Secondary Logo

Journal Logo

Development and validation of a score to predict postoperative respiratory failure in a multicentre European cohort

A prospective, observational study

Canet, Jaume; Sabaté, Sergi; Mazo, Valentín; Gallart, Lluís; de Abreu, Marcelo Gama; Belda, Javier; Langeron, Olivier; Hoeft, Andreas; Pelosi, Paolo For the PERISCOPE group

European Journal of Anaesthesiology (EJA): July 2015 - Volume 32 - Issue 7 - p 458–470
doi: 10.1097/EJA.0000000000000223
Complications and errors

BACKGROUND Postoperative respiratory failure (PRF) is the most frequent respiratory complication following surgery.

OBJECTIVE The objective of this study was to build a clinically useful predictive model for the development of PRF.

DESIGN A prospective observational study of a multicentre cohort.

SETTING Sixty-three hospitals across Europe.

PATIENTS Patients undergoing any surgical procedure under general or regional anaesthesia during 7-day recruitment periods.

MAIN OUTCOME MEASURES Development of PRF within 5 days of surgery. PRF was defined by a partial pressure of oxygen in arterial blood (PaO2) less than 8 kPa or new onset oxyhaemoglobin saturation measured by pulse oximetry (SpO2) less than 90% whilst breathing room air that required conventional oxygen therapy, noninvasive or invasive mechanical ventilation.

RESULTS PRF developed in 224 patients (4.2% of the 5384 patients studied). In-hospital mortality [95% confidence interval (95% CI)] was higher in patients who developed PRF [10.3% (6.3 to 14.3) vs. 0.4% (0.2 to 0.6)]. Regression modelling identified a predictive PRF score that includes seven independent risk factors: low preoperative SpO2; at least one preoperative respiratory symptom; preoperative chronic liver disease; history of congestive heart failure; open intrathoracic or upper abdominal surgery; surgical procedure lasting at least 2 h; and emergency surgery. The area under the receiver operating characteristic curve (c-statistic) was 0.82 (95% CI 0.79 to 0.85) and the Hosmer–Lemeshow goodness-of-fit statistic was 7.08 (P = 0.253).

CONCLUSION A risk score based on seven objective, easily assessed factors was able to predict which patients would develop PRF. The score could potentially facilitate preoperative risk assessment and management and provide a basis for testing interventions to improve outcomes.

The study was registered at (identifier NCT01346709).

From the Department of Anaesthesiology and Postoperative Care Unit, Hospital Universitari Germans Trias i Pujol (JC, VM), Department of Anaesthesiology, Fundació, Puigvert (SS), Department of Anaesthesiology, Hospital del Mar, IMIM (Institut Hospital del Mar d’ Investigacions Mèdiques), Universitat Autònoma de Barcelona, Barcelona, Spain (LG), Department of Anaesthesiology and Intensive Care Medicine, University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany (MGDA), Department of Anaesthesia and Critical Care, Hospital Clínico Universitario, University of Valencia, Valencia, Spain (JB), Department of Anaesthesiology and Critical Care, Université, Pierre et Marie Curie-Paris VI, CHU Pitié-Salpêtrière, Paris, France (OL), Department of Anaesthesiology and Intensive Care Medicine, University of Bonn, Bonn, Germany (AH), and Department of Surgical Sciences and Integrated Diagnostics, IRCCS San Martino Hospital-IST, University of Genoa, Genoa, Italy (PP)

*Members of the PERISCOPE (Prospective Evaluation of a RIsk Score for postoperative pulmonary COmPlications in Europe) group are listed in the Appendix.

Correspondence to Jaume Canet, Department of Anaesthesiology and Postoperative Care Unit, Hospital Universitari Germans Trias i Pujol, Badalona, 08916 Barcelona, Spain Tel: +34690953029; fax: +34934978749; e-mail:

Published online 13 February 2015

Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal's Website (

This article is accompanied by the following Invited Commentary:

Staehr-Rye AK, Eikermann M. Eliminate postoperative respiratory complications: preoperative screening opens the door to clinical pathways that individualise perioperative treatment. Eur J Anaesthesiol 2015; 32:455–457.

Back to Top | Article Outline


Postoperative respiratory failure (PRF) is the most frequent postoperative pulmonary complication (PPC) and has a major impact on outcome and health costs.1–7 The pathogenesis of PRF depends on factors related to patient status as well as anaesthetic and surgical procedure.8–10 The incidence of PRF in general surgical populations ranges between 0.2 and 3.4%8 and several scoring systems for predicting PRF have been proposed.1,3–7,11 However, previous studies developing scores to predict PRF have defined this complication in different ways. Definitions that have been used include unexpected tracheal reintubation,1,5,7,11 the need for postoperative mechanical ventilation1,3 or postoperative acute lung injury (ALI) and acute respiratory distress syndrome (ARDS).4,6 In addition, the majority of the available scoring systems have been developed from retrospective databases that contain administrative information and coding.1,3,5–7,11 Retrospectively identified predictors have certain limitations,12–15 including low positive predictive values and moderate reliability, and they are subject to errors in data collection, higher percentages of missing values and the lack of information on variables of clinical interest.

Current thinking on the diagnosis of PRF calls for the use of objective measures of newly developing hypoxaemia detected during the postoperative course,8 specifically a partial pressure of oxygen in arterial blood (PaO2) less than 8 kPa (60 mmHg), which usually corresponds to an arterial oxygen saturation less than 90%. Furthermore, according to the most recent international consensus on ARDS, the severity of PRF may be further classified as mild, moderate or severe based on the ratio of PaO2 to the inspiratory oxygen fraction (FIO2).16 Stratifying risk for different degrees of PRF severity could potentially facilitate the early detection and management of this complication.

In this study, we used a large European database of general surgical cases (PERISCOPE cohort – Prospective Evaluation of a RIsk Score for postoperative pulmonary COmPlications in Europe)17 that had been created to externally validate the ARISCAT risk score2 for a PPC composite. Hypothesising that it would be possible to use the PERISCOPE data to build a simple risk score to predict PRF alone, we designed the present secondary analysis. Our aims were to identify perioperative risk factors for PRF and build and internally validate a specific predictive model. We also stratified PRF at three levels of severity on the basis of the presence of hypoxaemia and type of respiratory support in order to assess differences in outcome.

Back to Top | Article Outline

Materials and methods

Study design

A cohort of surgical patients was created for the observational multicentre PERISCOPE study. Sixty-three European hospitals (Appendix) recruited patients during continuous 7-day periods, choosing a convenient date to begin data collection between 2 May and 15 August 2011. Follow-up ended in November 2011. The participating hospitals constituted a convenience sample of volunteer centres found through the European Society of Anaesthesiology (ESA). Candidates were approached directly by national study coordinators. The study was registered at (identifier NCT01346709).

Back to Top | Article Outline

PERISCOPE cohort inclusion and exclusion criteria

Consecutive patients undergoing in-hospital elective or emergency surgery under general (including combined general anaesthesia) or regional (neuroaxial or plexus block) anaesthesia were recruited.

Exclusion criteria were as follows: age under 18 years; obstetric procedures or any procedure during pregnancy; procedures in which only local or peripheral nerve anaesthesia would be used; procedures outside an operating theatre; procedures related to a previous postoperative complication; organ transplantation; patients who had undergone tracheal intubation preoperatively; and outpatient procedures, defined as those requiring a hospital stay less than 24 h.

Back to Top | Article Outline

Ethical considerations

Ethical requirements differed in the 21 countries, but formal approval from a research ethics review board was applied for and given in each. The locally responsible investigator applied for and obtained approval from the ethics committee of each participating hospital. Written informed consent was obtained from each patient.

Back to Top | Article Outline

Organisation, data collection and quality assurance

The research team consisted of a steering committee in addition to nationally and locally responsible investigators, who were all anaesthesiologists. Data collectors, who did not modify a centre's customary management of patients, used a structured questionnaire to record the following information: administrative data [dates of surgery and discharge; status (alive or dead) at discharge], general information (sex, date of birth date, height and weight), preoperative variables [oxyhaemoglobin saturation measured by pulse oximetry (SpO2) breathing air in supine position after 1 min resting breathing air, or in patients on oxygen, SpO2 after 10 min without oxygen]; respiratory symptoms based on a simplified version of the Medical Research Council questionnaire;18 respiratory infection in the last month; haemoglobin concentration; cough test; chronic pulmonary disease; smoking status; American Society of Anesthesiologists ASA class; and intraoperative variables [surgical incision, surgical duration in hours, type of surgery (scheduled or emergent), description of procedure, surgical specialty and anaesthetic technique]. Definitions of all variables can be found in the online supplement (Supplementary Table 1,

The data collectors also sought all PPCs by searching medical records daily to find relevant events until hospital discharge; information on PRF was, therefore, recorded, as this complication developed throughout the hospital stay. Data were collected on paper forms and then transferred anonymously to secure online case records (OpenClinica, Boston, Massachusetts, USA). This electronic system incorporated quality control algorithms to validate online data entry and identify missing data. An off-site data manager checked entries to confirm completeness and asked the local team contact to provide additional information if necessary. An expert on the International Classification of Diseases (Ninth Revision, Clinical Modification) coded all diagnoses and procedures at the end of the collection period.

Back to Top | Article Outline


The primary outcome of interest for this secondary analysis was PRF, which was defined as new-onset hypoxaemia appearing within 5 postoperative days at three levels of severity: mild (PaO2 <8 kPa or SpO2 <90% on room air but responding to mask/nasal supplemental oxygen); moderate (necessitating noninvasive or invasive mechanical ventilation to treat a PaO2 <8 kPa or SpO2 <90%); or severe [requiring invasive mechanical ventilation to manage a PaO2/FiO2 <26.7 kPa (200 mmHg) regardless of the level of positive end-expiratory pressure (PEEP)]. Hypoventilation and heart failure were excluded in all cases. Hypoventilation considered likely to be due to residual effects of anaesthetics or opiates was evaluated clinically by the investigators, and heart failure was defined as signs of diffuse alveolar interstitial infiltrates with dyspnoea and rates related to left ventricular failure confirmed by one of the following: echocardiography; pulmonary artery catheter monitoring; or clinical improvement with specific treatment.

Secondary outcomes of interest were postoperative ICU admission, postoperative length of stay (LOS) and in-hospital mortality.

Back to Top | Article Outline

Statistical analysis

The size of the PERISCOPE cohort had been calculated to provide at least 10 events per variable that we expected to enter into the logistic regression model.19 It was estimated that the 63 PERISCOPE centres would be able to collect around 5000 cases and that the incidence of PRF would be around 3%.1,2,20,21 Recording at least 150 PRF events would allow around 15 predictor variables to be entered into logistic regression. Demographic and clinical characteristics are expressed in percentages and median (interquartile range, IQR).

Potential PRF predictors were selected according to the investigators’ consensus on measurable preoperative variables or the results of previous studies.2,22 Independent continuous variables (age, SpO2 and duration of surgery) were grouped into categories on the basis of the investigators’ understanding of relevant clinical cut points.

To compare patients with and without PRF, all categorical variables were analysed with the Chi-square test or the Fisher exact test, as appropriate, for associations with the outcome. Bivariate odds ratios (ORs) and 95% confidence intervals (95% CIs) were also estimated. The possibility of colinearity between categorical variables was tested with the Cramer V test (nominal variables) or Kendall's tau-b (ordinal variables).

The logistic regression model was constructed using a backward stepwise selection procedure in which the presence of PRF was the dependent variable. Independent predictors were entered into the model if a significant association (P < 0.05) was identified on bivariate analysis and the correlation coefficient between them (colinearity) was less than 0.25. Potential predictors were removed if this exclusion did not result in a significant change in the log-likelihood ratio test. The cut-off for variable removal was set at a significance level of 0.05. Adjusted ORs and 95% CIs were also calculated.

To avoid overfitting and to obtain reliable internal validation of the subset of factors, we used a bootstrap method,23 deriving 1000 computer-generated samples by random selection with replacement, each including the same number of patients. Within each bootstrap sample, the β coefficient was calculated using all selected independent variables. The robustness of the model and, thus, the reliability of predictor variables in the final regression model were estimated by the 95% CI of the β coefficient derived from the bootstrap samples.

A simplified predictive risk score for clinical use was then calculated by multiplying each β coefficient (corrected after bootstrapping) by 10 and rounding to the nearest integer. The integers were added together to produce an overall PRF risk score for each patient. To evaluate the ability of the score to predict increasing PRF risk, we used the minimum description length principle24 to divide the sample into three risk levels, each with a similar number of patients. The logistic regression model's calibration was then assessed by the Hosmer–Lemeshow goodness-of-fit statistic and by plotting the actual frequency of PRF in each of the three risk levels against the predicted probability of PRF in that risk group.

To assess the ability of the simplified PRF risk score to discriminate between patients with and without PRF, we used the c-statistic, which was also displayed graphically as the area under the receiver operating characteristic (ROC) curve. In addition, to check the performance of the model if it were used without information for any single factor such as SpO2 (which might not be recorded in all centres), we also checked the discriminative performance by calculating the c-statistics and calibration statistics for alternative six-factor models.

The Mann–Whitney U test was used to compare postoperative LOS between patients with and without PRF. An actuarial life table was constructed to assess in-hospital mortality after development of mild, moderate or severe PRF. The Wilcoxon–Gehan test was used to compare overall survival curves.

Statistical analyses were performed using the SPSS software package (version 20.0; IBM Corp., Armonk, New York, USA). Bootstrapping was performed using R, version 3.0.2 (R Project for Statistical Computing).

Back to Top | Article Outline


Of 5859 initially eligible patients, 5384 (91.9%) were included in the final analysis (Fig. 1). The characteristics of patients and procedures are detailed in Table 1.

Fig. 1

Fig. 1

Table 1

Table 1

PRF developed in 224 patients (4.2% of the cohort) and was classified as mild in 155 (2.9%), moderate in 43 (0.8%) and severe in 26 (0.5%). The time between surgery and the onset of PRF was a median of 0.5 days (0 to 1). In 54.9% of the patients with PRF, symptoms began within 24 h and in 94.6% onset was within 3 days.

Back to Top | Article Outline

Postoperative respiratory failure, ICU stay, postoperative length of stay and mortality

Intensive care admission was required in 181 (80.8%) of the patients who developed PRF and in 318 (6.2%) of the patients who did not. The ICU stay was significantly longer in patients who developed PRF (P < 0.001). These patients were in the ICU a median of 44 (24 to 96.5) h, whereas the median stay for patients without PRF was 22 (12 to 46) h.

The median in-hospital postoperative stay was also longer in patients with PRF [9 (5 to 14) vs. 4 (2 to 7) days] (P < 0.001). Forty-six patients died in the hospital; 23 of them had PRF (10.3% of the 224 patients with PRF) and 23 did not (0.44% of the 5160 without PRF) (P < 0.001). Figure 2 shows survival curves for in-hospital mortality according to PRF severity. Differences in hospital mortality between PRF severity levels were statistically significant (P < 0.001).

Fig. 2

Fig. 2

Back to Top | Article Outline

Risk factors and postoperative respiratory failure score

The independent variables entered into logistic regression are summarised in Table 2, along with variables that were not significant on bivariate analysis or that were significant but rejected because of high colinearity with other variables. Multivariable logistic regression selected seven independent predictors of PRF, four were related to the patient's presurgical health status (low preoperative SpO2 breathing air, respiratory symptoms, heart failure and chronic liver disease) and three were procedure-related (open thoracic or abdominal surgery, duration of surgery, and emergency surgery). All were retained in more than 95% of the bootstrap subsamples. Table 3 summarises the ORs for these predictors. The seven-variable regression model had good discrimination (c-statistic 0.82) and calibration (Hosmer–Lemeshow, P = 0.253). The area under the ROC curve (c-statistic) and the calibration plot are presented in Fig. 3. Supplementary Table 2, shows the statistics reflecting the performance of the model without inclusion of preoperative SpO2 or any other single factor; the c-statistic fell to 0.81 for that model and all other alternative six-variable models created by removing one of the factors.

Table 2

Table 2

Table 2

Table 2

Table 3

Table 3

Fig. 3

Fig. 3

The incidence of PRF increased significantly between risk levels (low <12 points; intermediate 12 to 22 points; and high ≥23 points). The incidences (95% CIs) were 1.1% (0.7 to 1.5), 4.6% (3.4 to 5.6) and 18.8% (15.8 to 21.8), respectively, for each level. Table 4 summarises sensitivity, specificity and other statistics assessing the predictive utility of the cut-offs for moderate risk (≥ 12 points) and high risk (≥ 23 points).

Table 4

Table 4

Back to Top | Article Outline


The incidence of PRF in this prospective, multicentre surgical cohort receiving general or regional anaesthesia was 4.2% and the risk of developing PRF was predicted by a score based on seven easily recorded predictors. The PERISCOPE-PRF score performed well, as it was able to identify 82% of the patients who would develop PRF (as shown by the c-statistic of 0.82) and it was able to distinguish three levels of risk. Calibration measures showed good agreement between the predicted and observed values within the risk levels; bootstrapping confirmed the stability of the dataset and all seven predictors were retained after the procedure. PRF significantly increased the ICU admission rate, postoperative LOS and in-hospital mortality.

Several studies of risk have defined a composite PPC as the primary outcome.2,22,25,26 The complications most often included are respiratory infection, bronchospasm, PRF, atelectasis and pleural effusion. Although such an approach to risk modelling is useful for guiding preoperative management and vigilance, clinicians are aware that the pathogenesis and clinical impact of each component in the composite is substantially different. We therefore designed the present study to determine whether the PERISCOPE model, also designed to predict a composite, could be used to predict only PRF.

Most previous studies of PRF defined this complication as the need for more than 48 h of mechanical ventilation or unplanned reintubation,1,3,5,7,11 which would only identify the most severe forms of PRF. The predictive scores for PRF developed in these studies showed c-statistics ranging from 0.7911 to 0.893. The c-statistic of 0.82 for the PERISCOPE-PRF score fell within this range and is consistent with those earlier findings in spite of differences in definitions or study design.

The incidence of PRF in this cohort (4.2%) was higher than previous rates, which ranged from 2.6 to 3.4%.1,8,20 There are important methodological, population and outcome definition differences between our study and the earlier ones that can account for the higher rate. Our definition of PRF specified that new-onset hypoxaemia of noncardiac cause must have appeared within 5 postoperative days, marked objectively by a level of SpO2 less than 90% breathing air, which corresponds approximately to a PaO2/FIO2 of less than 40 kPa (300 mmHg). There is no consensus about the postoperative period within which a pulmonary complication can be considered attributable to surgery.8 Several studies analysed PRF developing within 30 days,1,3,11 whereas others limited the time frame to 3 to 7 days.4–7 We chose a 5-day period so that the complication and the surgical/anaesthetic events would be clearly linked, thereby excluding 8.9% of the PERISCOPE patients who later developed this complication. Although we included patients without previous lung injury and lacked information to calculate the PaO2/FIO2 for all patients, we did classify PRF in three levels of severity, in a way that was similar to the recent ARDS classification.16 Our stratification was based on the presence of hypoxaemia and the kind of respiratory support required to manage it (conventional oxygen therapy and noninvasive or invasive mechanical ventilation regardless of PEEP level), a classification consistent with current clinical management of PRF. Up to 74% of these patients can be managed with noninvasive ventilation,27 which several studies have found very effective for treating even severe levels of hypoxaemia.28–31 Recently, Kor et al.4 found a 2.6% incidence of ALI in patients undergoing high-risk surgery using a similar definition of impaired oxygen exchange (PaO2/FIO2 <40 kPa), but their definition also required the presence of pulmonary infiltrates. It is likely that the higher PRF incidence in our study was due to the fact that the measurable criterion was arterial oxygenation (SpO2). The incidence of severe PRF in our study (PaO2/FIO2 <26.7 kPa regardless of PEEP level) was 0.5%, similar to that seen in previous studies.6 However, because of the multicentre nature of our study, we cannot rule out that local clinical practices might have led to differences in the distribution of PRF severity. Practices might even have contributed to preventing the development of PRF, or variations in resources might have led to higher rates of rescue failure32 in some centres. We think it is important for the clinician to note that all levels of postoperative hypoxaemia had an impact on mortality in this cohort (Fig. 2), a finding that confirms that PRF prediction is of great importance.

Four of the seven predictors of PRF risk we identified were related to the patient's health status and these factors accounted for 57% of the total risk. To our knowledge, this is the first study reporting that low preoperative SpO2 breathing air and even a single respiratory symptom are strongly associated with risk for PRF, although slight oxygen desaturation (SpO2 ≤95%) has been found to be an independent predictor of a PPC composite outcome.2 In addition, clinical prediction using this objective variable is even more precise when three levels of SpO2 (>95, ≤95 and ≤90%) are considered.2 In other clinical settings, a low SpO2 is emerging as a good predictor of outcome.33,34 The incidence of SpO2 of 95% or less in our surgical cohort (18.8%) was much higher than the incidence of 6.3% in a recent population-based study.35 We interpret this as a sign that the surgical population will tend towards impaired cardiorespiratory function. Exclusion of SpO2 from the score when this measurement is not available (e.g. in clinical settings wherein telephone screening is used) reduces its performance. Calibration suffers in particular, meaning that the model without SpO2 might not accurately assess level of risk (Supplementary data, Table 2, We think that routine measurement of preoperative SpO2 should be encouraged and that it will probably prove to be a robust predictor of poor postoperative outcome.

Preoperative heart failure is a well recognised risk factor for the development of PPCs.1,5,22 In our study, we analysed three levels of heart failure according to the New York Heart Association (NYHA) classification and found that PRF risk increased with increasing severity of cardiac failure. We also identified chronic liver disease as a predictor of PRF. Chronic liver disease has been linked to a poor postoperative prognosis.36 One retrospective study found an association between liver disease and unanticipated early postoperative tracheal intubation after nonemergency, noncardiac surgery5 and a retrospective study identified an 8% rate of ventilatory dependence (postoperative mechanical ventilation >24 h or unplanned intubation) and a similar rate for pneumonia in 733 cirrhotic patients undergoing any surgical procedure.37 However, chronic liver disease encompasses a wide spectrum of disorders ranging from fatty liver disease to cirrhosis. No study has sought to define a relationship between the different kinds of liver disease and PRF or other PPCs to date. We did not record different types of liver disease in our study, but the strong association we found between this factor and PRF suggests that more detailed records should be used in future studies.

The three remaining independent risk factors were associated with the surgical procedure. In most previous studies, surgical incision, duration of surgery and emergency status have been proposed as predictors of PPCs.22 However, in the PRF score we present, we further distinguished between open and closed surgery because closed surgery has been associated with less postoperative pneumonia, PRF and mortality.38 This is consistent with our finding that closed abdominal surgery approximately halved the risk for PRF and closed thoracic surgery reduced risk fourfold.

Thus, although the identified risk factors differ slightly from study to study, we do see commonalities. Patient-associated risk factors (which depend fundamentally on comorbidity) and procedure-associated risk factors are very similar across the studies. High risk and emergency surgery were identified as risk factors in most of the studies.1,3,4,7

A strength of our study is that all variables were chosen and defined a priori and cases were identified prospectively by daily searches of records. Moreover, we included patients undergoing a broad spectrum of surgeries rather than limiting the study to a specific patient population or procedure.39 This approach sought to enhance the reliability of the findings so that they would be generalisable to the real world of anaesthetics and surgery.

A limitation of this study is that postoperative follow-up ended at hospital discharge. Second, the cohort was recruited by volunteer hospitals that did not cover the entire territory of Europe. Third, possible intraoperative events that might be related to PRF, such as respiratory complications, blood loss or ventilatory management, were not taken into account. Fourth, the present study reports internal validation of the score; external validation remains to be performed.

Identifying patients at a high risk for developing PRF is of great value in clinical decision making about perioperative measures to be applied. Among the measures that have been shown to reduce the incidence of PRF, we mention preoperative optimisation of some health conditions such as smoking and alcohol cessation,40,41 intraoperative ventilatory management42–44 and postoperative analgesia and physiotherapy.45,46 Although strategies to reduce PRF risk have also been shown to reduce health costs,47–50 randomised trials to test the efficacy of preventive measures are still lacking. The PERISCOPE-PRF score developed in this study can be useful for classifying patients systematically in such trials.

In conclusion, PRF is a frequent complication and is associated with a poor prognosis, but the PERISCOPE-PRF score is likely to help identify surgical patients at risk so that stricter measures to prevent this life-threatening complication can be considered.

Back to Top | Article Outline

Acknowledgements relating to this article

Assistance with the study: Brigitte Leva provided assistance with data management and Mary Ellen Kerans provided manuscript editing and advice on the English usage in some versions of the manuscript.

Financial support and sponsorship: the PERISCOPE study was funded and supported through the ESA's Clinical Trials Network. The off-site data manager who checked entries for completeness was an ESA staff member. The ESA also supported organisational meetings, communication between centres and manuscript editing for English language expression.

Conflict of interest: none.

Presentation: none.

Back to Top | Article Outline


List of Participating Centres and Contributors to the PERISCOPE Cohort Study

Chief Investigator

Jaume Canet, Hospital Universitari Germans Trias i Pujol, Barcelona, Spain

Steering Committee Members

Jaume Canet, Spain

Sergi Sabaté, Spain

Olivier Langeron, France

Marcelo Gama de Abreu, Germany

Lluís Gallart, Spain

F. Javier Belda, Spain

Paolo Pelosi, Italy

Andreas Hoeft, Germany

Valentin Mazo, Spain

Off-site Data Management

Brigitte Leva, European Society of Anaesthesiology aisbl (Brussels), Belgium


University Hospital centre “Mother Theresa” (Tirana): Jonela Burimi, Toma Halefi, Aleksander Hoxha*, Kliti Pilika, Imelda Selmani


Cliniques Universitaires Saint Luc A.S.B.L Université Catholique de Louvain (Brussels): Véronique Daout, Caroline Gauthier, David Kahn, Mona Momeni*, Christine Watremez

Bosnia and Herzegovina

Clinical Centre University Sarajevo “Heart Centre” (Sarajevo): Slavenka Straus*

General Hospital “Prim.dr Abdulah Nakas”(Sarajevo): Dejana Djonovic-manovic, Marina Juros-Zovko*


University Hospital Rijeka (Rijka): Helga Komen-Ušljebrka*, Vlasta Orlić, Ivana Stuck

Czech Republic

Faculty Hospital Brno (Brno): Lenka Baláková, Martina Kosinová, Ivo Křikava, Roman Štoudek, Petr Štourač*, Katarina Zadražilová

Masaryks hospital Usti nad labem (Usti Nad Labem): Sanober Janvekar*


Tartu University Hospital (Tartu): Juri Karjagin, Kadri Rõivassepp, Alar Sõrmus*


Hôpital Pitié-Salpêtrière (Paris): Philippe Cuvillon, Cristina Ibáñez-Esteve, Olivier Langeron*, Mathieu Raux, Armelle Nicolas-Robin


Klinikum Darmstadt GmbH (Darmstadt): André Winter*

Medical Centre of the Johannes Gutenberg University Mainz (Mainz): Malte Brunier, Kristin Engelhard, Rita Laufenberg Feldmann*, Raphaele Lindemann, Susanne Mauff, Anne Sebastiani, Camila Zamperoni

University Hospital Bonn (Bonn): Andreas Hoeft*, Florian Kessler, Maria Wittmann

University Hospital Carl Gustav Carus - Dresden University of Technonology (Dresden): Thomas Bluth, Marcelo Gama de Abreu*, Andreas Güldner, Thomas Kiss


MISEK Kft. (Miskolc): Kristina Bráz, Csilla Ruszkai*


Azienda Ospedaliera (Padova): Massimo Micaglio, Carlo Ori, Matteo Parotto*, Paolo Persona

Azienda Ospedaliera S. Croce e Carle (Cuneo): Coletta Giuseppe*

Azienda USL n. 5 di Pisa Ospedale F. Lotti (Pontedera): Paolo Carnesecchi, Denise Lazzeroni, Irene Lorenzi*

European Institute of Oncology (Milano): Gianluca Castellani, Daniele Sances*, Gianluca Spano, Stefano Tredici, Dario Vezzoli

IRCCS San Martino (Genova): Iole Brunetti, Anna Di Noto, Angelo Gratarola, Alexandre Molin*, Luca Montagnani, Giulia Pellerano, Paolo Pelosi

Ospedale Sant’Orsola - Malpighi (Bologna): Maurizio Fusari*

University of Insubria (Varese): Laura Camici, Luca Guzzetti, Fabio Marangoni, Paolo Severgnini*

University of Milano, Ospedale San Paolo (Milano): Piero Di Mauro, Francesca Rapido, Concezione Tommasino*


Pauls Stradins Clinical University hospital (Riga): Ieva Nemme, Janis Nemme*


Kaunas Medical University Hospital (Kaunas):, Justinas Blieka, Jurgita Borodičienė, Brigita Budrytė, Aurika Karbonskiene*, Inga Kiudulaitė, Eglė Milieškaitė, Renata Rasimavičiūtė, Ugnė Sirevičienė, Ramunė Stašaitytė, Edgaras Usas, Giedrė Zarskienė

Vilnius University Hospital Santariskiu Clinics (Vilnius): Egle Kontrimaviciute, Jurate Sipylaite*, Gabija Tomkutė


ZithaKlinik(Luxembourg): Petra Bardea, Marco Klop, Marc Koch*


10 Wojskowy Szpital Kliniczny z Polikliniką w Bydgoszczy (Bydgoszcz): Dominika Bożiłow, Robert Goch*


Hospitais da Universidade de Coimbra, EPE. (Coimbra): João Bonifácio, Sofia Marques, Tânia Teresa dos Santos Ralha*

Centro Hospitalar de Lisboa Ocidental (Lisbon): Daniel Alves, Inês Carvalho, Josefina Suzana Da Cruz Parente*, Sara Tomé

Hospital Fernando Fonseca (Lisbon): Cristina Carmona*

Instituto Português de Oncologia Do Porto (Porto): Miranda Costa*, Maria Lina, Sofia Sierra


Emergency Clinical Hospital of Constanta (Constanta): Alina Balcan, Iulia Cindea, Viorel Ionel Gherghina*, Catalin Grasa

Emergency County Hospital Clinic of Anaesthesia and Intensive Care (Târgu Mures): Ruxandra Copotoiu, Sanda-Maria Copotoiu*, Judit Kovacs*, Janos Szederjesi, Arthur Theil

Emergency Institute of Cardiovascular Diseases Prof Dr C. C. Iliescu (Bucharest): Daniela Filipescu*


Krasnoyarsk State Medical University (Krasnoyarsk): Alexey Grytsan*, Tatiana Kapkan, Sergey Rostovtsev, Anastasia Yushkova


Clinica Universidad de Navarra (Pamplona): Ricardo Calderón, Elena Cacho, Carolina Marginet, Pablo Monedero*, Maria José Yepes

Consorcio Hospital General Universitario de Valencia (Valencia): Jose Miguel Esparza Miñana, Manuel Granell Gil*, Gabriel Rico Portolés

Corporació Sanitària Parc Taulí (Barcelona Sabadell): Alberto Lisi*, Gisela Perez, Nuria Poch

Fundacio Althaia (Manresa): Mauricio Roberto Argañaraz Quinteros, Carme Font Bosch, Jordi Torrellardona Llobera*

Fundació Puigvert (Barcelona): Sergi Sabaté*, Pilar Sierra

Hospital Arnau de Vilanova (Lleida): Mercedes Matute*

Hospital Clinic de Barcelona (Barcelona): Amalia Alcon Dominguez*, María José Arguis, Isabel Belda, Enrique Carrero, Jacobo Moreno, Irene Rovira, Marta Ubre, Roberto Castillo, Sílvia Herrero

Hospital Clínic Universitari de Valencia (Valencia): Maria Teresa Ballester Luján*, F.Javier Belda, José Carbonell, Geri Gencheva, Andrea Gutierrez, Julio Llorens, Sofia Machado

Hospital de Denia (Denia): Francisca Llobell*, Daniel Paz Martin

Hospital del Tajo Aranjuez (Madrid): Francisco Javier García-Miguel*

Hospital General de La Palma Breña Alta (La Palma, Canarias): Aníbal Pérez García*

Hospital General Universitario Alicante (Alicante): Roque Company*, Aixa Ahamdanech Idrissi, Josefina del Fresno Cañaveras, Jose Alejandro Navarro Martinez ; Estefania Paya Martinez, Ester Sanchez Garcia

Hospital San Jorge (Huesca); Jorge Vera Bella*

Hospital Sant Pau (Barcelona): Inmaculada India Aldana, J. Manuel Campos, Xavier Pelaez Vaamonde*

Hospital Santa Maria (Lleida): Montserrat Torra*

Hospital Universitari del Mar ‘Parc de Salut Mar (Barcelona): Raquel Arroyo, Juan Carlos Cabrera, Jesús Carazo Cordobes*, Lluís Gallart, Amelia Rojo, Francisco Javier Santiveri

Hospital Universitari Germans Trias i Pujol (Badalona): Jaume Canet*, Miriam González, Anabel Jiménez, Yolanda Jiménez, Agnès Martí, Valentin Mazo, Enrique Moret, Monica Rodriguez Nuñez*, Joaquin Velasco

Hospital Universitario 12 de Octubre (Madrid): Adriana Calderón, Matide González, Olga González, Ana Hermira Anchuelo*, Eloisa López, Esther Sánchez

Hospital Universitario de La Princesa (Móstoles-Madrid): Blanca Aznárez Zango*, Francisco José García Corral, Esperanza Mata Mena, Antonio Planas Roca

Hospital Universitario de Móstoles (Madrid): Raquel Fernández Rocío Ayala Soto*, Borja Quintana

Hospital Universitario Marques De Valdecilla (Santander): Jose Manuel Rabanal Llevot*, Mónica Mercedes Williams Camus, Alba Palacios Blanco, Angela Largo Ruiz

Hospital Universitario Rio Hortgea (Valladolid): Jesus Rico Feijoo*

Hospital Universitario Virgen del Rocio (Sevilla): Elvira Castellano Garijo*

Hospital Son Llatzer (Palma de Mallorca): Julio Belmonte Cuenca*, Marcos José Bonet Binimelis, Ivaylo Grigorov, Josep Lluis Aguilar

Vall d’Hebron University Hospital (Barcelona): Míriam De Nadal Clanchet, Encarnación Guerrero Viñas, Susana Manrique Muñiz, Víctor Martín Mora, Francisca Munar Bauzà, Sonia Núñez Aguado, Montserrat Olivé Vidal*, María luisa Paños Gozalo, Marcos Sánchez Marín, María Carmen Suescun López


Ospedale Regionale di Lugano (Lugano): Paolo Maino*


St.Katherine Hospital of Cardiology (Odessa): Yevhen Eugene Yevstratov*


Medical Faculty of Istanbul, Istanbul University (Istanbul): Semra Kucukgoncu, Nuzhet Mert Sentürk*, Zerrin Sungur Ulke

*Site leader.

Back to Top | Article Outline


1. Arozullah AM, Daley J, Henderson WG, Khuri SF. Multifactorial risk index for predicting postoperative respiratory failure in men after major noncardiac surgery. The National Veterans Administration Surgical Quality Improvement Program. Ann Surg 2000; 232:242–253.
2. Canet J, Gallart L, Gomar C, et al. Prediction of postoperative pulmonary complications in a population-based surgical cohort. Anesthesiology 2010; 113:1338–1350.
3. Gupta H, Gupta PK, Fang X, et al. Development and validation of a risk calculator predicting postoperative respiratory failure. Chest 2011; 140:1207–1215.
4. Kor DJ, Warner DO, Alsara A, et al. Derivation and diagnostic accuracy of the surgical lung injury prediction model. Anesthesiology 2011; 115:117–128.
5. Ramachandran SK, Nafiu OO, Ghaferi A, et al. Independent predictors and outcomes of unanticipated early postoperative tracheal intubation after nonemergent, noncardiac surgery. Anesthesiology 2011; 115:44–53.
6. Blum JM, Stentz MJ, Dechert R, et al. Preoperative and intraoperative predictors of postoperative acute respiratory distress syndrome in a general surgical population. Anesthesiology 2013; 118:19–29.
7. Brueckmann B, Villa-Uribe JL, Bateman BT, et al. Development and validation of a score for prediction of postoperative respiratory complications. Anesthesiology 2013; 118:1276–1285.
8. Canet J, Gallart L. Postoperative respiratory failure: pathogenesis, prediction, and prevention. Curr Opin Crit Care 2014; 20:56–62.
9. Roussos C, Koutsoukou A. Respiratory failure. Eur Respir J Suppl 2003; 47:3s–14s.
10. Trillo-Alvarez C, Cartin-Ceba R, Kor DJ, et al. Acute lung injury prediction score: derivation and validation in a population-based sample. Eur Respir J 2011; 37:604–609.
11. Hua M, Brady JE, Li G. A scoring system to predict unplanned intubation in patients having undergone major surgical procedures. Anesth Analg 2012; 115:88–94.
12. Utter GH, Cuny J, Sama P, et al. Detection of postoperative respiratory failure: how predictive is the Agency for Healthcare Research and Quality's Patient Safety Indicator? J Am Coll Surg 2010; 211:347–354.
13. Pelosi P, Gama de Abreu M. Lung injury prediction models to improve perioperative management: let's hit the bull's-eye!. Anesthesiology 2011; 115:10–11.
14. Koch CG, Li L, Hixson E, et al. What are the real rates of postoperative complications: elucidating inconsistencies between administrative and clinical data sources. J Am Coll Surg 2012; 214:798–805.
15. Young A, Ramachandran SK. Clinical prediction of postoperative respiratory failure. Anesthesiology 2013; 118:1247–1249.
16. Ranieri VM, Rubenfeld GD, Thompson BT, et al. Acute respiratory distress syndrome: the Berlin Definition. JAMA 2012; 307:2526–2533.
17. Mazo V, Sabaté S, Canet J, et al. Prospective external validation of a predicitive score for postoperative pulmonary complications. Anesthesiology 2014; 121:219–231.
18. Cotes JE. Medical Research Council Questionnaire on Respiratory Symptoms (1986). Lancet 1987; 2:1028.
19. Peduzzi P, Concato J, Kemper E, et al. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol 1996; 49:1373–1379.
20. Johnson RG, Arozullah AM, Neumayer L, et al. Multivariable predictors of postoperative respiratory failure after general and vascular surgery: results from the patient safety in surgery study. J Am Coll Surg 2007; 204:1188–1198.
21. Money SR, Rice K, Crockett D, et al. Risk of respiratory failure after repair of thoracoabdominal aortic aneurysms. Am J Surg 1994; 168:152–155.
22. Smetana GW, Lawrence VA, Cornell JE. Preoperative pulmonary risk stratification for noncardiothoracic surgery: systematic review for the American College of Physicians. Ann Intern Med 2006; 144:581–595.
23. Steyerberg EW, Harrell FE Jr, Borsboom GJ, et al. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis. J Clin Epidemiol 2001; 54:774–781.
24. Liu H, Hussain F, Tan CL, Dash M. Discretization: an enabling technique. Data Min Knowl Discov 2002; 6:393–423.
25. Mitchell CK, Smoger SH, Pfeifer MP, et al. Multivariate analysis of factors associated with postoperative pulmonary complications following general elective surgery. Arch Surg 1998; 133:194–198.
26. McAlister FA, Bertsch K, Man J, et al. Incidence of and risk factors for pulmonary complications after nonthoracic surgery. Am J Respir Crit Care Med 2005; 171:514–517.
27. Redondo Calvo FJ, Madrazo M, Gilsanz F, et al. Helmet noninvasive mechanical ventilation in patients with acute postoperative respiratory failure. Respir Care 2012; 57:743–752.
28. Squadrone V, Coha M, Cerutti E, et al. Continuous positive airway pressure for treatment of postoperative hypoxemia: a randomized controlled trial. JAMA 2005; 293:589–595.
29. Michelet P, D’Journo XB, Seinaye F, et al. Noninvasive ventilation for treatment of postoperative respiratory failure after oesophagectomy. Br J Surg 2009; 96:54–60.
30. Jaber S, Michelet P, Chanques G. Role of noninvasive ventilation (NIV) in the perioperative period. Best Pract Res Clin Anaesthesiol 2010; 24:253–265.
31. Glossop AJ, Shephard N, Bryden DC, Mills GH. Noninvasive ventilation for weaning, avoiding reintubation after extubation and in the postoperative period: a meta-analysis. Br J Anaesth 2012; 109:305–314.
32. Wakeam E, Hevelone ND, Maine R, et al. Failure to rescue in safety-net hospitals: availability of hospital resources and differences in performance. JAMA Surg 2014; 149:229–235.
33. Millman AL, Payne B, Qu Z, et al. Oxygen saturation as a predictor of adverse maternal outcomes in women with preeclampsia. J Obstet Gynaecol Can 2011; 33:705–714.
34. Seymour CW, Kahn JM, Cooke CR, et al. Prediction of critical illness during out-of-hospital emergency care. JAMA 2010; 304:747–754.
35. Vold ML, Aasebo U, Hjalmarsen A, Melbye H. Predictors of oxygen saturation </=95% in a cross-sectional population based survey. Respir Med 2012; 106:1551–1558.
36. Pandey CK, Karna ST, Pandey VK, et al. Perioperative risk factors in patients with liver disease undergoing nonhepatic surgery. World J Gastrointest Surg 2012; 4:267–274.
37. Ziser A, Plevak DJ, Wiesner RH, et al. Morbidity and mortality in cirrhotic patients undergoing anesthesia and surgery. Anesthesiology 1999; 90:42–53.
38. Masoomi H, Buchberg B, Nguyen B, et al. Outcomes of laparoscopic versus open colectomy in elective surgery for diverticulitis. World J Surg 2011; 35:2143–2148.
39. Canet J, Gallart L. Predicting postoperative pulmonary complications in the general population. Curr Opin Anaesthesiol 2013; 26:107–115.
40. Musallam KM, Rosendaal FR, Zaatari G, et al. Smoking and the risk of mortality and vascular and respiratory events in patients undergoing major surgery. JAMA Surg 2013; 148:755–762.
41. Tonnesen H, Nielsen PR, Lauritzen JB, Moller AM. Smoking and alcohol intervention before surgery: evidence for best practice. Br J Anaesth 2009; 102:297–306.
42. Severgnini P, Selmo G, Lanza C, et al. Protective mechanical ventilation during general anesthesia for open abdominal surgery improves postoperative pulmonary function. Anesthesiology 2013; 118:1307–1321.
43. Futier E, Constantin JM, Paugam-Burtz C, et al. A trial of intraoperative low-tidal-volume ventilation in abdominal surgery. N Engl J Med 2013; 369:428–437.
44. Hemmes SN, Gama de Abreu M, Pelosi P, Schultz MJ. High versus low positive end-expiratory pressure during general anaesthesia for open abdominal surgery (PROVHILO trial): a multicentre randomised controlled trial. Lancet 2014; 384:495–503.
45. Makhabah DN, Martino F, Ambrosino N. Peri-operative physiotherapy. Multidiscip Respir Med 2013; 8:4.
46. Cassidy MR, Rosenkranz P, McCabe K, et al. I COUGH: reducing postoperative pulmonary complications with a multidisciplinary patient care program. JAMA Surg 2013; 148:740–745.
47. Fuller RL, McCullough EC, Bao MZ, Averill RF. Estimating the costs of potentially preventable hospital acquired complications. Healthcare Financ Rev 2009; 30:17–32.
48. Carey K, Stefos T, Shibei Z, et al. Excess costs attributable to postoperative complications. Med Care Res Rev 2011; 68:490–503.
49. Sasaki N, Meyer MJ, Eikermann M. Postoperative respiratory muscle dysfunction: pathophysiology and preventive strategies. Anesthesiology 2013; 118:961–978.
50. Sabate S, Mazo V, Canet J. Predicting postoperative pulmonary complications: implications for outcomes and costs. Curr Opin Anaesthesiol 2014; 27:201–209.

Supplemental Digital Content

Back to Top | Article Outline
© 2015 European Society of Anaesthesiology