIVS.

The Impact of Silent Liver Disease on Hospital Length of Stay Following Isolated Coronary Artery Bypass Grafting Surgery

Objectives: Risk assessment models for cardiac surgery do not distinguish between degrees of liver dysfunction. We have previously shown that preoperative liver stiffness is associated with hospital length of stay following cardiac surgery. The authors hypothesized that a liver stiffness measurement (LSM) ≥ 9.5 kPa would rule out a short hospital length of stay (LOS < 6 days) following isolated coronary artery bypass grafting (CABG) surgery. Methods: A prospective observational study of one hundred sixty-four adult patients undergoing non-emergent isolated CABG surgery at a single university hospital center. Preoperative liver stiffness measured by ultrasound elastography was obtained for each participant. Multivariate logistic regression models were used to assess the adjusted relationship between LSM and a short hospital stay. Results: We performed multivariate logistic regression models using short hospital LOS (<6 days) as the dependent variable. Independent variables included LSM (< 9.5 kPa, ≥ 9.5 kPa), age, sex, STS predicted morbidity and mortality, and baseline hemoglobin. After adjusting for included variables, LSM ≥ 9.5 kPa was associated with lower odds of early discharge as compared to LSM < 9.5 kPa (OR: 0.22, 95% CI: 0.06-0.84, p = 0.03). The ROC curve and resulting AUC of 0.76 (95% CI: 0.68-0.83) suggest the final multivariate model provides good discriminatory performance when predicting early discharge. Conclusions: A preoperative LSM ≥ 9.5 kPa ruled out a short length of stay in nearly 80% of patients when compared to patients with a LSM < 9.5 kPa. Preoperative liver stiffness may be a useful metric to incorporate into preoperative risk stratification.

Preemptive Pharmacogenetic-Guided Metoprolol Management for Postoperative Atrial Fibrillation in Cardiac Surgery: The Preemptive Pharmacogenetic-Guided Metoprolol Management for Atrial Fibrillation in Cardiac Surgery Pilot Trial

OBJECTIVES

To test the hypothesis that implementation of a cytochrome P-450 2D6 (CYP2D6) genotype-guided perioperative metoprolol administration will reduce the risk of postoperative atrial fibrillation (AF), the authors conducted the Preemptive Pharmacogenetic-Guided Metoprolol Management for Atrial Fibrillation in Cardiac Surgery pilot study.

DESIGN

Clinical pilot trial.

SETTING

Single academic center.

PARTICIPANTS

Seventy-three cardiac surgery patients.

MEASUREMENTS AND MAIN RESULTS

Patients were classified as normal, intermediate, poor, or ultrarapid metabolizers after testing for their CYP2D6 genotype. A clinical decision support tool in the electronic health record advised providers on CYP2D6 genotype-guided metoprolol dosing. Using historical data, the Bayesian method was used to compare the incidence of postoperative AF in patients with altered metabolizer status to the reference incidence. A logistic regression analysis was performed to study the association between the metabolizer status and postoperative AF while controlling for the Multicenter Study of Perioperative Ischemia AF Risk Index. Of the 73 patients, 30% (n = 22) developed postoperative AF; 89% (n = 65) were normal metabolizers; 11% (n = 8) were poor/intermediate metabolizers; and there were no ultrarapid metabolizer patients identified. The estimated rate of postoperative AF in patients with altered metabolizer status was 30% (95% CI 8%-60%), compared with the historical reference incidence (27%). In the risk-adjusted analysis, there was insufficient evidence to conclude that modifying metoprolol dosing based on poor/intermediate metabolizer status was associated significantly with the odds of postoperative AF (odds ratio 0.82, 95% CI 0.15-4.55, p = 0.82).

CONCLUSIONS

A CYP2D6 genotype-guided metoprolol management was not associated with a reduction of postoperative AF after cardiac surgery.

Derivation and validation of pragmatic clinical models to predict hospital length of stay after cardiac surgery in Ontario, Canada: a population-based cohort study

BACKGROUND

Cardiac surgery is resource intensive and often requires multidisciplinary involvement to facilitate discharge. To facilitate evidence-based resource planning, we derived and validated clinical models to predict postoperative hospital length of stay (LOS).

METHODS

We used linked, population-level databases with information on all Ontario residents and included patients aged 18 years or older who underwent coronary artery bypass grafting, valvular or thoracic aorta surgeries between October 2008 and September 2019. The primary outcome was hospital LOS. The models were derived by using patients who had surgery before Sept. 30, 2016, and validated after that date. To address the rightward skew in LOS data and to identify top-tier resource users, we used logistic regression to derive a model to predict the likelihood of LOS being more than the 98th percentile (> 30 d), and γ regression in the remainder to predict continuous LOS in days. We used backward stepwise variable selection for both models.

RESULTS

Among 105 193 patients, 2422 (2.3%) had an LOS of more than 30 days. Factors predicting prolonged LOS included age, female sex, procedure type and urgency, comorbidities including frailty, high-risk acute coronary syndrome, heart failure, reduced left ventricular ejection fraction and psychiatric and pulmonary circulatory disease. The C statistic was 0.92 for the prolonged LOS model and the mean absolute error was 2.4 days for the continuous LOS model.

INTERPRETATION

We derived and validated clinical models to identify top-tier resource users and predict continuous LOS with excellent accuracy. Our models could be used to benchmark clinical performance based on expected LOS, rationally allocate resources and support patient-centred operative decision-making.

Effect of Algoplaque Hydrocolloid Dressing Combined with Nanosilver Antibacterial Gel under Predictive Nursing in the Treatment of Medical Device-Related Pressure Injury

It was aimed at the clinical value of predictive nursing and Algoplaque hydrocolloid dressing (AHD) combined with nanosilver antibacterial gel in treating medical device-related pressure injury (MDRPI). 100 patients, who underwent surgery in Chongqing Qijiang District People's Hospital from February 2019 to February 2020, were selected as the research objects and were randomly divided into the experimental group (50 cases) and the control group (50 cases). For the characterization test, a nanosilver antibacterial gel was created first. Patients in both groups received predictive nursing, but those in the experimental group received AHD and nanosilver antibacterial gel, and those in the control group received gauzes. MDRPI incidence, pressed skin injury severity, comfort level, clothing changes, nursing satisfaction, and other factors were all compared. The particle size of the nanosilver gel was 45-85 nm, with a relatively homogeneous distribution with the medium size, according to the findings. The incidence of MDRPI in the experimental group was lower than that in the control group significantly (6% vs. 30%, P < 0.05). The degree of injury of pressured skin in the experimental group was milder than that in the control group (P < 0.05), the degree of comfort and nursing satisfaction was higher in the experimental group than in the control group (P < 0.05), and dressing change count was lower than that in the control group (P < 0.05). In the treatment of MDRPI, predictive nursing and AHD using nanosilver antibacterial gel showed high clinical application value.

A novel nomogram for predicting 3-year mortality in critically ill patients after coronary artery bypass grafting

Background

The long-term outcomes for patients after coronary artery bypass grafting (CABG) have been received more and more concern. The existing prediction models are mostly focused on in-hospital operative mortality after CABG, but there is still little research on long-term mortality prediction model for patients after CABG.

Objective

To develop and validate a novel nomogram for predicting 3-year mortality in critically ill patients after CABG.

Methods

Data for developing novel predictive model were extracted from Medical Information Mart for Intensive cart III (MIMIC-III), of which 2929 critically ill patients who underwent CABG at the first admission were enrolled.

Results

A novel prognostic nomogram for 3-year mortality was constructed with the seven independent prognostic factors, including age, congestive heart failure, white blood cell, creatinine, SpO₂, anion gap, and continuous renal replacement treatment derived from the multivariable logistic regression. The nomogram indicated accurate discrimination in primary (AUC: 0.81) and validation cohort (AUC: 0.802), which were better than traditional severity scores. And good consistency between the predictive and observed outcome was showed by the calibration curve for 3-year mortality. The decision curve analysis also showed higher clinical net benefit than traditional severity scores.

Conclusion

The novel nomogram had well performance to predict 3-year mortality in critically ill patients after CABG. The prediction model provided valuable information for treatment strategy and postdischarge management, which may be helpful in improving the long-term prognosis in critically ill patients after CABG.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12893-021-01408-8.

Predicting Prolonged Length of ICU Stay through Machine Learning

This study aimed to construct machine learning (ML) models for predicting prolonged length of stay (pLOS) in intensive care units (ICU) among general ICU patients. A multicenter database called eICU (Collaborative Research Database) was used for model derivation and internal validation, and the Medical Information Mart for Intensive Care (MIMIC) III database was used for external validation. We used four different ML methods (random forest, support vector machine, deep learning, and gradient boosting decision tree (GBDT)) to develop prediction models. The prediction performance of the four models were compared with the customized simplified acute physiology score (SAPS) II. The area under the receiver operation characteristic curve (AUROC), area under the precision-recall curve (AUPRC), estimated calibration index (ECI), and Brier score were used to measure performance. In internal validation, the GBDT model achieved the best overall performance (Brier score, 0.164), discrimination (AUROC, 0.742; AUPRC, 0.537), and calibration (ECI, 8.224). In external validation, the GBDT model also achieved the best overall performance (Brier score, 0.166), discrimination (AUROC, 0.747; AUPRC, 0.536), and calibration (ECI, 8.294). External validation showed that the calibration curve of the GBDT model was an optimal fit, and four ML models outperformed the customized SAPS II model. The GBDT-based pLOS-ICU prediction model had the best prediction performance among the five models on both internal and external datasets. Furthermore, it has the potential to assist ICU physicians to identify patients with pLOS-ICU risk and provide appropriate clinical interventions to improve patient outcomes.