Impact of Unplanned Extubation and Reintubation after Weaning on Nosocomial Pneumonia Risk in the Intensive Care Unit: A Prospective Multicenter Study

A prospective study in a multicenter database (OUTCOMEREA®) was conducted during a 2-yr period in six French medical or surgical ICUs. Starting in January 1997, all patients who were older than 16 yr, had an ICU stay length longer than 48 h, and received mechanical ventilation at any time during the ICU stay were entered in the study. Patients who were tracheostomized at ICU admission were excluded from the analyses.

Data were collected daily by senior physicians in each participating unit. For each patient, an investigator used these data to complete a case report form. All codes and definitions were established prior to study initiation. Before entry into the database, each case report form was reviewed by a member of the steering committee.

The case report forms contained no information identifying the patient. They were used to record age, sex, and admission category (medical, scheduled surgery, or unscheduled surgery); underlying diseases according to SAPS II definitions 17; other comorbidities including respiratory, cardiovascular, hepatic, renal, and immunosuppression using APACHE II definitions 18; and severity of illness at ICU admission and daily during the ICU stay as measured using the SAPS II and LOD system. 19 

All patients were screened daily throughout the ICU stay for suspected nosocomial pneumonia and for self-extubation, accidental extubation, or reintubation after weaning. Self-extubation was defined as deliberate extubation by a patient for whom the intensivists considered that intubation was beneficial, accidental extubation as inadvertent extubation by an ICU caregiver during procedures at the bedside, and reintubation after weaning as a need for reintubation within 48 h after planned extubation. The need for noninvasive mechanical ventilation alone was not considered as an extubation failure.

The sedative regimen used in the overwhelming majority of patients included a combination of a benzodiazepine and an opioid. The level of sedation was routinely titrated to achieve a Ramsay score of 2–3 and was tailored to the needs of each patient depending on the presenting condition. 20,21In patients with clinical improvement in the initial cause of their respiratory failure, four widely accepted criteria for a trial of extubation were evaluated daily: a ratio of the partial pressure of arterial oxygen over the fraction of inspired oxygen above 200, a positive end-expiratory pressure below 5 cm H²O, adequate coughing during suctioning, and absence of infusion of sedatives with a Glasgow coma score greater than 14. 22In addition, in the overwhelming majority of patients subjected to a trial of extubation, the ratio of respiratory frequency over tidal volume was equal to or lower than 105 breaths per minute per liter. 23Planned extubation was performed only after a successful T-tube trial of 20–30 min.

The sedated Glasgow coma score was defined as the value truly observed, including effect of sedation, and the expected Glasgow coma score was defined as the value of the score expected if the patient was not receiving sedative drugs.

Clinical suspicion of nosocomial pneumonia was based on the criteria described by Andrews et al.  24Nosocomial pneumonia was defined as pneumonia occurring after at least 48 h of mechanical ventilation, according to the recommendations of the First International Consensus Conference on the Clinical Investigation of Ventilator-Associated Pneumonia. 25The diagnosis of nosocomial pneumonia was based on a chest radiograph showing new pulmonary infiltrates not otherwise explained, positive results of quantitative cultures of a plugged telescopic catheter or protected-specimen brush specimen (> 10³cfu/ml) or of a bronchoalveolar lavage specimen (> 10⁴cfu/ml), and at least two of the following criteria: fever greater than 38°C, purulent bronchial secretions, or peripheral leukocyte count greater than 10,000/mm³.

Finally, length of mechanical ventilation, length of ICU and hospital stays, and outcomes at ICU and hospital discharges were recorded.

Results were expressed as numerical values and percentages for categorical variables, and as medians and quartiles (25th–75th percentile) for continuous variables. Comparisons were based on the Fisher exact test or chi-square test for categorical data and on paired Wilcoxon tests or Kruskal-Wallis tests for continuous data where appropriate. The relation between iatrogenic events and nosocomial pneumonia was computed using a Cox model with a time-dependent covariate. Time to nosocomial pneumonia was defined as the interval from mechanical ventilation initiation to the first nosocomial pneumonia episode even if nosocomial pneumonia occurred after successful extubation. In patients who had nosocomial pneumonia episodes before and after unplanned extubation and reintubation after weaning (n = 5), this time was defined as the interval between mechanical ventilation initiation and the nosocomial pneumonia episode after event. Nosocomial pneumonia episodes that occurred before unplanned extubation or reintubation after weaning were entered into the model as a dichotomous variable. Patients who had no nosocomial pneumonia episodes were censored at ICU discharge. Time to the first self-extubation, accidental extubation, or reintubation after weaning, and time to the first nosocomial pneumonia episode were computed using Kaplan-Meier estimates.

Time to occurrence of the first event (unplanned extubation or reintubation after weaning) and each category of event were successively introduced into the Cox model as time-dependent covariates. In addition, the following variables were studied: SAPS II and LOD on the first mechanical ventilation day, occurrence of pneumonia before mechanical ventilation or within 48 h after mechanical ventilation, enteral feeding, antiacid and antibiotic therapy within 48 h after mechanical ventilation, Knaus comorbidities, main reasons for ICU admission, and whether the patient was transferred to the ICU from a ward. These variables have been reported to affect the risk of mechanical ventilation-related pneumonia. 26A stepwise forward procedure was used to select variables for entry into the model. Results at the last step are reported.

Factors associated with hospital mortality were evaluated using a Cox model in which unplanned extubation or reintubation after weaning were entered as a time-dependent covariate in addition to the following patient characteristics at ICU admission: SAPS II and LOD, Knaus comorbidities, reasons for ICU admission, and whether the patient was transferred from a ward. Stepwise forward selection was used. At the last step, unplanned extubation and reintubation after weaning were forced into the model to allow an estimation of its effects adjusted on patient characteristics.

All statistical tests were two-tailed, and P  values less than 5% were considered significant. Statistical tests were performed using the SAS 8.00 (SAS Inc., Cary, NC) and S-plus 2000 (MathSoft Inc., Seattle, WA) software packages for personal computers.

During the 2-yr study period, 1,227 patients were admitted for longer than 48 h to the participating ICUs. During their ICU stay, 750 patients required conventional mechanical ventilation for more than 24 h, including 567 (75.6%) who were ventilated on ICU admission. The median duration of ventilation was 7 days (25th–75th percentile: 3–14 days). Overall, 7,953 patient-ventilation days were investigated, excluding days of ventilation through a tracheostomy and days of noninvasive ventilation.

Of the 750 ventilated patients, 105 (14.0%) experienced 131 events, with 51 self-extubations, 24 accidental extubations, and 56 reintubations after weaning, yielding incidence density rates of 6.4, 3.0, and 7.0 per 1,000 intubation days, respectively. Twenty-three patients experienced more than one category of event. The first event was self-extubation in 38 patients, with a median time to occurrence since mechanical ventilation initiation of 3 days (25th–75th percentile: 2–6); accidental extubation in 22 patients, with a median time of 3 days (25th–75th percentile: 1–4); and reintubation after weaning in 45 patients, with a median time of 8 days (25th–75th percentile: 6–11) (fig. 1).

The 105 patients with unplanned extubation or reintubation after weaning and the 645 patients without unplanned extubation or reintubation after weaning (controls) differed regarding the main causes of ICU admission and the presence of three comorbidities, namely, respiratory, cardiovascular, and immunosuppression (table 1). The SAPS II and LOD scores on the day before events were significantly higher in the accidental extubation group, whereas the Po²/Fio²ratio and the sedated Glasgow coma score were significantly lower (table 2).

When we recorded the maximum daily SAPS II and LOD score values during the 2 days preceding and the 3 days following the event, we found a significant increase in these values only in the reintubation after weaning subgroup, in which the maximum daily scores were significantly higher after the event than on the day before the event (P < 0.01 and P = 0.03 for the SAPS II and LOD, respectively).

Five patients with unplanned extubation or reintubation after weaning (3 planned extubation failure and 2 self-extubations) were treated with noninvasive mechanical ventilation.

Reintubation within 48 h after the event was required in all accidental extubations but in only 24 of 38 (63.2%) self-extubations. Among the 91 patients who underwent reintubation, including reintubation after weaning, six (5.7%) were tracheostomized within 48 h after event (reintubation after weaning: 1; accidental extubation: 1; and self-extubation: 4).

Among the 750 patients, 125 (16.7%) acquired 163 episodes of nosocomial pneumonia (table 3). The crude incidence of nosocomial pneumonia was significantly higher in the patients with unplanned extubation or reintubation after weaning (36 of 105; 34.3%) than in the controls (89 of 645; 13.8%) (P < 0.01). This significant difference persisted after exclusion of nosocomial pneumonia episodes that preceded the first event (29 of 105; 27.6%;P = 0.0003).

Nosocomial pneumonia after event was significantly more common in the accidental extubation subgroup (11 of 22; 50%) than in the reintubation after weaning subgroup (11 of 45; 24.4%) and self-extubation subgroup (7 of 38, 18.4%) (P = 0.02). When introduced in a Cox model as a time-dependent covariate, unplanned extubation and reintubation after weaning increased the risk of nosocomial pneumonia (relative risk, 1.80; 95% confidence interval, 1.15–2.80;P = 0.009) even after adjustment on SAPS II and LOD on the first mechanical ventilation day, enteral feeding, gastric protection by antihistamine type 2 receptors and proton pump inhibitors, and antimicrobial therapy within 48 h after mechanical ventilation initiation (table 4, which shows only significant variables).

When we studied these three categories of events separately, we found that the increase in nosocomial pneumonia risk associated with unplanned extubation and reintubation after weaning was entirely attributable to an association with accidental extubation (relative risk, 5.28; 95% confidence interval, 2.83–9.89;P < 0.01;table 3). Conversely, no associations with nosocomial pneumonia were found for reintubation after self-extubation (relative risk, 1.83; 95% confidence interval, 0.85–3.96;P = 0.12) or reintubation after weaning (relative risk, 1.35; 95% confidence interval, 0.7–2.61;P = 0.36). Figure 2shows that accidental extubations are at earlier risk of nosocomial pneumonia and that the proportion of patients with nosocomial pneumonia is higher with accidental extubations.

Total mechanical ventilation duration was significantly longer in the unplanned extubation and reintubation after weaning groups (median, 17 days; 25th–75th percentile: 8–24) than in the control group (median, 6 days; 25th–75th percentile: 3–12) (P < 0.01). Significant differences in the same direction were found for ICU and hospital length of stay (table 4). Moreover, mechanical ventilation duration was significantly higher in the subgroup with reintubation after weaning than in the other two subgroups. Tracheostomy was performed in a larger proportion of patients with unplanned extubation or reintubation after weaning (22.9%; reintubation after weaning: 13; accidental extubation: 5; and self-extubation: 6) than of controls (6.0%; n = 39) (P < 0.01). The overall additional mechanical ventilation durations after the first event were 11 days (range, 7–16 days), 7 days (range, 5–13 days), and 4 days (range, 1–13 days) for reintubation after weaning, accidental extubation, and self-extubation, respectively.

Unplanned extubation or reintubation after weaning did not influence ICU or hospital mortality rates (table 4), even after adjustment on SAPS II at ICU admission, comorbidities, and reasons for ICU admission.

Self-extubation, accidental extubation, and reintubation after weaning are common in the ICU. These three categories of events put patients at significant risk of morbidity and can contribute to death. 5Although unplanned extubation or reintubation after weaning have been extensively investigated since 1990, most studies focused on incidence, risk factors, prevention, and prediction of early reintubation. Data on adverse events, including nosocomial pneumonia, are sparse. Adverse event rates after unplanned extubation have ranged from 5% to 28%, 27and life-threatening complications have been reported. All studies of the outcome of unplanned extubation used a retrospective case-control design. 7,8,15,16We report the first large, prospective, multicenter study evaluating the occurrence and outcomes of all three categories of events, with the primary objective of looking for nosocomial pneumonia rate differences possibly related to differences in reintubation rates or to differences in the circumstances surrounding extubation.

The limitations of this study need to be discussed. First, the use of continuous intravenous sedation is associated with the prolongation of mechanical ventilation and a greater incidence of reintubation, and could be a risk factor for nosocomial pneumonia. 28We did not use daily interruption of sedative drug infusion as suggested by Kress et al. , 29since the results of their study had not been published at the time the data on our patients were collected. Sedative drug interruption may influence the rate of unplanned extubation. Second, the ICU patients in the present study were all hospitalized for longer than 48 h. Consequently, the self-extubation subgroup was not representative of the overall population with this event, which often occurs shortly after ICU admission of suicidal or postsurgical patients. This probably led us to underestimate the proportion of self-extubated patients who did not require reintubation. Third, among patients with unplanned extubation, those with accidental extubation were significantly more likely to require reintubation than those with self-extubation. 4The rate of accidental extubation was higher in our unplanned extubation group (37%) than in other series (0–22%). 2–6,15,27,30This explains why the overall reintubation rate among patients with unplanned extubation (77%) was at the upper end of the range reported in the literature (30–88%). 2–4,6,15,27,30,31Fourth, the Consensus Conference criteria from 1993 that we used for diagnosis of ventilatory associated pneumonia could be viewed as out-of-date. However, all patients were similarly screened for diagnosis of nosocomial pneumonia.

In our population, incidence rates were 4.8% for self-extubation and 2.8% for accidental extubation (7.6% for unplanned extubation). These rates are similar to the 3–16% rates of self-extubation found in other studies. 3,4,6,15,30,32–35Our incidence rate of reintubation after weaning was 6%, which is within the 2–19% range reported in the literature. 7–14,22This wide range reflects differences in the time to reintubation used to define reintubation after weaning (from within 24 h to within 72 h of planned extubation).

Recent studies have measured the frequency of unplanned extubation as incidence densities. Values ranged from 1.6 to 40 per 1,000 mechanical ventilation days, which is in agreement with our value of 9.4. 1,2,15,31,35–37 

In agreement with most previous studies, we found no increase in mortality in the patients who experienced unplanned extubation or reintubation after weaning, as compared with the controls. 3,15,27,30,33,38,39The recent literature found that reintubation was associated with an increase in mortality rate. 4,7,11,31In the present study, the mortality rate was nonsignificantly higher in the patients with accidental extubation or reintubation after weaning, all of whom required reintubation. Several recent series found that mortality was significantly higher in patients with reintubation after weaning, 7,8,10,11ranging from 32% to 43%, in keeping with our findings; mortality was only 4–12% in successfully extubated patients. 7,8,10,11,16 

Our study hypothesis was that the rate of nosocomial pneumonia would vary across these three categories as a result of differences in reintubation rates or in the circumstances surrounding the previous extubation. Reintubation can introduce colonized oropharyngeal secretions into the lower airways. 40Torres et al.  16reported that reintubation was among the most significant risk factors for nosocomial pneumonia, with a sixfold risk increase. However, these studies failed to distinguish between the risk of nosocomial pneumonia associated with reintubation per se  and the risk associated with the circumstances of extubation. Our findings support a role for these circumstances: although we found nosocomial pneumonia rate differences across categories, those differences did not mirror reintubation rates. The reintubation rate was far lower in the self-extubation category, and the rate of nosocomial pneumonia was also lower in these patients. However, all patients with reintubation after weaning or accidental extubation required reintubation, but nosocomial pneumonia was far more common in the accidental extubation category. Several factors may explain these differences. The pathogenesis of nosocomial pneumonia is related to aspiration of pharyngeal contents or gastric content 41into the distal airways. Factors that may affect the risk of nosocomial pneumonia after unplanned extubation or reintubation after weaning include altered glottic function after several days of intubation, trauma related to the inflated tube cuff in unplanned extubation, prior level of sedation, level of consciousness after extubation, 42presence of a nasogastric tube during the extubation period, 43body position, and time spent extubated. 16Reintubation after weaning occurred when the patients were awake, whereas in our accidental extubation subgroup alterations in consciousness were present in most patients (median sedated Glasgow coma score, 6) and the median Pao²/Fio²ratio was significantly lower than in the other two groups. Both parameters have been found to be associated with reintubation. 4Although by definition reintubation is always performed after reintubation after weaning, removal of the tube in this situation is done under optimal conditions, including stomach emptying, previous cessation of enteral feeding, careful airways suctioning, deflation of the tube cuff, and close monitoring. Most studies of nosocomial pneumonia rates in patients with unplanned extubation 7,8,10,11,16found no significant association 3,5,7,27but included few 7,27or no 3,5patients with accidental extubation. Accidental extubations may occur in sicker patients with heavier workload and therefore more opportunities of manipulations leading to accidental events. However, in the Cox model we used, adjustment on severity and organ failure scores has been performed. Therefore, this study shows clearly that accidental extubation remains an independent predictor of nosocomial pneumonia, adjusted on patients severity.

Total mechanical ventilation duration was significantly longer in our self-extubation, accidental extubation, and reintubation after weaning subgroups than in the controls, and all three events were significantly associated with longer ICU and hospital stays. Moreover, although the self-extubated patients were less likely to undergo reintubation, their mechanical ventilation length, ICU stay length, and hospital stay length were significantly longer than in the controls. In keeping with our findings, Epstein et al.  7reported than mechanical ventilation duration was about 2 weeks longer in patients who underwent reintubation. In contrast, in a case-control study by Chevron and coworkers, 4total mechanical ventilation length was significantly lower in patients with than without unplanned extubation (9 days and 16 days, respectively). This discrepancy may be ascribable in part to the low reintubation rate (37%) in the study by Chevron et al.  4and in part to exclusion from our study of patients with ICU stays shorter than 48 h. Nevertheless, it has been suggested that unplanned extubation not followed by reintubation indicates that planned extubation was overdue and that this may translate into a higher rate of ventilator-related complications, a longer ICU stay, and higher costs. 34 

Given that unplanned extubation and reintubation after weaning were associated with longer durations of mechanical ventilation and hospital stay, these events are markers for suboptimal quality of care. Both reintubation after weaning (with its risk of health status deterioration) and accidental extubation (with its risk of nosocomial pneumonia) can be reduced by improving the quality of care. Self-extubation is also partly preventable. 1Initiation of a continuous quality improvement program was followed by a decrease in the self-extubation rate from 2.6% to 1.2% per patient-ventilator day. 36It has been suggested that unplanned extubation should be viewed as quality-control issues. 44These events correspond to different facets of suboptimal care. Accidental extubation may be related to less-than-ideal nursing care. In a recent study, accidental extubation was the fourth most common incident observed in ICUs with nurse understaffing. 45In a study that used the PRN^REAscore, Chevron et al.  4demonstrated that accidental extubation was not related to the nurse workload, suggesting that this event may be a consequence of faulty nursing care and could be used as an indicator for nursing care quality. Self-extubation without subsequent reintubation may indicate that intubation was maintained too long because the intensivists underestimated the patient's ability to tolerate weaning; reintubation after weaning may reflect overestimation of this ability. Attempts to prevent self-extubation may result either in premature planned extubation responsible for an increase in the rate of reintubation after weaning or in increased use of deep sedation responsible for an increase in mechanical ventilation duration. 46In both cases, mechanical ventilation duration would be increased. Whether clinical judgment used alone or in combination with mechanical respiratory indices, such as frequency/tidal volume, is effective in predicting successful extubation remains controversial. 10,12–14,23,47Continuous quality control programs should evaluate clinical procedures, including clinical indicators, predictive weaning indexes based on pulmonary function parameters, and daily routine evaluation of criteria for a trial of extubation, the goal being to extubate patients as soon as possible but not too soon. 13,14,22,29,47It would be of interest to evaluate both self-extubation and reintubation after weaning as part of a continuous quality improvement program to determine the ratio of self-extubation over reintubation after weaning that minimizes ventilator days, length of stay, and mechanical ventilation-related complications. To be relevant, a quality indicator should be a common event that is associated with morbidity, easy to detect prospectively, and amenable to prevention. 48Unplanned extubation and reintubation after weaning meet all of these requirements and consequently deserve evaluation as quality indicator for quality improvement programs.

The following is the list of people who belong to the OUTCOME- REA® Study Group, including all authors of the present manuscript. Scientific committee: Jean-François Timsit, M.D., Staff Physician, ICU, Hôpital Bichat, Paris, France; Pierre Moine, M.D., Ph.D., Staff Physician, DAR, Hôpital Lariboisiere, Paris, France; Arnaud de Lassence, M.D., Staff Physician, ICU, Hôpital Louis Mourier, Colombes, France; Élie Azoulay, M.D., Staff Physician, ICU, Hôpital Saint Louis, Paris, France; Yves Cohen, M.D., Professor, ICU, Hôpital Avicenne, Bobigny, France; Maïté Garrouste-Orgeas, M.D., Staff Physician, ICU, Hôpital Saint Joseph, Paris, France; Jean-Philippe Fosse, M.D., Staff Physician, ICU, Hôpital Avicenne, Bobigny, France; Lilia Soufir, M.D., Staff Physician, ICU, Hôpital des Diaconesses, Paris, France; Jean-Ralph Zahar, M.D., Staff Physician, Infectious Diseases Deparment, CH François Quesnay, Mantes la Jolie, France; Christophe Adrie, M.D., Staff Physician, ICU, Hôpital de la Fontaine, Saint Denis, France; Gilles Troché, M.D., Staff Physician, Hôpital A. Beclere, Clamart, France; Jean Carlet, M.D., Staff Physician, ICU, Hôpital Saint Joseph, Paris, France; François L'Heriteau, M.D., Attending Physician, ICU, Hôpital Bichat, Paris, France. Biostatistical and informatical expertise: Sylvie Chevret, Ph.D., Professor, DBIM, Hôpital Saint Louis, Paris, France; Corinne Alberti, M.D., Staff Physician, DBIM, Hôpital Saint Louis, Paris, France; Frederik Lecorre, Informatician, Supelec, France. Investigators of the OutcomeRea Data Base: Caroline Bornstain, M.D., Staff Physician, ICU, HEGP, Paris, France; Marie Thuong, M.D., Staff Physician, ICU, Hôpital de la Fontaine, Saint Denis, France; Marie-Alliette Costa de Beauregard, M.D., Staff Physician, Nephrology, Hôpital Tenon, Paris, France; Jean-Pierre Colin, M.D., Staff Physician, ICU, CH de Dourdan, Dourdan, France; Eric Le Miere, M.D., Attending Physician, ICU, Hôpital Louis Mourier, Colombes, France; Caubel Antoine, M.D., Attending Physician, ICU, Hôpital Saint Joseph, Paris, France; Christine Cheval, M.D., Staff Physician, SICU, Hôpital Saint Joseph, Paris, France; Eric Vatalon, M.D., Staff Physician, SICU, Hôpital Saint Joseph, Paris, France; François Vincent, M.D., Staff Physician, Nephrology, Hôpital Tenon, Paris, France; Amar Salah, M.D., Staff Physician, ICU, Hôpital Louis Mourier, Colombes, France; Laurent Montessino, M.D., Attending Physician, ICU, Hôpital Bichat, Paris, France; Etienne Pigné, M.D., Staff Physician, ICU, Hôpital Louis Mourier, Colombes, France; Alexandre Boyer, M.D., Staff Physician, ICU, Hôpital Pellegrin, Bordeaux, France; Olivier Tuill, M.D., Staff Physician, ICU, Hôpital Avicenne, Bobigny, France; Bruno Mourvillier, M.D., Staff Physician, ICU, Hôpital Bichat, Paris, France; Samir Jamali, M.D., Staff Physician, ICU, CH de Dourdan, Dourdan, France; Delphine Moreau, M.D., Staff Physician, ICU, Hôpital Saint Louis, Paris, France; Guillaume Thiery, M.D., Staff Physician, ICU, Hôpital Saint Louis, Paris, France; Alexandre Duguet, M.D., Attending Physician, Hôpital Pitié-Salpetriere, Paris, France; Christian Laplace, M.D., Attending Physician, ICU, Hôpital Kremlin-Bicetre, Bicetre, France.