INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Many of the risk factors for respiratory tract colonization and nosocomial pneumonia overlap; they include patient-related conditions, infection control-related problems, intervention related alterations in host defenses, and bacterial exposure (1). The presence of endotracheal tubes and duration of this intervention have been associated with the highest risk (2, 3) of developing nosocomial pneumonia, and the procedure of intubation itself increases this risk significantly, as has been demonstrated in patients requiring reintubation (2, 4, 5). Some studies (2, 3, 5) using multivariate techniques have evaluated which risk factors are associated with the development of pneumonia in intubated patients beyond 24 to 48 h. A systematic review of the literature (10) shows that four studies evaluated risk factors in the intensive care unit (ICU) independently of the intubation procedure and most cohort studies were performed in ventilated patients. Recently, we reported (7) that airway management (continuous aspiration drainage and routine intracuff pressure monitoring) played a crucial role in preventing pneumonia in patients intubated between 24 h and 7 d. However, to the best of our knowledge, no information is available about which risk factors influence the development of pneumonia in the immediate postintubation period.

Our hypothesis was that risk factors for pneumonia during the intubation period change in accordance with the length of exposure to risk (i.e., intubation). This assumption is based on the observation that etiologies are distributed heterogeneously throughout the ventilation period, and different risk factors have been documented for each etiology (12, 13). Our objective in this study was to identify the potential risk factors associated with development of pneumonia within the first 48 h of intubation.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Population

In July 1996, a prospective study was begun in the medical/surgical ICU at the Hospital de Sabadell involving the follow-up of all intubated patients within the first 48 h and the identification of patients who met clinical criteria for pneumonia. The study lasted 16 mo. Patients intubated before hospital admission were excluded, as were patients whose follow-up period was less than 24 h. Although the follow-up period comprised the rest of the patient's stay in the ICU, only the first 48 h after intubation were considered for analysis. Several clinical and demographic variables, including age, gender, Acute Physiology and Chronic Health Evaluation II (APACHE II) score, underlying disease, presence of comorbidities or cardiac arrest, clinical signs, days of intubation, type of sedatives, use of corticosteroids, and antibiotic regimens were recorded for each intubated patient on a standardized form. Specific details of the intubation procedure (such as experience of the physician, number of attempts, reason for intubation, presence of blood or gastric contents inside the airways, or emergency/elective procedure) were also recorded. All mechanically ventilated patients had a nasogastric tube in place and received sucralfate. We did not use selective decontamination of the digestive tract or antibiotic prophylaxis (except in patients undergoing surgical interventions that required it). All patients were in the semi-upright position for most of the time.

Definitions

The clinical diagnosis of pneumonia (14) required the detection by chest radiography of a new, persistent pulmonary infiltrate (other than those of noninfectious origin). Chest X-rays were interpreted by a radiologist. At least two of the following criteria were also required: (1) fever  38° C; (2) leukocytosis  10,000 per mm³; (3) purulent respiratory secretions. An additional requirement was that the clinical criteria prior to intubation should rule out the development of nosocomial pneumonia at that time. The diagnosis of pneumonia was prospectively determined by consensus between three of the researchers (J.R., E.D., and J.V.). Fiberoptic bronchoscopic examination using a protected specimen brush or bronchoalveolar lavage was performed in some of these episodes within the first 12 h of the development of a new pulmonary infiltrate, for etiologic identification. In these patients, bronchoscopy, processing of samples, cultures and identification of bacteria were performed as previously described (15).

Acute respiratory distress syndrome (ARDS) was defined according to standard definitions (16). Coma was defined as a score of < 9 on the Glasgow Coma Scale (17). Prior antibiotic use was defined as antibiotic use for longer than 24 h in the two preceding weeks (18). Patients who received continuous sedation (over a 12-h period) were defined as "receiving sedation." Sedation refers to infusions of benzodiazepines, opiates, barbiturates, or propofol. Paralytic agents were evaluated independently.

Statistical Analysis

Continuous variables were compared with t tests and one-way analysis of variance (ANOVA) or the nonparametric counterpart (Mann-Whitney U test). A two-tailed Fisher exact test was used to compare differences between groups for discrete variables.

Independent variables recorded at baseline included age; sex; primary diagnosis; location before admission to the ICU; medical or surgical status; presence of trauma; prior infection; chronic comorbid conditions, such as alcoholism, a smoking history of 10 or more pack- years, asthma, bronchiectasis, pulmonary fibrosis, chronic bronchitis, cancer, cirrhosis, cardiac failure, renal failure, and human immunodeficiency virus (HIV) infection; and recent corticosteroid therapy or chemotherapy. Additional independent variables were classified as intubation variables (elective/emergency procedure, continuous secretion drainage, size of endotracheal tube, intubation attempts, presence of blood in the airway, prior esophageal intubation, change or reinsertion of endotracheal tube, and reason for intubation), or postintubation variables. These included Glasgow Coma Scale score (dichotomized with a cutoff of 9), witnessed large volume aspiration, and drug exposure (infusions of sedative agents, paralytic agents, and antibiotics) categorized as presence/absence. Cardiopulmonary resuscitation was registered if chest compressions were required either during the intubation maneuver or after intubation. The experience of the physician performing the intubation was categorized according to the specialty and years of practice; only anesthesiologists, critical care specialists, and senior fellows (> 1 yr of experience) of these two specialties were categorized as "experienced."

Multiple logistic regression analysis was used to evaluate the impact of potential risk factors on the development of pneumonia, controlling for the remaining variables. In constructing the model, variables were selected with the backward stepwise method, assessing significance by means of the likelihood ratio test. A variable was eligible for entry into a logistic regression model if it was significantly associated with pneumonia (p < 0.05) in the univariate analysis and if at least 10% of the patients exhibited the characteristic. Odds ratios (OR) and 95% confidence interval (CI) were calculated in accordance with the standard methods. Interactions were explored between the substantive variables that remained in multivariate analyses.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A total of 250 intubated patients were followed up in a medical-surgical ICU to identify potential risk factors for developing pneumonia within the first 48 h postintubation. One hundred eighty-one were intubated at the first attempt, 41 required two attempts, and 28 required more than two. Ninety-two other patients were excluded because their follow-up period was less than 24 h (43 died, 41 were extubated, five were transferred to other hospitals, and three presented self-extubation). Four additional patients were not enrolled because intubation was performed outside the hospital, and detailed data on the procedure were not available. All patients received orotracheal (median diameter = 80 mm) intubation. One hundred sixty-eight were medical patients, 65 underwent surgery, and only 17 had trauma. A detailed list of the underlying diseases of the study population is shown in Table 1. The study cohort did not include neutropenic patients. Fifty-five received continuous sedation with midazolam, propofol, or morphine; none received intravenous barbiturates. One hundred sixty-three were male, and median age was 64. 

Thirty-two (12.8%) developed pneumonia during the study period (18 within the first 24 h postintubation). The relationship between the development of pneumonia and the cause of intubation is shown in Table 2. The most frequently isolated organisms were methicillin sensitive Staphylococcus aureus (20%), Haemophilus influenzae (15%), Streptococcus pneumoniae (15%), mixed aerobic/anaerobic flora (15%), and Pseudomonas aeruginosa (15%). Thirty-three procedures were reintubations (21 weaning failures, 11 self-extubations, and one tube obstruction) and four (12.1%) of these patients developed pneumonia within the next 48 h. Intubation caused by respiratory/cardiac arrest or decrease in the level of consciousness was significantly associated with pneumonia (21.6% versus 7.2%, p < 0.01) when compared with other causes. Other variables such as large volume aspiration, emergency procedure, presence of sedation, cardiopulmonary resuscitation (CPR), and Glasgow coma score < 9 after intubation were significantly (p < 0.05) associated with pneumonia in the univariate analysis. In contrast, prior infection and prior antimicrobial use were associated (p < 0.05) with a protective effect. Detailed information on these variables is shown in Table 3. Continuous subglottic secretion drainage was performed in 205 (82.0%) patients; it had no impact (12.7% versus 13.3%, p > 0.20) on the development of pneumonia within the 48 h postintubation. Similarly, the remaining independent variables tested had no significant effect (p > 0.20). One hundred and twenty-seven patients were in the hospital longer than 24 h prior to intubation and the incidence of pneumonia within the 48 h after intubation did not differ when compared with those patients with a shorter period of prior stay (11% versus 14.6%, p > 0.20).

Two patients developed pneumonia in the period comprised between Days 3 and 6 after intubation. One of these episodes was due to mixed aerobic/anaerobic flora and the other was caused by P. aeruginosa. Seven additional episodes of pneumonia (all caused by P. aeruginosa) were identified in patients who underwent longer than 1 wk of intubation. None of the six significant variables associated with an increased risk of pneumonia within 48 h of intubation remained significant when evaluated after this period. After 48 h of intubation, no association was found between development of pneumonia and prior infection. In contrast, prior antimicrobial use was associated (OR = 1.42, p < 0.05) with an increased risk of pneumonia after 48 h of intubation.

The eight significant variables at the univariate analysis were included in the logistic regression model, pneumonia development being the dependent variable. The final model (Table 4) identified CPR (OR = 5.13, 95% CI = 2.14, 12.26) and continuous sedation (OR = 4.40, 95% CI = 1.83, 10.59) as significant risk factors for pneumonia within 48 h of intubation, while antibiotic use (OR = 0.29, 95% CI = 0.12, 0.69) showed a significant protective effect. There was no significant effect modification (interaction) between these variables.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study describes the largest cohort to date of patients with development of pneumonia immediately following intubation. The study presents a number of new findings. First, in our cohort, a previously unreported variable (i.e., CPR) was the most important risk factor for the development of the complication. Second, in contrast with episodes developed later (7), factors related with airway manipulation such as continuous subglottic drainage (19) had no significant effect. Third, antibiotic use protected against early pneumonia, but not against development of the condition at later stages. Indeed, a comparison of this study with earlier publications (5, 8, 10, 18) suggests that different risk factors emerge over the period of intubation. These observations should encourage researchers to investigate whether the factors influence intubated patients during specific periods of their exposure to risk; it may be that the classic approach of grouping all patients together is inappropriate. More importantly, these data suggest strongly that different strategies of prevention should be combined in order to reduce the incidence of the complication.

Since the initial study reported by Craven and coworkers (6) in 1986, other investigators (10, 11) have emphasized the importance of a range of risk factors for the development of pneumonia in different cohorts of intubated patients. However, only one of these studies focused on a specific period of intubation (7), and none included the first 24 h immediately after intubation. Joshi and coworkers (20) included the first 2 d when evaluating risk factors for nosocomial pneumonia in the ICU, but intubated patients represented only 58% of the study population. The present study is the first one to restrict analysis to ventilated patients over the 48 h after intubation and we found that a previously unreported variable (CPR) was the most important single determinant of pneumonia. Indeed, prior observations (21) have reported a very high incidence of pneumonia in patients after cardiac arrest and CPR. In spite of this, this is the first study to look for this association using regression methods. Like patients suffering head trauma (22), this subpopulation may benefit from short-term antibiotic prophylaxis; our current findings suggest that this possibility should be investigated in further clinical trials. Multivariate analysis was able to demonstrate that association between CPR and pneumonia was not the result of residual confounding due to the presence of large volume aspiration or decreased level of consciousness. However, many of these cases of pneumonia, including those occurring in the setting of continuous sedation, may actually be associated with smaller episodes of gastric aspiration which may go undetected. More sophisticated methods for detecting the presence of aspiration in the peri-intubation period might have identified this as a more important risk factor.

In addition, our model indicates that continuous sedation with midazolam, propofol, or morphine has an independent harmful effect (OR = 4.40, 95% CI = 1.83, 10.59). This agrees with other reports suggesting that profound sedation has adverse effects on local airway defenses, and increases the incidence of pneumonia in ventilated patients (23). The identification of continuous sedation was independent of the use of muscle relaxants. Although a recent study (11) identified paralytic agents (OR 1.57) as an independent risk factor for late-onset ventilator-associated pneumonia (VAP), paralytic agents could be a confounder of using sedative agents because all patients receiving paralytic agents require profound sedation and this variable was not investigated in their model, whereas this association has not been demonstrated in other studies (6, 23).

Interestingly, we found that subglottic secretion drainage, which prevented pneumonia in later episodes (until 10 d after intubation) (19), had no significant effect (p > 0.20) on the first 48 h postintubation. This is probably because the risk of small aspirations of subglottic secretions is low, and because it takes the risk accumulated over more than 1 d (7) to produce a clinically relevant effect. The current study is the first to evaluate in detail a number of variables associated with the intubation procedure, particularly those related to the procedure's technical difficulties and the skill of the person performing it. However, excepting the presence of large volume aspiration (which was not selected as an independent variable in the multivariate model), none of these variables were associated with pneumonia. These observations suggest that although airway manipulation predisposes to the development of infection in episodes that develop later (7), it does not play a role in the development of the complication in the immediate postintubation period.

Even though earlier work (2, 5, 8, 18, 24) has emphasized the importance of the association between previous antibiotic use and development of pneumonia, in this study we observed that exposure to antibiotics independently prevented development of pneumonia during the first 2 d of ventilation. Indeed, other studies (7, 22) have reported the protective effect of antibiotics specifically on episodes caused by primary endogenous flora. In contrast, most episodes caused by multiresistant pathogens developed later and these patients were usually receiving prolonged broad-spectrum antibiotics (25). We demonstrated that the apparent protective effect of antibiotics within the first 48 h of intubation disappears when the intubation period was prolonged and most of the late episodes were caused by P. aeruginosa. We did not systematically collect data on specific antimicrobial agents administered. Thus, a potential association between some prescribing practices and development of pneumonia cannot be explored. A recent Canadian study (11) found antibiotic administration to be associated with lower rates of VAP. These investigators were the first to demonstrate that modeling the effect of antibiotic administration over time shows that the apparent protective effect of antibiotics attenuates as the time in the ICU increases. When comparing these findings with our current observations and the information previously reported in the literature (10), the administration of antibiotics illustrates the fact that an intervention can have a bimodal effect, i.e., that its effect depends on the moment of exposure.

All these observations provide further evidence that the pathogenesis of pneumonia in intubated patients is complex and cannot be simplified to a uniform model; several patterns coexist in a dynamic fashion, and it is probably fair to say that the key role of time has been largely underestimated in previous studies. Interestingly, only one variable (antibiotic exposure) that influenced the period prior to intubation was selected in the multivariate model. This information illustrates that VAP is more related to postintubation events than to preintubation factors or intubation related variables, even when early episodes are accounted. When combined with prior reports (10, 19), our current findings emphasize that risk factors for developing pneumonia change over the period of intubation, at least during the first days of ventilation. Consequently, we suggest that different strategies of prevention should be combined to effectively reduce this complication.

Finally, some limitations of the study should be noted. First, comparisons of early- and late-onset pneumonia were only evaluated for the eight variables associated with a significant modification of the risk of pneumonia within 48 h of intubation, but it was not possible to compare the remaining nonsignificant variables in the same population. Second, the relatively small sample size may not have sufficient statistical power to identify all potentially important risk factors for the development of early-onset pneumonia. Nonetheless, the magnitude of the increased risk of these factors would be substantially lower than with the variables selected in the present study. Third, most patients (> 80%) received continuous subglottic drainage with routine monitoring of intracuff pressure, and these interventions may underrate the risk associated with airway manipulation-related variables. Similarly, another potential limitation of this study was that it was performed in a single center, which raises the possibility of institutional bias either in patient selection or in other institutional practices. In addition, we used the duration of intubation as a clinical variable to categorize each episode of VAP, whereas the potential impact of hospitalization prior to intubation was not investigated. Finally, the diagnosis of pneumonia was based on clinical grounds and we cannot completely rule out the possibility that diagnosis is subject to changing observer bias. However, alternative diagnoses were reasonably excluded, and restricting cases only to episodes confirmed by more specific methods would represent an unacceptable bias owing to its limited sensitivity; in addition, we used similar laboratory standards and clinical criteria to those used in previous studies (2), thus facilitating comparison of risk factors with other studies focusing on intubated patients.

In summary, our current findings support our hypothesis that risk factors for pneumonia in intubated patients change over time. Further studies should test this hypothesis in larger study populations. This observation can explain why risk factors for pneumonia differ from series to series, and even why different researchers should find certain variables (e.g., antibiotic use) to have directly opposite effects. As a consequence, we suggest that the time factor should be considered when stratifying patients and targeting VAP prophylaxis. Finally, all these findings reinforce the belief that the prevention of pneumonia requires a combined approach.