Prediction models for prolonged intensive care unit stay after cardiac surgery: systematic review and validation study

Transfusion of 1 and 2 units of red blood cells does not increase mortality and organ failure in patients undergoing isolated coronary artery bypass grafting

OBJECTIVES

In cardiac surgery, the association between red blood cell (RBC) transfusion and clinical outcome is elusive. We investigated in a large cohort of patients who underwent isolated coronary artery bypass grafting (CABG) the effect of transfusion of 1-2 units of leucocyte-depleted RBCs on mortality and multiorgan failure.

METHODS

The investigation included all patients from July 2009 to June 2014 who underwent CABG at our institution and received no (n = 1478) or 1-2 units of RBCs (n = 1528). The primary end-point was 30-day mortality; secondary end-points were major organ dysfunction. A subgroup analysis assessed the effect of the duration of RBC storage on patient outcome. Statistical analysis was performed using propensity score (PS) adjustment.

RESULTS

The 30-day mortality rate was 0.3% in the RBC- group and 0.2% in the RBC+ group. Compared with the RBC- group, PS-adjusted odds ratio (OR) of 30-day mortality in the RBC+ group was 0.29 [95% confidence interval (CI): 0.06-1.50; P = 0.14]. PS-adjusted OR of a 'prolonged intensive care unit (ICU) stay' (>48 h) was significantly higher in the RBC+ group than in the RBC- group [OR 1.49 (95% CI: 1.14-1.95); P = 0.004], but major clinical complications such as low cardiac output syndrome, stroke, haemofiltration, wound infection and prolonged mechanical ventilator support (>24 h) did not differ significantly between groups. Duration of blood storage was not independently associated with clinical outcome.

CONCLUSIONS

Our data do not indicate a transfusion-related increase in mortality and multiorgan failure in patients undergoing isolated CABG.

Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD): The TRIPOD Statement

CONTEXT

Prediction models are developed to aid health care providers in estimating the probability or risk that a specific disease or condition is present (diagnostic models) or that a specific event will occur in the future (prognostic models), to inform their decision making. However, the overwhelming evidence shows that the quality of reporting of prediction model studies is poor. Only with full and clear reporting of information on all aspects of a prediction model can risk of bias and potential usefulness of prediction models be adequately assessed.

OBJECTIVE

The Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) Initiative developed a set of recommendations for the reporting of studies developing, validating, or updating a prediction model, whether for diagnostic or prognostic purposes.

EVIDENCE ACQUISITION

This article describes how the TRIPOD Statement was developed. An extensive list of items based on a review of the literature was created, which was reduced after a Web-based survey and revised during a 3-day meeting in June 2011 with methodologists, health care professionals, and journal editors. The list was refined during several meetings of the steering group and in e-mail discussions with the wider group of TRIPOD contributors.

EVIDENCE SYNTHESIS

The resulting TRIPOD Statement is a checklist of 22 items, deemed essential for transparent reporting of a prediction model study. The TRIPOD Statement aims to improve the transparency of the reporting of a prediction model study regardless of the study methods used. The TRIPOD Statement is best used in conjunction with the TRIPOD explanation and elaboration document.

CONCLUSIONS

To aid the editorial process and readers of prediction model studies, it is recommended that authors include a completed checklist in their submission (also available at www.tripod-statement.org).

PATIENT SUMMARY

The Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) Initiative developed a set of recommendations for the reporting of studies developing, validating, or updating a prediction model, whether for diagnostic or prognostic purposes.

Evaluation of Serial Arterial Lactate Levels as a Predictor of Hospital and Long-Term Mortality in Patients After Cardiac Surgery

OBJECTIVES

Although hyperlactatemia is common after cardiac surgery, its value as a prognostic marker is unclear. The aim of the present study was to determine whether postoperative serial arterial lactate (AL) measurements after cardiac surgery could predict outcome.

DESIGN

Prospective, observational study.

SETTING

Surgical intensive care unit in a tertiary-level university hospital.

PARTICIPANTS

Participants included 2,935 consecutive patients.

INTERVENTIONS

AL was measured on admission to the intensive care unit and 6, 12, and 24 hours after surgery, and evaluated together with clinical data and outcomes including in-hospital and long-term mortality.

MEASUREMENTS AND MAIN RESULTS

In-hospital and long-term mortality (mean follow-up 6.3±1.7 years) were 5.9% and 8.7%, respectively. Compared with survivors, nonsurvivors showed higher mean AL values in all measurements (p<0.001). Hyperlactatemia (AL>3.0 mmol/L) was a predictor for in-hospital mortality (odds ratio = 1.468; 95% confidence interval = 1.239-1.739; p<0.001) and long-term mortality (hazard ratio = 1.511; 95% confidence interval = 1.251-1.825; p<0.001). Recent myocardial infarction and longer cardiopulmonary bypass time were predictors of hyperlactatemia. The pattern of AL dynamics was similar in both groups, but nonsurvivors showed higher AL values, as confirmed by repeated measures analysis of variance (p<0.001). The area under the curve also showed higher levels of AL in nonsurvivors (80.9±68.2 v 49.71±25.8 mmol/L/h; p = 0.038). Patients with hyperlactatemia were divided according to their timing of peak AL, with higher mortality and worse survival in patients in whom AL peaked at 24 hours compared with other groups (79.1% v 86.7%-89.2%; p = 0.03).

CONCLUSIONS

The dynamics of the postoperative AL curve in patients undergoing cardiac surgery suggests a similar mechanism of hyperlactatemia in survivors and nonsurvivors, albeit with a higher production or lower clearance of AL in nonsurvivors. The presence of a peak of hyperlactatemia at 24 hours is associated with higher in-hospital and long-term mortality.

Transparent reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): the TRIPOD statement

Prediction models are developed to aid health care providers in estimating the probability or risk that a specific disease or condition is present (diagnostic models) or that a specific event will occur in the future (prognostic models), to inform their decision making. However, the overwhelming evidence shows that the quality of reporting of prediction model studies is poor. Only with full and clear reporting of information on all aspects of a prediction model can risk of bias and potential usefulness of prediction models be adequately assessed. The Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) Initiative developed a set of recommendations for the reporting of studies developing, validating, or updating a prediction model, whether for diagnostic or prognostic purposes. This article describes how the TRIPOD Statement was developed. An extensive list of items based on a review of the literature was created, which was reduced after a Web-based survey and revised during a 3-day meeting in June 2011 with methodologists, health care professionals, and journal editors. The list was refined during several meetings of the steering group and in e-mail discussions with the wider group of TRIPOD contributors. The resulting TRIPOD Statement is a checklist of 22 items, deemed essential for transparent reporting of a prediction model study. The TRIPOD Statement aims to improve the transparency of the reporting of a prediction model study regardless of the study methods used. The TRIPOD Statement is best used in conjunction with the TRIPOD explanation and elaboration document. To aid the editorial process and readers of prediction model studies, it is recommended that authors include a completed checklist in their submission (also available at www.tripod-statement.org).

Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement

Prediction models are developed to aid health care providers in estimating the probability or risk that a specific disease or condition is present (diagnostic models) or that a specific event will occur in the future (prognostic models), to inform their decision making. However, the overwhelming evidence shows that the quality of reporting of prediction model studies is poor. Only with full and clear reporting of information on all aspects of a prediction model can risk of bias and potential usefulness of prediction models be adequately assessed. The Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD) Initiative developed a set of recommendations for the reporting of studies developing, validating, or updating a prediction model, whether for diagnostic or prognostic purposes. This article describes how the TRIPOD Statement was developed. An extensive list of items based on a review of the literature was created, which was reduced after a Web-based survey and revised during a 3-day meeting in June 2011 with methodologists, health care professionals, and journal editors. The list was refined during several meetings of the steering group and in e-mail discussions with the wider group of TRIPOD contributors. The resulting TRIPOD Statement is a checklist of 22 items, deemed essential for transparent reporting of a prediction model study. The TRIPOD Statement aims to improve the transparency of the reporting of a prediction model study regardless of the study methods used. The TRIPOD Statement is best used in conjunction with the TRIPOD explanation and elaboration document. To aid the editorial process and readers of prediction model studies, it is recommended that authors include a completed checklist in their submission (also available at www.tripod-statement.org).

Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): the TRIPOD statement

Prediction models are developed to aid healthcare providers in estimating the probability or risk that a specific disease or condition is present (diagnostic models) or that a specific event will occur in the future (prognostic models), to inform their decision-making. However, the overwhelming evidence shows that the quality of reporting of prediction model studies is poor. Only with full and clear reporting of information on all aspects of a prediction model can risk of bias and potential usefulness of prediction models be adequately assessed. The Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD) initiative developed a set of recommendations for the reporting of studies developing, validating or updating a prediction model, whether for diagnostic or prognostic purposes. This article describes how the TRIPOD Statement was developed. An extensive list of items based on a review of the literature was created, which was reduced after a web-based survey and revised during a 3-day meeting in June 2011 with methodologists, healthcare professionals and journal editors. The list was refined during several meetings of the steering group and in e-mail discussions with the wider group of TRIPOD contributors. The resulting TRIPOD Statement is a checklist of 22 items, deemed essential for transparent reporting of a prediction model study. The TRIPOD Statement aims to improve the transparency of the reporting of a prediction model study, regardless of the study methods used. The TRIPOD Statement is best used in conjunction with the TRIPOD explanation and elaboration document. To aid the editorial process and readers of prediction model studies, it is recommended that authors include a completed checklist in their submission (also available at www.tripod-statement.org).

Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD Statement

BACKGROUND

Prediction models are developed to aid healthcare providers in estimating the probability or risk that a specific disease or condition is present (diagnostic models) or that a specific event will occur in the future (prognostic models), to inform their decision-making. However, the overwhelming evidence shows that the quality of reporting of prediction model studies is poor. Only with full and clear reporting of information on all aspects of a prediction model can risk of bias and potential usefulness of prediction models be adequately assessed.

MATERIALS AND METHODS

The Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) initiative developed a set of recommendations for the reporting of studies developing, validating or updating a prediction model, whether for diagnostic or prognostic purposes. This article describes how the TRIPOD Statement was developed. An extensive list of items based on a review of the literature was created, which was reduced after a Web-based survey and revised during a 3-day meeting in June 2011 with methodologists, healthcare professionals and journal editors. The list was refined during several meetings of the steering group and in e-mail discussions with the wider group of TRIPOD contributors.

RESULTS

The resulting TRIPOD Statement is a checklist of 22 items, deemed essential for transparent reporting of a prediction model study. The TRIPOD Statement aims to improve the transparency of the reporting of a prediction model study regardless of the study methods used. The TRIPOD Statement is best used in conjunction with the TRIPOD explanation and elaboration document.

CONCLUSIONS

To aid the editorial process and readers of prediction model studies, it is recommended that authors include a completed checklist in their submission (also available at www.tripod-statement.org).

Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD): the TRIPOD Statement

BACKGROUND

Prediction models are developed to aid healthcare providers in estimating the probability or risk that a specific disease or condition is present (diagnostic models) or that a specific event will occur in the future (prognostic models), to inform their decision-making. However, the overwhelming evidence shows that the quality of reporting of prediction model studies is poor. Only with full and clear reporting of information on all aspects of a prediction model can risk of bias and potential usefulness of prediction models be adequately assessed. The Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) Initiative developed a set of recommendations for the reporting of studies developing, validating or updating a prediction model, whether for diagnostic or prognostic purposes. This article describes how the TRIPOD Statement was developed.

METHODS

An extensive list of items based on a review of the literature was created, which was reduced after a web-based survey and revised during a 3-day meeting in June 2011 with methodologists, healthcare professionals and journal editors. The list was refined during several meetings of the steering group and in e-mail discussions with the wider group of TRIPOD contributors.

RESULTS

The resulting TRIPOD Statement is a checklist of 22 items, deemed essential for transparent reporting of a prediction model study.

CONCLUSION

The TRIPOD Statement aims to improve the transparency of the reporting of a prediction model study regardless of the study methods used. The TRIPOD Statement is best used in conjunction with the TRIPOD explanation and elaboration document. A complete checklist is available at http://www.tripod-statement.org.

Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement

Prediction models are developed to aid health-care providers in estimating the probability or risk that a specific disease or condition is present (diagnostic models) or that a specific event will occur in the future (prognostic models), to inform their decision making. However, the overwhelming evidence shows that the quality of reporting of prediction model studies is poor. Only with full and clear reporting of information on all aspects of a prediction model can risk of bias and potential usefulness of prediction models be adequately assessed. The Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) Initiative developed a set of recommendations for the reporting of studies developing, validating, or updating a prediction model, whether for diagnostic or prognostic purposes. This article describes how the TRIPOD Statement was developed. An extensive list of items based on a review of the literature was created, which was reduced after a Web-based survey and revised during a 3-day meeting in June 2011 with methodologists, health-care professionals, and journal editors. The list was refined during several meetings of the steering group and in e-mail discussions with the wider group of TRIPOD contributors. The resulting TRIPOD Statement is a checklist of 22 items, deemed essential for transparent reporting of a prediction model study. The TRIPOD Statement aims to improve the transparency of the reporting of a prediction model study regardless of the study methods used. The TRIPOD Statement is best used in conjunction with the TRIPOD explanation and elaboration document. To aid the editorial process and readers of prediction model studies, it is recommended that authors include a completed checklist in their submission (also available at www.tripod-statement.org).

Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. The TRIPOD Group

BACKGROUND

Prediction models are developed to aid health care providers in estimating the probability or risk that a specific disease or condition is present (diagnostic models) or that a specific event will occur in the future (prognostic models), to inform their decision making. However, the overwhelming evidence shows that the quality of reporting of prediction model studies is poor. Only with full and clear reporting of information on all aspects of a prediction model can risk of bias and potential usefulness of prediction models be adequately assessed.

METHODS

The Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) Initiative developed a set of recommendations for the reporting of studies developing, validating, or updating a prediction model, whether for diagnostic or prognostic purposes. This article describes how the TRIPOD Statement was developed. An extensive list of items based on a review of the literature was created, which was reduced after a Web-based survey and revised during a 3-day meeting in June 2011 with methodologists, health care professionals, and journal editors. The list was refined during several meetings of the steering group and in e-mail discussions with the wider group of TRIPOD contributors.

RESULTS

The resulting TRIPOD Statement is a checklist of 22 items, deemed essential for transparent reporting of a prediction model study. The TRIPOD Statement aims to improve the transparency of the reporting of a prediction model study regardless of the study methods used. The TRIPOD Statement is best used in conjunction with the TRIPOD explanation and elaboration document.

CONCLUSIONS

To aid the editorial process and readers of prediction model studies, it is recommended that authors include a completed checklist in their submission (also available at www.tripod-statement.org).

Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): explanation and elaboration

The TRIPOD (Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis) Statement includes a 22-item checklist, which aims to improve the reporting of studies developing, validating, or updating a prediction model, whether for diagnostic or prognostic purposes. The TRIPOD Statement aims to improve the transparency of the reporting of a prediction model study regardless of the study methods used. This explanation and elaboration document describes the rationale; clarifies the meaning of each item; and discusses why transparent reporting is important, with a view to assessing risk of bias and clinical usefulness of the prediction model. Each checklist item of the TRIPOD Statement is explained in detail and accompanied by published examples of good reporting. The document also provides a valuable reference of issues to consider when designing, conducting, and analyzing prediction model studies. To aid the editorial process and help peer reviewers and, ultimately, readers and systematic reviewers of prediction model studies, it is recommended that authors include a completed checklist in their submission. The TRIPOD checklist can also be downloaded from www.tripod-statement.org.

Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): the TRIPOD statement

Prediction models are developed to aid health care providers in estimating the probability or risk that a specific disease or condition is present (diagnostic models) or that a specific event will occur in the future (prognostic models), to inform their decision making. However, the overwhelming evidence shows that the quality of reporting of prediction model studies is poor. Only with full and clear reporting of information on all aspects of a prediction model can risk of bias and potential usefulness of prediction models be adequately assessed. The Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) Initiative developed a set of recommendations for the reporting of studies developing, validating, or updating a prediction model, whether for diagnostic or prognostic purposes. This article describes how the TRIPOD Statement was developed. An extensive list of items based on a review of the literature was created, which was reduced after a Web-based survey and revised during a 3-day meeting in June 2011 with methodologists, health care professionals, and journal editors. The list was refined during several meetings of the steering group and in e-mail discussions with the wider group of TRIPOD contributors. The resulting TRIPOD Statement is a checklist of 22 items, deemed essential for transparent reporting of a prediction model study. The TRIPOD Statement aims to improve the transparency of the reporting of a prediction model study regardless of the study methods used. The TRIPOD Statement is best used in conjunction with the TRIPOD explanation and elaboration document. To aid the editorial process and readers of prediction model studies, it is recommended that authors include a completed checklist in their submission (also available at www.tripod-statement.org).

Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD Statement

Prediction models are developed to aid health care providers in estimating the probability or risk that a specific disease or condition is present (diagnostic models) or that a specific event will occur in the future (prognostic models), to inform their decision making. However, the overwhelming evidence shows that the quality of reporting of prediction model studies is poor. Only with full and clear reporting of information on all aspects of a prediction model can risk of bias and potential usefulness of prediction models be adequately assessed. The Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) Initiative developed a set of recommendations for the reporting of studies developing, validating, or updating a prediction model, whether for diagnostic or prognostic purposes. This article describes how the TRIPOD Statement was developed. An extensive list of items based on a review of the literature was created, which was reduced after a Web-based survey and revised during a 3-day meeting in June 2011 with methodologists, health care professionals, and journal editors. The list was refined during several meetings of the steering group and in e-mail discussions with the wider group of TRIPOD contributors. The resulting TRIPOD Statement is a checklist of 22 items, deemed essential for transparent reporting of a prediction model study. The TRIPOD Statement aims to improve the transparency of the reporting of a prediction model study regardless of the study methods used. The TRIPOD Statement is best used in conjunction with the TRIPOD explanation and elaboration document. To aid the editorial process and readers of prediction model studies, it is recommended that authors include a completed checklist in their submission (also available at www.tripod-statement.org).

A straightforward approach to designing a scoring system for predicting length-of-stay of cardiac surgery patients

Background

Length-of-stay prediction for cardiac surgery patients is a key point for medical management issues, such as optimization of resources in intensive care units and operating room scheduling. Scoring systems are a very attractive family of predictive models, but their retraining and updating are generally critical. The present approach to designing a scoring system for predicting length of stay in intensive care aims to overcome these difficulties, so that a model designed in a given scenario can easily be adjusted over time or for internal purposes.

Methods

A naïve Bayes approach was used to develop a simple scoring system. A set of 36 preoperative, intraoperative and postoperative variables collected in a sample of 3256 consecutive adult patients undergoing heart surgery were considered as likely risk predictors. The number of variables was reduced by selecting an optimal subset of features. Scoring system performance was assessed by cross-validation.

Results

After the selection process, seven variables were entered in the prediction model, which showed excellent discrimination, good generalization power and suitable sensitivity and specificity. No significant difference was found between AUC of the training and testing sets. The 95% confidence interval for AUC estimated by the BCa bootstrap method was [0.841, 0.883] and [0.837, 0.880] in the training and testing sets, respectively. Chronic dialysis, low postoperative cardiac output and acute myocardial infarction proved to be the major risk factors.

Conclusions

The proposed approach produced a simple and trustworthy scoring system, which is easy to update regularly and to customize for other centers. This is a crucial point when scoring systems are used as predictive models in clinical practice.

Critical appraisal and data extraction for systematic reviews of prediction modelling studies: the CHARMS checklist

Carl Moons and colleagues provide a checklist and background explanation for critically appraising and extracting data from systematic reviews of prognostic and diagnostic prediction modelling studies. Please see later in the article for the Editors' Summary.

EM for regularized zero-inflated regression models with applications to postoperative morbidity after cardiac surgery in children

This paper proposes a new statistical approach for predicting postoperative morbidity such as intensive care unit length of stay and number of complications after cardiac surgery in children. In a recent multi-center study sponsored by the National Institutes of Health, 311 children undergoing cardiac surgery were enrolled. Morbidity data are count data in which the observations take only nonnegative integer values. Often, the number of zeros in the sample cannot be accommodated properly by a simple model, thus requiring a more complex model such as the zero-inflated Poisson regression model. We are interested in identifying important risk factors for postoperative morbidity among many candidate predictors. There is only limited methodological work on variable selection for the zero-inflated regression models. In this paper, we consider regularized zero-inflated Poisson models through penalized likelihood function and develop a new expectation-maximization algorithm for numerical optimization. Simulation studies show that the proposed method has better performance than some competing methods. Using the proposed methods, we analyzed the postoperative morbidity, which improved the model fitting and identified important clinical and biomarker risk factors.

Postoperative delirium in a substudy of cardiothoracic surgical patients in the BAG-RECALL clinical trial

BACKGROUND

Postoperative delirium in the intensive care unit (ICU) is a frequent complication after cardiac or thoracic surgery and is associated with increased morbidity and mortality.

METHODS

In this single-center substudy of the BAG-RECALL trial (NCT00682825), we screened patients after cardiac or thoracic surgery in the ICU twice daily for delirium using the Confusion Assessment Method for the ICU. The primary outcome was the incidence of delirium in patients who had been randomized to intraoperative Bispectral Index (BIS)-guided and end-tidal anesthetic concentration-guided depth of anesthesia protocols. As a secondary analysis, a Bayesian stochastic search variable selection strategy was used to rank a field of candidate risk factors for delirium, followed by binary logistic regression.

RESULTS

Of 310 patients assessed, 28 of 149 (18.8%) in the BIS group and 45 of 161 (28.0%) in the end-tidal anesthetic concentration group developed postoperative delirium in the ICU (odds ratio 0.60, 95% confidence interval, 0.35-1.02, P= 0.058). Low average volatile anesthetic dose, intraoperative transfusion, ASA physical status, and European System for Cardiac Operative Risk Evaluation were identified as independent predictors of delirium.

CONCLUSIONS

A larger randomized study should determine whether brain monitoring with BIS or an alternative method decreases delirium after cardiac or thoracic surgery. The association between low anesthetic concentration and delirium is a surprising finding and could reflect that patients with poor health are both more sensitive to the effects of volatile anesthetic drugs and are also more likely to develop postoperative delirium. Investigation of candidate methods to prevent delirium should be prioritized in view of the established association between postoperative delirium and adverse patient outcomes.

Development of an open-heart intraoperative risk scoring model for predicting a prolonged intensive care unit stay

BACKGROUND

Based on a pilot study with 34 patients, applying the modified sequential organ failure assessment (SOFA) score intraoperatively could predict a prolonged ICU stay, albeit with only 4 risk factors. Our objective was to develop a practicable intraoperative model for predicting prolonged ICU stay which included more relevant risk factors.

METHODS

An extensive literature review identified 6 other intraoperative risk factors affecting prolonged ICU stay. Another 168 patients were then recruited for whom all 10 risk factors were extracted and analyzed by logistic regression to form the new prognostic model.

RESULTS

The multivariate logistic regression analysis retained only 6 significant risk factors in the model: age ≥ 60 years, PaO2/FiO2 ratio ≤ 200 mmHg, platelet count ≤ 120,000/mm(3), requirement for inotrope/vasopressor ≥ 2 drugs, serum potassium ≤ 3.2 mEq/L, and atrial fibrillation grading ≥ 2. This model was then simplified into the Open-Heart Intraoperative Risk (OHIR) score, comprising the same 6 risk factors for a total score of 7-a score of ≥ 3 indicating a likely prolonged ICU stay (AUC for ROC of 0.746).

CONCLUSIONS

We developed a new, easy to calculate OHIR scoring system for predicting prolonged ICU stay as early as 3 hours after CPB. It comprises 6 risk factors, 5 of which can be manipulated intraoperatively.

External validation of multivariable prediction models: a systematic review of methodological conduct and reporting

BACKGROUND

Before considering whether to use a multivariable (diagnostic or prognostic) prediction model, it is essential that its performance be evaluated in data that were not used to develop the model (referred to as external validation). We critically appraised the methodological conduct and reporting of external validation studies of multivariable prediction models.

METHODS

We conducted a systematic review of articles describing some form of external validation of one or more multivariable prediction models indexed in PubMed core clinical journals published in 2010. Study data were extracted in duplicate on design, sample size, handling of missing data, reference to the original study developing the prediction models and predictive performance measures.

RESULTS

11,826 articles were identified and 78 were included for full review, which described the evaluation of 120 prediction models. in participant data that were not used to develop the model. Thirty-three articles described both the development of a prediction model and an evaluation of its performance on a separate dataset, and 45 articles described only the evaluation of an existing published prediction model on another dataset. Fifty-seven percent of the prediction models were presented and evaluated as simplified scoring systems. Sixteen percent of articles failed to report the number of outcome events in the validation datasets. Fifty-four percent of studies made no explicit mention of missing data. Sixty-seven percent did not report evaluating model calibration whilst most studies evaluated model discrimination. It was often unclear whether the reported performance measures were for the full regression model or for the simplified models.

CONCLUSIONS

The vast majority of studies describing some form of external validation of a multivariable prediction model were poorly reported with key details frequently not presented. The validation studies were characterised by poor design, inappropriate handling and acknowledgement of missing data and one of the most key performance measures of prediction models i.e. calibration often omitted from the publication. It may therefore not be surprising that an overwhelming majority of developed prediction models are not used in practice, when there is a dearth of well-conducted and clearly reported (external validation) studies describing their performance on independent participant data.

Meta-analysis and aggregation of multiple published prediction models

Published clinical prediction models are often ignored during the development of novel prediction models despite similarities in populations and intended usage. The plethora of prediction models that arise from this practice may still perform poorly when applied in other populations. Incorporating prior evidence might improve the accuracy of prediction models and make them potentially better generalizable. Unfortunately, aggregation of prediction models is not straightforward, and methods to combine differently specified models are currently lacking. We propose two approaches for aggregating previously published prediction models when a validation dataset is available: model averaging and stacked regressions. These approaches yield user-friendly stand-alone models that are adjusted for the new validation data. Both approaches rely on weighting to account for model performance and between-study heterogeneity but adopt a different rationale (averaging versus combination) to combine the models. We illustrate their implementation in a clinical example and compare them with established methods for prediction modeling in a series of simulation studies. Results from the clinical datasets and simulation studies demonstrate that aggregation yields prediction models with better discrimination and calibration in a vast majority of scenarios, and results in equivalent performance (compared to developing a novel model from scratch) when validation datasets are relatively large. In conclusion, model aggregation is a promising strategy when several prediction models are available from the literature and a validation dataset is at hand. The aggregation methods do not require existing models to have similar predictors and can be applied when relatively few data are at hand.

Development of a nursing intervention to prepare frail older patients for cardiac surgery (the PREDOCS programme), following phase one of the guidelines of the Medical Research Council

Background:

In older patients undergoing elective cardiac surgery, the timely identification and preparation of patients at risk for frequent postoperative hospital complications provide opportunities to reduce the risk of these complications.

Aims:

We developed an evidence-based, multi-component nursing intervention (Prevention of Decline in Older Cardiac Surgery Patients; the PREDOCS programme) for application in the preadmission period to improve patients’ physical and psychosocial condition to reduce their risk of postoperative complications. This paper describes in detail the process used to design and develop this multi-component intervention.

Methods:

In a team of researchers, experts, cardiac surgeons, registered cardiac surgery nurses, and patients, the revised guidelines for developing and evaluating complex interventions of the Medical Research Council (MRC) were followed, including identifying existing evidence, identifying and developing theory and modelling the process and outcomes. Additionally, the criteria for reporting the development of complex interventions in healthcare (CReDECI) were followed.

Results:

The intervention is administered during a consultation by the nurse two to four weeks before the surgery procedure. The consultation includes three parts: a general part for all patients, a second part in which patients with an increased risk are identified, and a third part in which selected patients are informed about how to prepare themselves for the hospital admission to reduce their risk.

Conclusions:

Following the MRC guidelines, an extended, stepwise, multi-method procedure was used to develop the multi-component nursing intervention to prepare older patients for cardiac surgery, creating transparency in the assumed working mechanisms. Additionally, a detailed description of the intervention is provided.

Association between preoperative plasma sRAGE levels and recovery from cardiac surgery

Background. The receptor for advanced glycation end products (RAGE) is an inflammation-perpetuating receptor, and soluble RAGE (sRAGE) is a marker of cellular RAGE expression. This study investigated whether raised plasma levels prior to surgery of sRAGE or S100A8/A9 (a RAGE ligand) were associated with longer duration of hospital care in patients undergoing cardiac surgery necessitating cardiopulmonary bypass. Methods. Patients (n = 130) undergoing elective cardiac surgery were enrolled prospectively. Plasma sRAGE and S100A8/A9 concentrations were measured before and 2 h after surgery. Results. Preoperative plasma sRAGE increased significantly (P < 0.0001) from 1.06 ng/mL (IQR, 0.72–1.76) to 1.93 ng/mL (IQR, 1.14–2.63) 2 h postoperatively. Plasma S100A8/9 was also significantly (P < 0.0001) higher 2 h postoperatively (2.37 μg/mL, IQR, 1.81–3.05) compared to pre-operative levels (0.41 μg/mL, IQR, 0.2–0.65). Preoperative sRAGE, but not S100A8/A9, was positively and significantly correlated with duration of critical illness (r = 0.3, P = 0.0007) and length of hospital stay (LOS; r = 0.31, P < 0.0005). Multivariate binary logistic regression showed preoperative sRAGE to be, statistically, an independent predictor of greater than median duration of critical illness (odds ratio 16.6, P = 0.014) and to be, statistically, the strongest independent predictor of hospital LOS. Conclusion. Higher preoperative plasma sRAGE levels were associated with prolonged duration of care in adults undergoing cardiac surgery requiring cardiopulmonary bypass.

Prolonged intensive care treatment of octogenarians after cardiac surgery: a reasonable economic burden?

OBJECTIVES

In accordance with the rising prevalence of octogenarians undergoing cardiac surgery, these patients utilize an increasing portion of intensive care unit (ICU) capacities, provoking economic and ethical concerns. In this study, we evaluated the outcomes and costs generated by the prolonged postoperative ICU treatment of octogenarians.

METHODS

Between July 2009 and August 2010, 109 of 1063 patients required ICU treatment of at least 5 days after cardiac surgery. Patients were retrospectively assigned to either Group A (age <80, n = 86) or Group B (age ≥80, n = 23). Operative risk, mortality, length and costs of ICU treatment were analysed and compared. After 1 year, survival, quality of life (QOL) and functional status were assessed.

RESULTS

Hospital mortality was 31.4% in Group A and 56.5% in Group B. Survivals of discharged patients after 1 year were 83% (Group A) and 80% (Group B), respectively. Log EuroSCORE I of octogenarians was significantly higher (30 ± 17 vs 20 ± 16, P < 0.001). No significant differences (Group A vs Group B) were found between the groups concerning length of ICU treatment (20 ± 21 vs 16 ± 14 days, P = 0.577) or costs (27 205 ± 29 316€ vs 21 821 ± 16 259€, P = 0.812). Functional capacity, calculated by using Barthel index, was high (Group A: 87 ± 22 and Group B: 67 ± 31, P = 0.108) and did not differ significantly between groups. QOL, measured with the short form-12 health survey, did not differ significantly between groups (physical health summary score: P = 0.27; mental health score: P = 0.885) and was comparable with values of the age-adjusted general population.

CONCLUSIONS

Presented data propose that advanced age is correlated with a higher mortality, but not with prolonged ICU treatment or higher costs after cardiac surgery. Considering the encouraging functional status and QOL of the survivors, the financial burden caused by octogenarians is justified.

Very prolonged stay in the intensive care unit after cardiac operations: early results and late survival

BACKGROUND

Prolonged intensive care unit (ICU) stay is a surrogate for advanced morbidity or perioperative complications, and resource utilization may become an issue. It is our policy to continue full life support in the ICU, even for patients with a seemingly grim outlook. We examined the effect of duration of ICU stay on early outcomes and late survival.

METHODS

Between 1993 and 2011, 6,385 patients were admitted to the ICU after cardiac surgery. Patients were grouped according to length of stay in the ICU: group 1, 2 days or less (n = 4,631; 73%); group 2, 3 to 14 days (n = 1,423; 22%); group 3, more than 14 days (n = 331; 5%). Length of stay in ICU for group 3 patients was 38 ± 24 days (range, 15 to 160; median 31). Clinical profile and outcomes were compared between groups.

RESULTS

Patients requiring prolonged ICU stay were older, underwent more complex surgery, had greater comorbidity, and a higher predicted operative mortality (p < 0.0001). They had a higher incidence of adverse events and increased mortality (p < 0.0001). Of the 331 group 3 patients, 60% were discharged: survival of these patients at 1, 3, and 5 years was 78%, 65%, and 52%, respectively. Operative mortality as well as late survival of discharged patients was proportional to duration of ICU stay.

CONCLUSIONS

Current technology enables keeping sick patients alive for extended periods of time. Nearly two thirds of patients requiring prolonged ICU leave hospital, and of these, 50% attain 5-year survival. These data support offering full and continued support even for patients requiring very prolonged ICU stay.

Preoperative endogenous ouabain predicts acute kidney injury in cardiac surgery patients

OBJECTIVES

Acute kidney injury is a frequent complication of cardiac surgery and increases morbidity and mortality. As preoperative biomarkers predicting the development of acute kidney injury are not available, we have tested the hypothesis that preoperative plasma levels of endogenous ouabain may function as this type of biomarker.

RATIONALE AND DESIGN

Endogenous ouabain is an adrenal stress hormone associated with adverse cardiovascular outcomes. Its involvement in acute kidney injury is unknown. With studies in patients and animal settings, including isolated podocytes, we tested the above mentioned hypothesis.

PATIENTS

Preoperative endogenous ouabain was measured in 407 patients admitted for elective cardiac surgery and in a validation population of 219 other patients. We also studied the effect of prolonged elevations of circulating exogenous ouabain on renal parameters in rats and the influence of ouabain on podocyte proteins both "in vivo" and "in vitro."

MAIN RESULTS

In the first group of patients, acute kidney injury (2.8%, 8.3%, 20.3%, p < 0.001) and ICU stay (1.4±0.38, 1.7±0.41, 2.4±0.59 days, p = 0.014) increased with each incremental preoperative endogenous ouabain tertile. In a linear regression analysis, the circulating endogenous ouabain value before surgery was the strongest predictor of acute kidney injury. In the validation cohort, acute kidney injury (0%, 5.9%, 8.2%, p < 0.0001) and ICU stay (1.2±0.09, 1.4±0.23, 2.2±0.77 days, p = 0.003) increased with the preoperative endogenous ouabain tertile. Values for preoperative endogenous ouabain significantly improved (area under curve: 0.85) risk prediction over the clinical score alone as measured by integrate discrimination improvement and net reclassification improvement. Finally, in the rat model, elevated circulating ouabain reduced creatinine clearance (-18%, p < 0.05), increased urinary protein excretion (+ 54%, p < 0.05), and reduced expression of podocyte nephrin (-29%, p < 0.01). This last finding was replicated ex vivo by incubating podocyte primary cell cultures with low-dose ouabain.

CONCLUSIONS

Preoperative plasma endogenous ouabain levels are powerful biomarkers of acute kidney injury and postoperative complications and may be a direct cause of podocyte damage.

Critical care management of patients following transcatheter aortic valve replacement

Transcatheter aortic valve replacement (TAVR) is rapidly gaining popularity as a technique to surgically manage aortic stenosis (AS) in high risk patients. TAVR is significantly less invasive than the traditional approach to aortic valve replacement via median sternotomy. Patients undergoing TAVR often suffer from multiple comorbidities, and their postoperative course may be complicated by a unique set of complications that may become evident in the intensive care unit (ICU). In this article, we review the common complications of TAVR that may be observed in the ICU, and different strategies for their management.

Customized Prediction of Short Length of Stay Following Elective Cardiac Surgery in Elderly Patients Using a Genetic Algorithm

Objective

To develop a customized short LOS (<6 days) prediction model for geriatric patients receiving cardiac surgery, using local data and a computational feature selection algorithm.

Design

Utilization of a machine learning algorithm in a prospectively collected STS database consisting of patients who received cardiac surgery between January 2002 and June 2011.

Setting

Urban tertiary-care center.

Participants

Geriatric patients aged 70 years or older at the time of cardiac surgery.

Interventions

None.

Measurements and Main Results

Predefined morbidity and mortality events were collected from the STS database. 23 clinically relevant predictors were investigated for short LOS prediction with a genetic algorithm (GenAlg) in 1426 patients. Due to the absence of an STS model for their particular surgery type, STS risk scores were unavailable for 771 patients. STS prediction achieved an AUC of 0.629 while the GenAlg achieved AUCs of 0.573 (in those with STS scores) and 0.691 (in those without STS scores). Among the patients with STS scores, the GenAlg features significantly associated with shorter LOS were absence of congestive heart failure (CHF) (OR = 0.59, p = 0.04), aortic valve procedure (OR = 1.54, p = 0.04), and shorter cross clamp time (OR = 0.99, p = 0.004). In those without STS prediction, short LOS was significantly correlated with younger age (OR = 0.93, p < 0.001), absence of CHF (OR = 0.53, p = 0.007), no preoperative use of beta blockers (OR = 0.66, p = 0.03), and shorter cross clamp time (OR = 0.99, p < 0.001).

Conclusion

While the GenAlg-based models did not outperform STS prediction for patients with STS risk scores, our local-data-driven approach reliably predicted short LOS for cardiac surgery types that do not allow STS risk calculation. We advocate that each institution with sufficient observational data should build their own cardiac surgery risk models.

Fast-track practice in cardiac surgery: results and predictors of outcome

OBJECTIVES

Various studies have shown different parameters as independent risk factors in predicting the success of fast-track postoperative management in cardiac surgery. In the present study, we evaluated our 7-year experience with the fast-track protocol and investigated the preoperative predictors of successful outcome.

METHODS

Between 2004 and 2010, 5367 consecutive patients undergoing cardiac surgery were preoperatively selected for postoperative admission in the postanaesthesia care unit (PACU) and were included in this study. These patients were then transferred to the ordinary ward on the same day of the operation. The primary end-point of the study was the success of the PACU protocol, defined as discharge to the ward on the same day, no further admission to the intensive care unit and no operative mortality. Logistic regression analysis was performed to detect the independent risk factors for failure of the PACU pathway.

RESULTS

Of 11,895 patients undergoing cardiac surgery, 5367 (45.2%) were postoperatively admitted to the PACU. The protocol was successful in 4510 patients (84.0%). Using the multivariate logistic regression analysis, older age and left ventricular dysfunction were found to be independent risk factors for failure of the PACU protocol [odds ratio of 0.98/year (0.97-0.98) and 0.31 (0.14-0.70), respectively].

CONCLUSIONS

Our fast-track management, called the PACU protocol, is efficient and safe for the postoperative management of selected patients undergoing cardiac surgery. Age and left ventricular dysfunction are significant preoperative predictors of failure of this protocol.

Fracture risk assessment: state of the art, methodologically unsound, or poorly reported?

Osteoporotic fractures, including hip fractures, are a global health concern associated with significant morbidity and mortality as well as a major economic burden. Identifying individuals who are at an increased risk of osteoporotic fracture is an important challenge to be resolved. Recently, multivariable prediction tools have been developed to assist clinicians in the management of their patients by calculating their 10-year risk of fracture (FRAX, QFracture, Garvan) using a combination of known risk factors. These prediction models have revolutionized the way clinicians assess the risk of fracture. Studies evaluating the performance of prediction models in this and other areas of medicine have, however, been characterized by poor design, methodological conduct, and reporting. We examine recently developed fracture prediction models and critically discuss issues in their design, validation, and transparency.

Length of intensive care unit stay following cardiac surgery: is it impossible to find a universal prediction model?

OBJECTIVES

Accurate models for prediction of a prolonged intensive care unit (ICU) stay following cardiac surgery may be developed using Cox proportional hazards regression. Our aims were to develop a preoperative and intraoperative model to predict the length of the ICU stay and to compare our models with published risk models, including the EuroSCORE II.

METHODS

Models were developed using data from all patients undergoing cardiac surgery at St. Olavs Hospital, Trondheim, Norway from 2000-2007 (n = 4994). Internal validation and calibration were performed by bootstrapping. Discrimination was assessed by areas under the receiver operating characteristics curves and calibration for the published logistic regression models with the Hosmer-Lemeshow test.

RESULTS

Despite a diverse risk profile, 93.7% of the patients had an ICU stay <2 days, in keeping with our fast-track regimen. Our models showed good calibration and excellent discrimination for prediction of a prolonged stay of more than 2, 5 or 7 days. Discrimination by the EuroSCORE II and other published models was good, but calibration was poor (Hosmer-Lemeshow test: P < 0.0001), probably due to the short ICU stays of almost all our patients. None of the models were useful for prediction of ICU stay in individual patients because most patients in all risk categories of all models had short ICU stays (75th percentiles: 1 day).

CONCLUSIONS

A universal model for prediction of ICU stay may be difficult to develop, as the distribution of length of stay may depend on both medical factors and institutional policies governing ICU discharge.

Predictors of Prolonged Stay in the Intensive Care Unit following Cardiac Surgery

The prediction of intensive care unit length of stay (ICU-LOS) could contribute to more efficient ICU resources' allocation and better planning of care among cardiac surgery patients. The aim of this study was to identify the preoperative and intraoperative predictors for prolonged cardiac surgery ICU-LOS. An observational cohort study was conducted among 150 consecutive patients, who were admitted to the cardiac surgery ICU of a tertiary hospital of Athens, Greece from September 2010 to January 2011. Multivariate regression analysis revealed that patients with increased creatinine levels preoperatively (odds ratio (OR) 3.0, P = 0.049), history of atrial fibrillation (AF) (OR 6.3, P = 0.012) and high EuroSCORE values (OR 2.6, P = 0.017) had a significant greater probability to stay in the ICU for more than 2 days. In addition, intraoperative hyperglycemia (OR 3.0, P = 0.004) was strongly associated with longer ICU-LOS. In conclusion, the high perioperative risk, the history of AF and renal dysfunction, and the intraoperative hyperglycemia are significant predictors of prolonged ICU stay. The early identification of patients at risk could allow the efficient ICU resources' allocation and the reduction of healthcare costs. This would contribute to nursing care planning depending on the availability of healthcare personnel and ICU bed capacity.

Independent risk factors for cardiac operations in adults with congenital heart disease: a retrospective study of 543 operations for 500 patients

Adults with congenital heart disease (CHD) are an increasing population requiring cardiac operations. To date, the perioperative risk factors for this group have not been identified. This study aimed to identify clinical, morphologic, and hemodynamic risk factors for an adverse outcome. This study retrospectively analyzed a cohort of 500 patients (ages >16 years) who underwent 543 operations between January 2004 and December 2008 at a single center. The composite end point of an adverse outcome was in-hospital death, a prolonged intensive care exceeding 4 days, or both. The composite end point was reached by 253 of the patients (50.6%). Of the 500 patients, 13 (2.6%) died within 30 days after the operation. After logistic regression analysis, the following eight items remained significant: male gender (P = 0.003; odds ratio [OR] 1.8; 95% confidence interval [CI] 1.2-2.6), cyanosis (P > 0.006; OR 3.7; 95% CI 1.5-9.4), functional class exceeding 2 (P = 0.004; OR 2.2; 95% CI 1.3-3.7), chromosomal abnormalities (P = 0.004; OR 3.3; 95% CI 1.4-7.7), impaired renal function (P = 0.019; OR 3.8; 95% CI 1.2-11.5), systemic right ventricle (RV) in a biventricular circulation (P = 0.027; OR 3.3; 95% CI 1.1-9.5), enlargement of the systemic ventricle (P = 0.011; OR 1.7; 95% CI 1.1-2.6), and operation with extracorporeal circulation (P = 0.002; OR 4.3; 95% CI 1.7-11.4). Early mortality in the current adult CHD population is low. Morbidity, however, is significant and influenced by the patients' conditions (male gender, chromosomal abnormalities), history (cyanosis, New York Hospital Association [NYHA] class), and underlying morphology (systemic RV). This information for a large cohort of patients could help progress toward more adequate counseling for adults with a congenital heart defect.

Clinical outcomes in patients with prolonged intensive care unit length of stay after cardiac surgical procedures

BACKGROUND

Advances in critical care medicine have allowed for improved care of patients requiring prolonged intensive care unit length of stay (prICULOS) after cardiac operations, yet little is known regarding their eventual outcomes. The purpose of this study was to examine short- and long-term outcomes in patients undergoing cardiac operations with prICULOS.

METHODS

All cases of coronary artery bypass grafting (CABG), aortic valve, mitral valve, and combined CABG/valve surgical procedures performed at a single institution from July 2002 to July 2007 were identified. All-cause mortality in patients discharged alive from the hospital was determined until December 2007 through linkage with the Social Security Death Index. Patients who experienced intraoperative death or those with missing or invalid social security numbers were excluded. The definition of prICULOS was total ICULOS greater than 7 days.

RESULTS

A total of 3,478 patients met inclusion criteria. One hundred thirty-seven of three thousand four hundred seventy-eight patients (3.9%) experienced prICULOS. These patients were more likely to be older than 70 years (55.5% versus 30.5%; p<0.0001) and to have had recent myocardial infarction (28.5% versus 20.1%; p=0.02), previous cardiac operation (18.3% versus 6.9%; p<0.0001), and emergent status (9.5% versus 1.6%; p<0.0001). They experienced greater in-hospital mortality (37.2% versus 1.7%; p<0.0001) and those who were discharged alive had worse long-term survival (log-rank, p<0.0001). After risk adjustment, prICULOS emerged as a significant predictor of in-hospital death (odds ratio [OR] 20.9; 95% confidence interval [CI], 12.9-33.7) and decreased long-term survival (hazard ratio [HR] 2.9; 95% CI, 2.0-4.3).

CONCLUSIONS

Patients with prICULOS after cardiac operations have worse overall outcomes. These data may be used to inform these patients and their families of realistic expectations regarding their clinical course.

Prolonged ventilation post cardiac surgery--tips and pitfalls of the prediction game

Background

Few available models aim to identify patients at risk of prolonged ventilation after cardiac surgery. We compared prediction models developed in ICU in two adjacent periods of time, when significant changes were observed both in population characteristics and the perioperative management.

Methods

We performed a retrospective review of two cohorts of patients in our department in two subsequent time periods (July 2007 - December 2008, n = 2165; January 2009 - July 2010, n = 2192). The study was approved by the Institutional Ethics Committee and the individual patient consent was not required. Patients were divided with regard to ventilation time of more or less than 48 hours. Preoperative and procedure-related variables for prolonged ventilation were identified and multivariate logistic regression analysis was performed separately for each cohort.

Results

Most recent patients were older, with more co-morbidities, more frequently undergoing off-pump surgery. At the beginning of 2009 we also changed the technique of postoperative ventilation. Percentage of patients with prolonged ventilation decreased from 5.7% to 2.4% (p < 0.0001).Preoperative and procedure-related variables for prolonged ventilation were identified. Prediction models for prolonged ventilation were different for each cohort. Most recent significant predictors were: aortic aneurysm surgery (OR 12.9), emergency surgery (OR 5.3), combined procedures (OR 5.1), valve procedures (OR 3.2), preoperative renal dysfunction (OR 2.9) and preoperative stroke or TIA (OR 2.8).

Conclusions

Prediction models for postoperative ventilation should be regularly updated, particularly when major changes are noted in patients' demographics and surgical or anaesthetic technique.

Computerized prediction of intensive care unit discharge after cardiac surgery: development and validation of a Gaussian processes model

Background

The intensive care unit (ICU) length of stay (LOS) of patients undergoing cardiac surgery may vary considerably, and is often difficult to predict within the first hours after admission. The early clinical evolution of a cardiac surgery patient might be predictive for his LOS. The purpose of the present study was to develop a predictive model for ICU discharge after non-emergency cardiac surgery, by analyzing the first 4 hours of data in the computerized medical record of these patients with Gaussian processes (GP), a machine learning technique.

Methods

Non-interventional study. Predictive modeling, separate development (n = 461) and validation (n = 499) cohort. GP models were developed to predict the probability of ICU discharge the day after surgery (classification task), and to predict the day of ICU discharge as a discrete variable (regression task). GP predictions were compared with predictions by EuroSCORE, nurses and physicians. The classification task was evaluated using aROC for discrimination, and Brier Score, Brier Score Scaled, and Hosmer-Lemeshow test for calibration. The regression task was evaluated by comparing median actual and predicted discharge, loss penalty function (LPF) ((actual-predicted)/actual) and calculating root mean squared relative errors (RMSRE).

Results

Median (P25-P75) ICU length of stay was 3 (2-5) days. For classification, the GP model showed an aROC of 0.758 which was significantly higher than the predictions by nurses, but not better than EuroSCORE and physicians. The GP had the best calibration, with a Brier Score of 0.179 and Hosmer-Lemeshow p-value of 0.382. For regression, GP had the highest proportion of patients with a correctly predicted day of discharge (40%), which was significantly better than the EuroSCORE (p < 0.001) and nurses (p = 0.044) but equivalent to physicians. GP had the lowest RMSRE (0.408) of all predictive models.

Conclusions

A GP model that uses PDMS data of the first 4 hours after admission in the ICU of scheduled adult cardiac surgery patients was able to predict discharge from the ICU as a classification as well as a regression task. The GP model demonstrated a significantly better discriminative power than the EuroSCORE and the ICU nurses, and at least as good as predictions done by ICU physicians. The GP model was the only well calibrated model.

Predicting prolonged intensive care unit stays in older cardiac surgery patients: a validation study

PURPOSE

In cardiac surgery prediction models identifying patients at risk of prolonged stay at the Intensive Care Unit (ICU) are used to optimize treatment and use of ICU resources. A recent systematic validation study of 14 of these models identified three models with a good predictive performance across patients of all ages. It is however unclear how these models perform in older patients, who nowadays form a considerable part of this patient population. The current study specifically validates the performance of these three models in older cardiac surgery patients and quantifies how their performance changes with increasing age of patients.

METHODS

The Parsonnet model, the EuroSCORE, and a model by Huijskes and colleagues were validated using prospectively collected data of 11,395 cardiac surgery patients. Performance of the models was described by discrimination (area under the ROC curve, AUC) and calibration.

RESULTS

For the Parsonnet model, the EuroSCORE and the Huijskes model discrimination clearly decreased with increasing age (AUCs of 0.76, 0.71 and 0.72 for ages 70-75 and 0.72, 0.70 and 0.72, respectively, for ages 75-80 and 0.68, 0.64 and 0.69, respectively, above 80 years). The models showed poor calibration in patients aged >70 (p values for fit of the models <0.006).

CONCLUSIONS

To optimize treatment and ICU resources, risk prediction for prolonged ICU stay after cardiac surgery using the existing models should be done with great care for older patients.