Assessment of a Multifactorial Risk Index for Predicting Postoperative Pneumonia after Open Abdominal Aortic Aneurysm Repair

Abstract

Pneumonia is a common postoperative complication, with a mortality of up to 40%. The Post-operative Pneumonia Risk Index (PPRI) was derived from a large cohort of general surgical patients but has not been validated in patients undergoing open abdominal aortic aneurysm (AAA) repair. The PPRI was applied to patients undergoing elective open AAA repair in a tertiary referral vascular unit. Pneumonia occurred in 20% of patients. Receiver operating characteristic curve analysis identified 36 as the optimum PPRI cutoff value. At this cutoff, the likelihood ratio for pneumonia was 1.35 (95% confidence interval 1.08–1.62). However, in a multivariate analysis, only weight loss in excess of 10% over the preceding 6 months was an independent predictor of postoperative pneumonia. Although the PPRI is of some value in identifying high-risk patients undergoing AAA repair, weight loss alone may be predictive, allowing targeted preventive measures in aneurysm patients at increased risk.

Keywords

postoperative complications/diagnosis risk assessment

Pulmonary complications are a common cause of postoperative death.¹ Pneumonia is the third most common postoperative infection, affecting between 9 and 40% of postoperative patients, with a mortality of between 30 and 40%.^2,3 A single retrospective study of patients undergoing major aortic surgery concluded that pulmonary risk factors were as important as cardiac risk factors in determining suitability for open aortic surgery.⁴ Identification of patients at high risk of this complication would facilitate preoperative counseling and allow intensive preventive measures to be taken.

A multifactorial risk index has been developed in patients undergoing major noncardiac surgery.⁵ The Post-operative Pneumonia Risk Index (PPRI) was derived and validated using a cohort of over 300,000 patients undergoing major noncardiac surgery in 100 Veterans' Administration hospitals between 1997 and 1999.⁵ Patients with preoperative pneumonia and those who developed pneumonia after a period of ventilator dependence were excluded. The PPRI assigns points in 14 different categories, the sum of which yields the final score. Pneumonia developed in 16% of patients with a score greater than 55. The authors suggested that the score could be used to target preventive measures on those patients at the highest risk. Open abdominal aortic aneurysm (AAA) repair was identified in the derivation cohort as the highest risk procedure for postoperative pneumonia. However, the discriminatory value of the score within the aneurysm subgroup was not investigated. We aimed to assess the utility of this scoring system for predicting pneumonia after elective open AAA repair.

Methods

The components of the PPRI are summarized in Table 1. The score was applied to a cohort of patients undergoing elective open AAA repair in a tertiary referral vascular unit between 1998 and 2005. The unit maintains a prospective database of patients undergoing aneurysm repair. An outcome of pneumonia is recorded in the database if a patient develops pyrexia, productive cough, a raised white cell count, and localizing signs on chest examination or chest radiography. All patients received general anesthesia. Midline incisions were used in all patients. Epidural analgesia was used preferentially, whereas nasogastric drainage was used selectively. At the end of the procedure, patients were extubated and transferred to a fast-track intensive care facility overnight, as reported previously.⁶

Table 1.

Components of the Post-operative Pneumonia Risk Index

Risk Factor	Point Value Assigned
Type of surgery Abdominal aortic aneurysm repair Thoracic Upper abdominal Neck Neurosurgery Peripheral vascular	151410838
Age (yr) > 80 70–79 60–69 50–59	171394
Functional status Totally dependent Partially dependent	106
Blood urea (mmol/L) < 2.86 7.85–10.7 > 10.7	423
Weight loss > 10% in 6 mo	7
History of chronic obstructive pulmonary disease	5
General anesthesia	4
Impaired sensorium	4
Previous cerebrovascular accident	4
Transfusion > 4 units	3
Emergency surgery	3
Chronic steroid use	3
Current smoker	3
Alcohol > 2 drinks/d	2

The PPRI was calculated for each patient in the database. A receiver operating characteristic (ROC) curve was constructed to identify the optimum cutoff value for the prediction of postoperative pneumonia. The area under the curve (AUC) was calculated using the extended trapezoidal rule. A likelihood ratio and positive and negative predictive values were calculated for the optimum cutoff. Potential predictors of postoperative pneumonia, including the PPRI, were tested in univariate logistic regression models. Variables with a p value ≤ 0.1 in the univariate models were then entered into a reverse stepwise multivariate model to test for independence. The significance level was set at 5% in the multivariate model. The statistical analysis was performed using Statsdirect 2.5.7 (Statsdirect Ltd., Altrincham, UK).

Results

The PPRI was applied to 202 patients who underwent elective open AAA repair. Postoperative pneumonia occurred in 41 patients (20%). PPRI scores ranged from 19 to 58 (mean 39). No patient with a PPRI score less than 26 developed pneumonia. The incidence of pneumonia increased in proportion to the PPRI (Figure 1). ROC curve analysis identified a PPRI of 36 as the optimum cutoff point for the prediction of pneumonia in the study cohort (Figure 2). At this cutoff, the PPRI had a sensitivity of 0.81 (95% confidence interval [CI] 0.65–0.91) and a specificity of 0.41 (95% CI 0.33–0.49) for postoperative pneumonia. The likelihood ratio was 1.35 (95% CI 1.08–1.62). A PPRI of 36 or greater had a positive predictive value of 26% (95% CI 18–34) and a negative predictive value of 89% (95% CI 79–95). Potential individual predictors of postoperative pneumonia were assessed by means of univariate and multivariate logistic regression models (Table 2). Based on the results of univariate analyses, three predictors (PPRI, weight loss greater than 10% in 6 months, and age) were entered into a reverse stepwise multivariate model to test for independence. Only weight loss retained independent significance in this model.

Figure 1.

Incidence of pneumonia according to Post-operative Pneumonia Risk Index (PPRI) group.

Figure 2.

Receiver operating characteristic curve for the Post-operative Pneumonia Risk Index. The area under the curve was 0.60 (95% confidence interval 0.50–0.67).

Table 2.

Predictors of Postoperative Pneumonia

Variable	Univariate p	Multivariate p	Odds Ratio (95% CI)
PPRI	0.09	0.68	1.02 (0.95–1.09)
Alcohol > 2 drinks/d	0.50
Current smoker	0.52
Chronic steroid use	0.27
Transfusion > 4 units	0.61
Blood urea nitrogen	0.54
Previous CVA	0.91
Impaired sensorium	0.35
History of COPD	0.16
Weight loss > 10% in 6 mo	0.05	0.05	3.41 (0.97–11.98)
Functional status	0.91
ASA grade	0.34
Age	0.07	0.10	1.04 (0.99–1.09)

ASA = American Society of Anesthesiologists; CI = confidence interval; COPD = chronic obstructive pulmonary disease; CVA = cerebrovascular accident; PPRI = Post-operative Pneumonia Risk Index.

Discussion

Major respiratory complications occur in about 16% of open AAA patients.⁴ Pneumonia occurs in about 9%.⁴ Aneurysm patients typically have a number of factors predisposing them to postoperative pneumonia. These include a history of cigarette smoking, long midline incisions, prolonged surgery requiring intubation, and cough reflex suppression. These factors lead to basal collapse, pooling of respiratory secretions, and an increased risk of pulmonary morbidity in the postoperative period. Given the associated mortality, identification of patients at high risk of postoperative pneumonia is important to allow prompt preventive measures to be taken. Whereas cardiac risk scores are well established, risk scores for other postoperative complications are less well established.

The PPRI was derived from a large cohort of patients undergoing a broad range of surgical procedures.⁵ Although the model was validated on a random sample of patients drawn from the same population, it has not been validated in specific patient groups. Open AAA repair is assigned the highest operative score in the PPRI, reflecting the high risk of pneumonia in this patient group. However, the score has never been validated in a homogeneous population of open aneurysm patients. Ideally, the score should also be assessed in patients undergoing endovascular aneurysm repair (EVAR). However, of 210 patients in our unit's EVAR database, only a single patient to date has developed pneumonia, rendering assessment of the PPRI impossible in this group.

Our data demonstrate that the PPRI is of some value in patients undergoing open AAA repair. A score of 36 or greater had a sensitivity of 81% for predicting pneumonia, although this sensitivity comes at the cost of a low specificity (40%). The PPRI did not independently correlate with pneumonia in our cohort. The only element of the PPRI independently associated with pneumonia was weight loss in excess of 10% in the previous 6 months. Weight loss had a sensitivity of 45% but a much higher specificity (81%). The likelihood ratio for pneumonia in patients with weight loss was 2.41 (95% CI 1.08–4.29). On the basis of these results, it may be more practicable to use a history of weight loss alone as a risk predictor in aneurysm patients.

Various strategies are available to minimize the risk of postoperative respiratory complications following laparotomy. A recent systematic review concluded that there was strong evidence in favor of lung expansion interventions, deep breathing exercises, and continuous positive pressure ventilation as preventive measures.⁷ In addition, there was moderate evidence that a selective policy of nasogastric intubation and the use of short-acting neuromuscular blockade may also be of benefit.⁷ Notably, although malnutrition increases the risk of pulmonary morbidity, no evidence was found to support the use of routine enteral or parenteral feeding preoperatively. There may be some benefit to immunonutrition.⁷ Transverse abdominal incisions are associated with a reduction in postoperative pain and pulmonary complications following AAA repair.⁸ Our unit did not use transverse incisions before 2005, although they are now used on a selective basis.

Scoring systems derived from heterogeneous patients in one geographic area do not always prove useful elsewhere. Our unit previously demonstrated that the POSSUM, V-POSSUM, and P-POSSUM equations had poor predictive power in a homogeneous aneurysm population in a geographic location different from the original derivation and validation cohorts.⁹ In fact, a completely new POSSUM formula had to be derived locally. The PPRI suffers from similar limitations. A history of weight loss alone in AAA patients has similar predictive value and is easier to use in the clinical setting than a 14-element scoring system. Although the PPRI may not enhance an individual surgeon's clinical assessment, it may be of some use as a risk stratification tool in controlled trials of interventions to prevent pneumonia following aneurysm repair. However, given the apparent limitations of the PPRI in AAA patients, it would be valuable to derive an aneurysm-specific score from a large cohort, such as the UK National Vascular Database. In the meantime, vascular surgeons should be aware that weight loss preceding aneurysm repair is an adverse prognostic indicator. Such patients would benefit from close observation, incentive spirometry, and early use of continuous positive pressure ventilation.