Preoperative predictive factors for intensive care unit admission after pulmonary resection*

Objective: To determine whether the use of a set of preoperative variables can predict the need for postoperative ICU admission. Methods: This was a prospective observational cohort study of 120 patients undergoing elective pulmonary resection between July of 2009 and April of 2012. Prediction of ICU admission was based on the presence of one or more of the following preoperative characteristics: predicted pneumonectomy; severe/very severe COPD; severe restrictive lung disease; FEV1 or DLCO predicted to be < 40% postoperatively; SpO2 on room air at rest < 90%; need for cardiac monitoring as a precautionary measure; or American Society of Anesthesiologists physical status ≥ 3. The gold standard for mandatory admission to the ICU was based on the presence of one or more of the following postoperative characteristics: maintenance of mechanical ventilation or reintubation; acute respiratory failure or need for noninvasive ventilation; hemodynamic instability or shock; intraoperative or immediate postoperative complications (clinical or surgical); or a recommendation by the anesthesiologist or surgeon to continue treatment in the ICU. Results: Among the 120 patients evaluated, 24 (20.0%) were predicted to require ICU admission, and ICU admission was considered mandatory in 16 (66.6%) of those 24. In contrast, among the 96 patients for whom ICU admission was not predicted, it was required in 14 (14.5%). The use of the criteria for predicting ICU admission showed good accuracy (81.6%), sensitivity of 53.3%, specificity of 91%, positive predictive value of 66.6%, and negative predictive value of 85.4%. Conclusions: The use of preoperative criteria for predicting the need for ICU admission after elective pulmonary resection is feasible and can reduce the number of patients staying in the ICU only for monitoring.

admission after pulmonary resection. Therefore, we decided to design a study to investigate whether the use of a set of preoperative variables can predict the need for immediate postoperative ICU admission.

Methods
A prospective observational cohort study of patients referred to the outpatient preoperative evaluation clinic of the Department of Pulmonology of the Universidade Federal de São Paulo (UNIFESP, Federal University of São Paulo) was carried out between July of 2009 and April of 2012. This study was approved by the UNIFESP Research Ethics Committee (Ruling no. 410/09).
We included patients over 18 years of age undergoing elective pulmonary resection, with diagnosed or suspected benign or malignant disease. The exclusion criteria were as follows: incomplete preoperative evaluation; one concomitant surgical procedure in addition to pulmonary resection; parenchymal-sparing procedures; and preoperative or intraoperative death.
After giving written informed consent, participants underwent a preoperative evaluation consisting of clinical assessment and physical examination, with the use of a structured form. Pulmonary function was assessed by using the modified algorithm presented in the American College of Chest Physicians (ACCP) guidelines, (12) with the following tests: pre-and post-bronchodilator spirometry; pulse oximetry; DLCO measurement; pH measurement and arterial blood gas analysis; cardiopulmonary exercise testing; and pulmonary perfusion mapping. On completing the evaluation, patients were categorized as being at high or acceptable risk.
All patients were operated on by the same team of thoracic surgeons and were referred to the ICU of the Pulmonology Section in the Introduction According to the 1997 American Thoracic Society (1) statement regarding ICU allocation decisions, the primary goal of the ICU is "to preserve meaningful human life by protecting and sustaining patients in a caring manner when they are threatened by an acute critical illness or injury or as a consequence of medical or surgical therapy". The 1999 update (2) adds that the ICU serves to monitor and care for patients with potentially severe physiological instability requiring technical and/or artificial life support.
The European Respiratory Society and the European Society of Thoracic Surgeons (3) do not recommend systematic ICU admission after thoracotomy. Patients who are estimated to be at low risk for complications should be referred to a dedicated thoracic surgery unit. Patients with reduced cardiopulmonary reserve undergoing complex resection and patients who are estimated to be at moderate to high risk for complications should be referred to a high-dependency unit, if available, whereas patients requiring support for organ failure should be admitted to the ICU.
It is known that among patients admitted to the ICU after surgery, only a minority develops acute complications requiring immediate intervention; the majority is referred to the ICU for "surveillance" of possible deterioration of their clinical condition or simply for monitoring. (4)(5)(6) It should be noted that patients referred to the ICU only for monitoring may experience unfavorable outcomes, such as increased stress due to the environment as well as to sleep and family deprivation. In addition, there is a substantial increase in hospital costs. (7) Thoracic surgery always results in pulmonary dysfunction, and depending on the degree of impairment, there can be difficulty in extubating the patient at the end of the surgical procedure and need for prolonged mechanical ventilation. In addition, many surgical candidates present with comorbidities and/or compromised cardiopulmonary reserve, which makes them more susceptible to developing perioperative complications. (2,3) However, the question regarding the proportion of these patients who would benefit from the ICU setting in the immediate postoperative period also remains unanswered.
Few studies have bothered to determine prognostic factors associated with the need for ICU admission, (8)(9)(10)(11) and in the literature there is no consensus regarding indications for ICU 33 http://dx.doi.org/10.1590/S1806-37132015000100005 ventilation after surgery or for reintubation; ARF or need for noninvasive positive pressure ventilation; hypotension with hemodynamic instability or associated with signs of shock, requiring blood transfusion as well as fluid resuscitation; unresolved intraoperative or immediate postoperative complications (clinical or surgical); or a recommendation by the anesthesiologist or surgeon to continue treatment in the ICU.

Statistical analysis
Categorical variables were summarized as absolute and relative frequencies (percentages). Numerical variable data were expressed as mean, standard deviation, median, and interquartile range. Diagnostic performance of the ICU admission prediction model was tested by calculating sensitivity, specificity, positive predictive value, negative predictive value, and accuracy, and was measured against the gold standard for mandatory ICU admission. All data were tabulated using Microsoft ® Excel 2007. For statistical calculations, we used the Statistical Package for the Social Sciences for Windows, version 19.0 (IBM SPSS Statistics, IBM Corp., Armonk, NY, USA). Intraoperative variables analyzed included anesthesia time, procedure performed, number of segments resected, and surgical and clinical complications.
The gold standard for mandatory admission to the ICU was based on the presence of one or more of the following postoperative characteristics: need for maintenance of invasive mechanical Chart 1 -Criteria for the diagnosis of complications observed after pulmonary resection. (15)(16)(17)(18) Complication Criteria PMV Need for mechanical ventilation for more than 48 hours after surgery for treatment of ARF. (15)

Pulmonary infection
Antimicrobial treatment for suspected pulmonary infection, associated with at least one of the following criteria: onset of purulent secretions or change in character of secretions; presence of a new opacity on chest X-ray or progression of a pre-existing one; fever; or leukocytosis > 12,000 cells/mm 3 . (15,16)

Atelectasis
Chest X-ray findings consistent with onset or worsening of respiratory symptoms. Bronchospasm Wheezing on lung auscultation associated with shortness of breath and requiring medications, as long as LHF and PTE are ruled out. ARF Clinical condition resulting from acutely impaired gas exchange and requiring invasive ventilatory support. Shock Absolute hypotension (systolic blood pressure less than 90 mmHg) or relative hypotension (a greater than 40 mmHg decrease in systolic blood pressure) associated with signs of hypoperfusion, oliguria, pallor, cold skin, altered mental status, and metabolic acidosis, with hemodynamic instability or requiring vasoactive medication. (17)
The 30-day mortality rate was 2.5% (3/120), and three deaths were due to septic shock from pulmonary infection, which, in 2 cases, was preceded by a bronchial fistula. The 3 patients in question had a diagnosis of bronchiectasis with aspergilloma. Another 4 deaths occurred after 30 days, but during the same hospital stay: 2 occurred during an operation for bronchiectasis and 2 occurred during an operation for cancer.
Among the 120 patients evaluated, 24 (20.0%) were predicted to require ICU admission, and the clinical criteria for this prediction are shown in Table 2.
Among those 24 cases, there were 16 true positives and 8 false positives. Of those 8 false positives, 1 underwent surgical treatment of bleeding bronchiectasis, had restrictive lung disease, and had a ppo-DLCO of 33%, and 1 had severe COPD, with a preoperative FEV 1 of 45% predicted and an SpO 2 of 88%. In the remaining 6, the proposed surgical procedure was pneumonectomy. All false positives were discharged from the ICU in the morning after the procedure (Table 3).
of the 120 patients included in the study are shown in Table 1. DLCO was measured in 31 patients (25.8%), and its mean ± SD was 64.4 ± 19% predicted. Maximal oxygen consumption was determined in 4 patients who underwent cardiopulmonary exercise testing, and its mean was 21 mL • kg -1 • min -1 or 74.5% predicted.
The mean anesthesia time was 6.1 ± 1.8 hours. The surgical procedures performed   Therefore, when analyzing studies of prognostic risk models, it is necessary to consider the population they apply to, and when it comes to validation studies, it is necessary to determine whether the circumstances under which they were carried out can be superimposed to those of the original study. However, it is important to bear in mind that most health care facilities do not have the technological means to run sophisticated models, nor do they follow strict inclusion protocols.
What could then be done in this context? Pieretti et al. (9) used a set of pre-established criteria for predicting the need for ICU admission and obtained satisfactory results. Therefore, at our facility-i.e., a university hospital-we decided to determine whether the use of a set of preoperative clinical variables for predicting the need for ICU admission would be able to accomplish this goal.
The choice of the clinical criteria used for predicting the need for ICU admission was based on data in the literature and on our daily clinical practice. Pneumonectomy accounts for the highest morbidity and mortality rates among the various possible resections. (20)(21)(22)(23) ASA physical status 3 means that the patient has severe systemic Among the 96 patients for whom ICU admission was not predicted, there were 96 true negatives and 14 false negatives. Those 14 false negatives were ASA physical status 2, had a mean age of 60 years, and had a mean ppo-FEV 1 of 70%; in addition, the proposed surgical procedure was not pneumonectomy ( Table 3).
The accuracy of the ICU admission prediction model was 81.6% (98/120), with sensitivity of 53.3%, specificity of 91%, positive predictive value of 66.6%, and negative predictive value of 85.4%.
Analysis of the false negative subgroup revealed that, in 2 cases, the operation was converted from lobectomy to pneumonectomy intraoperatively; 2 patients developed arrhythmia, which was promptly corrected, and 12 patients had shock or hypotension requiring transfusion, the shock being caused by intraoperative or immediate postoperative bleeding in 10 ( Table 4).

Discussion
Studies on criteria for predicting ICU admission after pulmonary resection are few in number and have included mainly patients with lung cancer. (7,10,11) When planning this study, we decided to include patients with benign diseases as well, because, in developing countries, surgical procedures for the treatment of bronchiectasis and other post-infectious pulmonary sequelae are still very common. Benign diseases are known to be more prevalent in patients younger than those presenting with cancer, but this does not make the rates of postoperative morbidity (18 to 46%) and mortality (zero to 26.3%) any lower. (19)(20)(21) In 2008, Brunelli et al. (10) developed and validated the first risk scale for predicting the need for ICU admission after pulmonary resection, principally for patients with lung cancer. Of the 1,297 participants, 82 (6.3%) required ICU admission, and, using the logistic regression model, those authors found that the independent predictors of need for ICU admission were age over 65 years, ppo-FEV 1 < 65%, ppo-DLCO < 50%, cardiac comorbidities, and pneumonectomy.
Okiror et al. (11) performed external validation of the scale developed in aforementioned study and concluded that the scale of Brunelli et al. had a moderate discriminatory power for predicting the need for ICU admission. However, the criteria used in the validation study were not the same as those of the original study (emergency ICU  Among the 96 patients for whom the need for ICU admission was not predicted, the non-prediction was incorrect in 14 (false negatives), a situation that is a greater cause for concern than is predicting the need for ICU admission when that is not the case, if the error endangers the postoperative course of such patients. All the complications that justified the patient staying in the ICU occurred during or soon after the surgical procedure, namely: shock and/or hypotension requiring blood transfusion as well as fluid resuscitation; and intraoperative conversion from lobectomy to pneumonectomy. And this scenario could not have been predicted preoperatively.
Although these complications ended up reducing the sensitivity of the criteria used, which was 53.3%, the specificity and negative predictive value of the criteria were encouraging, 91% and 85.4%, respectively, and their accuracy was 81.7%.
These findings make us think that prediction of the need for ICU admission after pulmonary resection should consider the possibility of major bleeding and not only the size of the disease resulting in functional limitation in activities of daily living. (14,(24)(25)(26) Values of < 50% predicted for FEV 1 and FVC, as determined by spirometry, are associated with severe or very severe disease. In a scenario of a patient with reduced functional reserve, who will undergo removal of nonfunctioning parenchyma, the use of ppo-FEV 1 or ppo-DLCO for predicting the need for postoperative ICU admission corrects this distortion, to some extent, and a threshold of 40% has been established by the ACCP guidelines. (12) An SpO 2 on room air at rest < 90% indicates reduced functional reserve and worsening of gas exchange during the removal of still functioning parenchyma. (28,29) Cardiac monitoring as a precautionary measure, the need for which is defined by the cardiologist, is recommended in patients with reduced cardiovascular reserve. (12) Patients were predicted to require ICU admission because there was a major surgical risk factor or because their clinical status was compromised either by reduced cardiopulmonary reserve or by comorbidities. However, the prediction was incorrect in 8 patients (false positives), and the predictive factors were ppo-DLCO < 40%, severe COPD with an SpO 2 on room air of 88%, and planned pneumonectomy. This finding could signal that, even in large operations, if there are Most studies addressing prediction of the need for ICU admission in pulmonary resection candidates have focused mainly on patients with malignant lung diseases, (7)(8)(9)(10)(11) and our study showed that, among patients with bronchiectasis, the rate of intraoperative bleeding was more than twice as high as that among patients with lung cancer. The major limitation of our study is its sample size, which did not allow the internal validation of the study. The inclusion of patients from a single facility can limit the size of the population sample and preclude the generalization of results. However, it has several advantages: it is convenient for the population that seeks our services; all surgical procedures are performed by the same surgical team; preoperative clinical assessment is also performed by the same team and always in the same way, regardless of whether that is a clinical study or that is simply health care; and the care provided in the ICU and the follow-up care by physiotherapists and by the pain team of the hospital are consistent with our peculiarities and difficulties. Another minimal limitation is the 53.3% sensitivity of the model, a value that reduces its reliability in predicting the need for ICU admission after pulmonary resection. However, this model ensures high specificity and high negative predictive value in predicting which patients will not require ICU admission.
In conclusion, the use of composite measures for predicting the need for ICU admission after pulmonary resection is feasible and accurate, and since this model uses clinical variables that do not require high technology (planned pneumonectomy; severe COPD; severe restrictive lung disease; ppo-FEV 1 or ppo-DLCO < 40%; SpO 2 on room air at rest < 90%; need for cardiac monitoring as a precautionary measure; ASA physical status ≥ 3), it may have wide applicability in daily clinical practice.