Published ahead of print on December 4, 2002, doi:10.1164/rccm.200209-985BC
© 2003 American Thoracic Society
Accuracy of the Preoperative Assessment in Predicting Pulmonary Risk after Nonthoracic SurgeryDivision of General Internal Medicine, University of Alberta, Edmonton; Division of General Internal Medicine, University of Calgary, Calgary, AB; Divisions of General Internal Medicine and Geriatrics, Department of Medical Imaging, University of Toronto, Toronto, ON, Canada; Division of Internal Medicine, Democritus University of Thrace, Alexandroupolis, Greece; Division of Pulmonary and Critical Care Medicine, Mayo Clinic, Rochester, Minnesota; Department of Anesthesia, Glasgow Royal Infirmary, Scotland Correspondence and requests for reprints should be addressed to Dr. Finlay A. McAlister, 2E3.24 WMC, University of Alberta Hospital, 8440 112 Street, Edmonton, AB, Canada T6G 2R7. E-mail: finlay.mcalister{at}ualberta.ca
We examined the accuracy of preoperative assessment in predicting postoperative pulmonary risk in a prospective cohort of 272 consecutive patients referred for evaluation before nonthoracic surgery. Outcomes were assessed by an independent investigator who was blinded to the preoperative data. There were 22 (8%) postoperative pulmonary complications. Statistically significant predictors of pulmonary complications (all p  0.005) were as follows: hypercapnea of 45 mm Hg or more (odds ratio, 61.0), a FVC of less than 1.5 L/minute (odds ratio, 11.1), a maximal laryngeal height of 4 cm or less (odds ratio, 6.9), a forced expiratory time of 9 seconds or more (odds ratio, 5.7), smoking of 40 pack-years or more (odds ratio, 5.7), and a body mass index of 30 or more (odds ratio, 4.1). Multiple regression analyses revealed three preoperative clinical factors that are independently associated with pulmonary complications: an age of 65 years or more (odds ratio, 1.8; p = 0.02), smoking of 40 pack-years or more (odds ratio, 1.9; p = 0.02), and maximum laryngeal height of 4 cm or less (odds ratio, 2.0; p = 0.007). Thus, preoperative factors can identify those patients referred to pulmonologists or internists who are at increased risk for pulmonary complications after nonthoracic surgery.
Key Words: complications, postoperative • clinical skills • pulmonary function tests Although pulmonary complications afflict almost 5% of patients after noncardiac surgery (1, 2), only seven studies have rigorously evaluated the accuracy of preoperative or operative variables in the prediction of postoperative pulmonary complications in patients undergoing nonthoracic surgery (3). The patients enrolled, variables tested, and outcomes assessed in these seven studies varied widely, and several key elements of the history, physical examination, and pulmonary function tests, which are often cited as predictive of pulmonary complications, either have not been evaluated or were not found to be associated with these complications (3). Recently published risk prediction equations for postoperative respiratory failure or pneumonia are significant advances in this field (1, 2). Arozullah and colleagues conducted secondary analyses of data from the 100-hospital strong Veterans Affairs National Surgical Quality Improvement Program to derive multifactorial risk indices for predicting respiratory failure (12 variables) or pneumonia (14 variables) after noncardiac surgery. Although these indices permit general estimates of risk in patients undergoing a variety of surgical procedures, they do not include data from the physical examination or selected laboratory tests (such as pulmonary function tests, arterial blood gases, or chest radiography) commonly employed by clinicians to assist in risk prediction before surgery. To address the continuing uncertainty about the ability of preoperative assessment to predict postoperative pulmonary complications, we examined the accuracy of selected elements of this evaluation in patients undergoing nonthoracic surgical procedures. We hypothesized that some factors detectable during the preoperative assessment would be significantly associated with postoperative pulmonary complications. Some of the results of this study have been previously reported in the form of an abstract at the 2001 American College of Chest Physicians Meeting (4).
Investigators who saw preoperative consultations at the behest of other physicians were recruited from the Clinical Assessment of the Reliability of the Examination (CARE) study group (see http://www.carestudy.com for details) as per previous studies from our group (5, 6). Ten groups (22 investigators, all of whom were general internists or pulmonologists) enrolled consecutive outpatients referred for preoperative consultation over a defined time period, regardless of pulmonary status. Patients were excluded if they were mechanically ventilated at the time of preoperative assessment, had a history of sleep apnea, had known neuromuscular disease, had significant other medical problems that precluded their participation (e.g., cognitive impairment or severe orthopedic injury), or were having intrathoracic surgery or if the patient did not subsequently undergo the planned operation. One investigator from each group collected preoperative history, physical examination, and laboratory data, as well as operative data (Tables 1–3) using a standardized protocol and data collection sheets. A second investigator from that group, blinded to the results of the first investigator's assessment or laboratory tests, collected the outcome data using prestandardized definitions and protocols. The primary outcome was the occurrence of a symptomatic and clinically significant postoperative pulmonary complication before hospital discharge or by the 7th postoperative day (whichever came first), including respiratory failure requiring invasive or noninvasive mechanical ventilation, pneumonia (2, 7), or atelectasis of lung or lobe requiring bronchoscopic intervention. Only the first of these complications occurring in any one patient was used for this analysis, and outcomes were included if any patients were readmitted to hospital with a pulmonary complication within 7 days of surgery. Laboratory tests (such as pulmonary function tests, chest radiographs, or blood gases) were not routinely ordered perioperatively, and all such test ordering was left to the discretion of the attending physicians caring for these patients.
We conducted bivariate analyses (using the chi-squared or Fisher exact test for dichotomous variables, the chi-squared test for trend for categorical variables, and the t-test for continuous variables) to examine the association between each variable in Tables 1–3 and the occurrence of postoperative pulmonary complications and calculated odds ratios (ORs). As we planned to conduct multiple comparisons, we a priori adjusted the p value for significance to p 0.005 (using the Bonferroni correction) for the variables in Tables 1–3. Multiple logistic regression analyses were performed to examine clinical factors that identified high-risk patients (i.e., without inclusion of laboratory data) with the p value for significance set at 0.05 or less (8). Additional detail on the data collection procedures, the variables collected (including how to perform the physical examination maneuvers and the types of surgeries included), the outcomes sought, and data analysis is provided in an online supplement.
We enrolled 272 consecutive patients who were referred to the participating investigators for preoperative consultation. The mean age was 62 years; 108 (40%) were men, and 22 patients (8%) suffered a postoperative pulmonary complication: six respiratory failure requiring ventilatory support, nine postoperative pneumonia, and seven atelectasis requiring bronchoscopic intervention. One patient with postoperative pneumonia subsequently deteriorated and required invasive ventilatory support in the intensive care unit. There was no appreciable difference in complication rates between sites. The results of bivariate analyses for the history, physical examination, operative, and laboratory variables collected are included in Tables 1–3. Although six factors were found to be significantly associated with pulmonary complications after surgery, several other variables commonly cited to increase risk (such as male sex, recent upper respiratory tract infection, and history of chronic productive cough) were not associated with the outcomes evaluated in this study. Results of pulmonary function tests and arterial blood gases were useful in predicting the occurrence of postoperative pulmonary complications (Table 3). However, in contrast to our a priori hypothesis, patients at a higher risk of pulmonary complications were not more likely to have these tests done preoperatively. For example, although patients who had preoperative pulmonary function tests had a postoperative pulmonary complication rate of 6.9%, the complication rate in those who didn't have pulmonary function tests done was 9.4% (and 6.3% versus 8.7% for arterial blood gases). The multiple regression model for clinical variables present at the preoperative assessment (i.e., not including laboratory tests) that best fit the data included these variables: had an age of 65 years or more (OR 1.8, p = 0.02), smoked 40 pack-years or more (OR 1.9, p = 0.02), and had a maximum laryngeal height of 4 cm or less (OR 2.0, p = 0.007). The area under the receiver operating characteristic curve for this model was 0.79 and, using a leave-one-out cross-validation procedure, was estimated to be 0.67 for new data. We decided a priori to not include laboratory data in the preoperative clinical factors multiple regression analysis as (1) not all patients would have had the tests done (since the performance of laboratory tests was left to the discretion of the attending physician caring for the patients) and because (2) we wanted to determine which clinical factors identified high-risk patients who should subsequently undergo pulmonary function tests or arterial blood gases to define their risk better. We did include dummy variables to determine whether being sent for particular laboratory tests predicted risk—none did.
This study demonstrates that a focused preoperative assessment can identify patients at increased risk for pulmonary complications after nonthoracic surgery. These findings complement the recently published risk prediction equations for respiratory failure and pneumonia (1, 22) and extend their findings, as we assessed a variety of clinically relevant factors (particularly physical examination maneuvers) not evaluated in their study. Although an "abnormal preoperative physical examination" has previously been shown to predict postoperative pulmonary complications (9), we were able to define the specific elements of the physical examination that were most useful. Similarly, although others have shown that smoking history correlates with pulmonary complications (1, 2, 10), we were able to provide a quantitative estimate of this increased risk. The ORs we found for other elements of the preoperative evaluation are consistent with those reported in earlier studies: a history of chronic obstructive airways disease (OR 4.2 in our study versus 4.5 in a study of 1,331 patients undergoing abdominal surgery) (11), dyspnea-limited exercise capacity two blocks or less (OR 3.0 in our study versus 2.9 in a study of 73 patients undergoing head and neck surgery) (10), and productive cough preoperatively (OR 1.9 in our study versus 2.2 in a study of 148 patients undergoing a variety of surgical procedures) (12). Similar to four of five previous studies (3), we were not able to demonstrate a greater risk of pulmonary complications in patients undergoing general anesthesia or having an upper abdominal incision. Finally, our study is the first using a blinded and independent comparison of preoperative tests and postoperative outcomes to demonstrate that hypercapnea and reduced FVC do correlate with pulmonary complications after nonthoracic surgery (3). There are a number of limitations with our study. First, this prediction model has not been validated in an independent data set, although a validation study is ongoing and our leave-one-out cross-validation procedure confirmed the internal validity of our results (13). Second, the selection bias inherent in studying a sample of patients referred for preoperative consultation (albeit consecutive patients) rather than unselected patients undergoing surgery may lead to overestimation of the ORs (14, 15), although our observed pulmonary complication rate (8.1%) was within the range reported in previous studies (3). Third, we do not have any information on whether other physicians sent these patients to the participating internists and pulmonologists because they were felt to be at higher pulmonary risk; however, the participating physicians reported that virtually all of these consultations were for the assessment and management of cardiovascular risk. Fourth, we relied on patient self-report for smoking history and exercise capacity, and some may question the accuracy of patient recall for these factors. However, there seems little reason to believe that patients in this study would systematically overestimate/underestimate their smoking history or exercise capacity compared with patients seen in usual practice, and as such, this "limitation" has little bearing on our findings about which elements of the history reported by patients to their internists/pulmonologists are useful in predicting which patients are at increased postoperative pulmonary risk. Fifth, we restricted our postoperative surveillance to those complications that were irrefutably "pulmonary" and did not record whether any patients who did not fit our definition of "postoperative pulmonary complication" had another complication (such as heart failure or pulmonary embolism), which may have necessitated the use of oxygen therapy in the postoperative period. Finally, preoperative pulmonary function tests or arterial blood gases were not done in all patients and were left to the discretion of the attending physicians looking after the patients: Their ORs may have been overestimated if they were selectively done in higher risk individuals (a theoretical concern [15] but one not supported by our finding that postoperative pulmonary complications were actually more common in patients who did not have pulmonary function tests or arterial blood gases performed). Although these tests do appear to be highly useful in predicting pulmonary risk, the potential expense and risks if these tests were performed in all patients preoperatively fuel the search for clinical prediction rules such as ours that can identify high-risk individuals in whom these tests may be most useful (14). Our study provides information about the accuracy of key elements of the preoperative assessment (history and physical examination) as well as laboratory testing in predicting pulmonary risk in patients referred to internists or pulmonologists. Our method was rigorous in that we collected preoperative data from consecutive patients in a standardized format, and postoperative outcomes were assessed by an independent investigator who was blinded to the preoperative/operative data. Once the pulmonary risk of patients can be accurately assessed preoperatively, interventions for the prevention of pulmonary complications can be efficiently tested in high-risk individuals.
Supported by the Alberta Heritage Foundation for Medical Research (F.A.M., N.A.K., and S.R.M.) and a Career Scientist Award from the Ontario Ministry of Health (S.E.S.) The CARE-PREOP1 Investigators (in order of number of patients enrolled) are as follows: F. McAlister, B. Fisher, V. Wilkinson, S. Majumdar, B. Wirzba, N. Gibson, N. Kassam, S. Wilkinson, R. Leung (Edmonton, AB, Canada); N. Khan, W. Ghali, M. Mintz (Calgary, AB, Canada); M. Papaioakim, N.-N. Papadopoulou (Alexandroupolis, Greece); M. Daniel, T. Quasim (Glasgow, Scotland); O. Gajic; L. Urrutia, B.J. Narr, S.G. Peters (Rochester, MN); S. Straus, A. Lalovic, A. Paus-Jenssen, R. Straus (Toronto, ON, Canada). This article has an online supplement, which is accessible from this issue's table of contents online at www.atsjournals.org Received in original form September 3, 2002; accepted in final form November 12, 2002
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
