2021 年 3 巻 3 号 p. 78-87
BACKGROUND
Although the majority of medical professionals recognize unplanned extubation as a critical accident, its relationship with the clinical outcomes of ventilated patients is controversial. The purpose of this study was to understand whether unplanned extubation, including self-extubation and accidental extubation, is a prognostic factor for clinical outcomes of mechanically ventilated adult patients.
METHODS
A pre-specified protocol was registered on PROSPERO (CRD42019120976). EMBASE, MEDLINE, CINAHL, and the ICTRP were searched on December 25, 2018 and February 5, 2020. The primary outcome was hospital mortality, and the secondary outcomes were ICU mortality, ICU and hospital length of stay, duration of mechanical ventilation, proportion of reintubation, and cost.
RESULTS
Of the 3216 articles retrieved, 11 were selected for the systematic review, and 9 met the criteria for the meta-analysis. Self-extubation was significantly associated with lower hospital mortality (OR = 0.49, 95% CI 0.30–0.81; certainty of evidence: moderate) and unplanned extubation was associated with ICU mortality (OR = 0.34, 95% CI 0.17–0.69; certainty of evidence: moderate). There were no significant between-group differences in lengths of hospital or ICU stay, with mean differences of 1.71 days (95% CI −7.68 to 11.69; certainty of evidence: very low) and 1.26 days (95% CI −3.58 to 6.10; certainty of evidence: very low), respectively.
CONCLUSIONS
Self-extubation is associated with lower patient mortality, but a definitive conclusion cannot be made due to methodological limitations.
Unplanned extubation of the tracheal tube is a common accident among patients who undergo invasive mechanical ventilation [1]. A previous systematic review reported that the median rate of unplanned extubation was 7.3 per 100 ventilated patients [2]. Although the majority of medical professionals recognize unplanned extubation as a critical accident, it has not been evaluated whether unplanned extubation has a relationship with, or is a prognostic factor for, the clinical outcomes of ventilated patients.
A previous systematic review without quantitative synthesis [2] reported inconsistent results on hospital mortality of patients with unplanned extubation compared to those without unplanned extubation, with studies showing higher or lower mortality rates. Most of the studies on hospital mortality included in the systematic review were case-control studies matched for clinical characteristics. However, to examine whether unplanned extubation is a prognostic factor for clinical outcomes, cohort matching with clinical characteristics must be avoided in case–control studies. Moreover, the control must include not only planned extubation but also death during mechanical ventilation in cohort studies. Therefore, we aimed to examine the role of unplanned extubation as a prognostic factor for short-term clinical outcomes.
This systematic review and meta-analysis study was conducted according to the guidelines in the Cochrane Handbook [3] and reported according to the Preferred Reporting Items for Systematic Reviews Meta-Analysis (PRISMA) statement [4] (Supplementary Appendix 1). The study protocol was registered on PROSPERO (CRD42019120976) in January 2019.
RESEARCH QUESTION AND ELIGIBILITY CRITERIAThe research question we evaluated was: Is unplanned extubation in adults who are on invasive mechanical ventilation in the ICU related to their short-term outcomes? Prognostic factors were defined as “factors associated with a risk of future health outcomes in individuals with a particular health condition [5].” The primary outcome of this study was hospital mortality, and the secondary outcomes included ICU mortality, ICU and hospital length of stay (LOS), duration of mechanical ventilation, proportion of reintubation, and costs. Studies were included if they enrolled only adult patients (age ≥18 years) on invasive mechanical ventilation in the ICU. Eligible studies included published and unpublished observational studies, including abstracts. Control patients were defined as patients with no history of unplanned extubation, including those who later died or underwent a tracheostomy with mechanical ventilation. Studies in all languages were included. We included studies with follow-up durations of any length. Case reports, case series, and qualitative studies were excluded. Additionally, we excluded matched case-control studies due to selection bias in choosing controls [6, 7]. These studies matched similar characteristics, hence do not reflect general mechanical ventilation populations. In this study, our aim was to evaluate unplanned extubation as a prognostic factor in general mechanical ventilated patients in acute care settings.
Unplanned extubation was defined as accidental or patient-initiated endotracheal tube removal. However, decannulation of the tracheal tube via a tracheostoma was not considered unplanned extubation. In this study, we defined self-extubation as patient-initiated extubation. We defined accidental extubation as unintentional extubation, excluding self-extubation. We used the term unplanned extubation to include both self-extubation and accidental extubation.
SEARCH STRATEGIES AND SELECTION OF STUDIESWe searched the following databases for eligible studies: MEDLINE via PubMed, EMBASE, Cochrane Central Register of Controlled Trials (CENTRAL), and CINAHL. We searched for ongoing trials registered in the World Health Organization International Clinical Trials Registry Platform on December 25, 2018 and updated the search on February 5, 2020. The key search terms used to identify potentially relevant studies are listed in Supplementary Appendix 2. We included all references from a previous systematic review [2]. We attempted to identify additional relevant studies by manually searching the reference lists of both the studies returned by the search and articles citing such studies (based on Google scholar). If sufficient information was unavailable, we contacted the authors of each study. Two of the 3 reviewers (TU, HS, and ST) independently screened the titles and abstracts to identify potentially relevant studies. Two reviewers (TU and HS) then independently assessed the eligibility based on a full-text review. Disagreements were resolved by discussion to reach consensus and, if necessary, ST and YK arbitrated these discussions.
DATA EXTRACTION AND QUALITY ASSESSMENTData were independently extracted by TU and HS into a pre-specified form. We contacted the authors of the studies in cases with insufficient information. Two authors (TU and HS) independently assessed the risk of bias for each outcome in the included studies with the Quality In Prognostic Studies tool [8]. Disagreements were resolved by discussion and, if necessary, ST arbitrated these discussions.
DATA ANALYSISWe created a tabular summary of findings table using primary and secondary outcomes. We used the grading of recommendations, assessment, development, and evaluation (GRADE) considerations to assess the quality of evidence [9, 10]. The analysis was conducted in Review Manager 5.3 software (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). Pooled risk ratios (RR) or odds ratios (OR) with 95% confidence intervals (95% CI) for dichotomous variables and mean differences (MD) with 95% CI for continuous variables were estimated by a random-effects model. If the pooled analysis included a case–control study, we reported a pooled OR and calculated the assumed risk from the pooled risk of the control. The corresponding risk was multiplied by the RR and converted to the OR by using a previously defined formula [11].
In cases where heterogeneity was detected (I² >50%), we attempted to identify the possible causes, such as differences in the proportion of reintubations, severity of illness, and age between studies. A p-value of less than .05 was considered statistically significant for all analyses.
For discrete variables, we conducted imputation of missing values as worst–worst scenario: all missing patients in the 2 groups have outcomes. For continuous variables, we did not undertake imputation of missing values. We did not pool studies that did not report continuous outcome variables of all the participants. All of the studies included in this systematic review had no missing data. Thus, we conducted meta-analyses using original data.
ANALYSIS OF SUBGROUPS OR SUBSETSTo explore the potential heterogeneity and determine the presence of a risk of bias, we conducted subgroup analyses based on the application of a systematic weaning approach for the participants. Additionally, we undertook subgroup analysis based on whether the studies were done before or after 2011. The ABCDE bundle was introduced in 2011 and is used to facilitate a strategic weaning process, including the choice of sedatives and the evaluation of pain and delirium [12].
DIFFERENCES BETWEEN PROTOCOL AND REVIEWWe planned to use a funnel plot to detect publication bias for the primary outcome; a subgroup analysis based on the characteristics of the majority of participants in each study (i.e., surgical, medical, or mixed); sensitivity analysis for the primary outcome by excluding studies which used imputed statistics and those published before 2001, the year that the first evidence-based weaning guidelines became available [13]. However, all the mentioned analyses could not be conducted due to insufficient data. Additionally, sensitivity analysis of the studies limited to self-extubation for the primary outcome was planned; however, it could not be conducted due to insufficient data. Thus, we conducted the sensitivity analysis for the ICU mortality instead of the primary outcome.
In instances where the meta-analytic result of the primary outcome was statistically significant, we planned to use meta-regression to adjust for certain confounders such as age, proportion of women, illness severity, proportion of re-intubation, and the length of mechanical ventilation. However, we were unable to conduct this meta-regression, as the included studies did not have sufficient data for this level of analysis.
The initial and updated search identified 3,216 articles and abstracts. A total of 16 studies were assessed for eligibility, and 2,731 patients across the 11 studies were included in quantitative analysis (Fig. 1, Table 1). Five studies were excluded from qualitative synthesis (Supplementary Appendix 3). Three studies used weaning protocols (Table 1). Additional data are shown in Supplementary Appendix 4. The reintubation ratio in the exposure group ranged from 10.3% to 76.6%. We did not identify any protocol articles or ongoing studies from the registration search. Excluded studies, in which control population was selected from mechanically ventilated subjects without unplanned extubation, but matched case-controlled studies were shown in Supplementary Appendix 5.
| Source | Study type | Type of ICU | Initiation of data collection | Definition of UEX | UEX, N | Control, N | Protocolized weaning | Outcome |
|---|---|---|---|---|---|---|---|---|
| Coppolo and May 1990, United States | Single-center prospective cohort study | Unclear | NR | SE | 12 | 100 | No | NR |
| Vassal et al., 1993, France | Single-center case–control study | Medical and Mixed | 1993 | SE + AE | 24 | 173 | NR | Duration of MV |
| Boulain et al., 1998, France | Multicenter prospective cohort study | Medical and Mixed | 1992 | SE + AE | 46 | 380 | NR | ICU mortality and proportion of re-intubation |
| Chevron et al., 1998, France | Single-center prospective cohort study | Medical | 1994 | SE + AE | 36 | 74 | NR | Hospital LOS and duration of MV |
| Bouza et al., 2007, Spain | Single-center prospective cohort study | Medical | NR | SE + AE | 34 | 310 | Yes | ICU mortality, ICU LOS and duration of MV |
| Jarachovic et al., 2011, United States | Single-center prospective cohort study | Medical | 2007 | SE + AE | 29 | 161 | Yes | lCU LOS and duration of MV |
| De Groot, et al., 2011, Netherlands | Single-center case-control study | Mixed | 2005 | SE | 74 | 296 | Yes | ICU mortality, hospital mortality, l ICU LOS, hospital LOS, and duration of MV |
| Cho and Yeo, 2014, Korea | Multicenter retrospective cohort study | Mixed | 2013 | SE | 32 | 418 | NR | Duration of MV |
| Lee et al., 2015, Korea | Single-center prospective cohort study | Medical | 2011 | SE | 30 | 420 | No | ICU LOS, hospital LOS, and duration of MV |
| Chuang, et al., 2015, Taiwan | Single-center case-control study | Mixed | 2012 | SE | 37 | 156 | NR | ICU mortality, hospital mortality, l ICU LOS, hospital LOS, and duration of MV |
| Aydoğan, et al., 2017, Turkey | Single-center case–control study | Mixed | 2012 | SE + AE | 30 | 60 | NR | NR |
UEX, unplanned extubation, AE, accidental extubation, SE, self-extubation, SD, standard deviation, NR, not reported, MV, mechanical ventilation, LOS, length of stay.
The risk of bias was assessed based on the outcomes that we had previously defined (Supplementary Appendix 6). Almost all of the pooled studies had a low risk of bias for study participation, study attrition, and outcome measurement. Moreover, all the pooled studies demonstrated a moderate risk of bias for study confounding, as the important confounding factors, such as delirium, duration of mechanical ventilation, severity of illness, and age, were not considered. Most pooled studies, except for those that considered hospital mortality and reintubation ratio, had a high risk of bias for statistical analysis and reporting as they did not use multivariate analysis, and their protocols were not registered. Among the 2 studies reporting hospital mortality, one had a moderate risk of bias in the measurement of prognostic factor. Almost half of the studies that reported hospital LOS, ICU LOS, and duration of mechanical ventilation had a moderate risk of bias for the measurement of prognostic factor.
HOSPITAL AND ICU MORTALITYOnly 2 case–controlled studies reported hospital mortality (Table 1). In these studies, case and control were self-extubation and no self-extubation, respectively. The studies reported outcomes of patients with and without self-extubation, as well as risk factors for self-extubation. ICU mortality was reported by two case-controlled and two prospective observational cohort studies (Table 1). The exposure in 2 of these studies included self-extubation only, whereas that of 2 other studies included both self- and accidental extubation. We planned to calculate the pooled RR for dichotomous variables. However, as all studies that reported hospital mortality and half of all studies that reported ICU mortality were case-controlled, we calculated the pooled OR for ICU and hospital mortality instead. Self-extubation was significantly associated with a decrease in hospital mortality (2 trials, n = 743; OR: 0.49, 95% CI: 0.30–0.81, p = 0.005; I2 = 0%; Fig. 2a); however, the certainty of evidence for hospital mortality was moderate (Table 2). Similarly, unplanned extubation was significantly associated with a decrease in ICU mortality (4 trials, n = 1,333; OR: 0.34, 95% CI: 0.17–0.69, p = 0.003; I2 = 53%; Fig. 2b). The certainty of evidence for ICU mortality was moderate (Table 2). We conducted 2 subgroup analyses based on whether: 1) the studies were conducted before or after the ABCDE bundle introduced in 2011, and 2) a systematic weaning approach was used, in accordance with the registered protocol (Supplementary Appendix 7). Additionally, we conducted the sensitivity analysis for ICU mortality based on the studies limited to self-extubation. In the sensitivity analysis, unplanned extubation was no longer associated with ICU mortality (2 trials, n = 563; OR: 0.35, 95%CI: 0.09–1.27, p = 0.09; I2 = 64%; Supplementary Appendix 8).
| Overview of study design | ||||||
| Patients or study population: adult patients on mechanical ventilation in the ICU Exposure: Unplanned extubation, defined as self-extubation, accidental extubation, or both. Comparison: No unplanned extubation |
||||||
| Outcome | Illustrative comparative risks* (95% CI) | Relative effect (95% CI) | No. of participants (studies) | Certainty of the evidence (GRADE) | Comments | |
| Assumed risk | Corresponding risk | |||||
| Control | Exposure | |||||
| Hospital mortality | Study population | OR 0.49 (0.30–0.81) |
743 (2 studies) | ⊕⊕⊕⊝ Moderatea,b |
||
| 540 per 1,000 | 365 per 1,000 (260–487) |
|||||
| ICU mortality | Study population | OR 0.34 (0.17–0.69) |
1,333 (4 studies) | ⊕⊕⊕⊝ Moderatea,b |
||
| 450 per 1,000 | 218 per 1,000 (122–361) |
|||||
| Hospital length of stay | Study population | 753 (3 studies) | ⊕⊝⊝⊝ Very Lowa,b,c,d |
|||
| MD: 1.71 (−7.68 to 11.09) |
||||||
| ICU length of stay | Study population | 1,177 (4 studies) | ⊕⊝⊝⊝ Very Lowa,b,c,d |
|||
| MD: 1.26 (−3.58 to 6.10) |
||||||
| Duration of mechanical ventilation | Study population | 1,934 (7 Studies) | ⊕⊝⊝⊝ Very Lowa,b,c,d |
|||
| MD: 0.44 (−2.68 to 3.56) |
||||||
| Reintubation | Study population | RR 6.17 (4.05–9.41) |
330 (1 Study) | ⊕⊕⊕ ⊕ Higha |
||
| 99 per 1,000 | 608 per 1,000 (329–928) |
|||||
| GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to the effect estimate. Moderate certainty: We are moderately confident of the effect estimate; the true effect is likely to be close to the effect estimate, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited; the true effect may be substantially different from the effect estimate. Very low certainty: We have very little confidence in the effect estimate; the true effect is likely to be substantially different from the effect estimate. |
||||||
ICU, intensive care unit; CI, confidence interval; RR, risk ratio; MD, mean difference; OR, odds ratio; RCT, randomized controlled trial
* absolute risk
a We did not consider the domain of study confounding in the risk of bias to judge the certainty of the evidence.
b Downgraded one point because of moderate or high risk of bias associated with the measurement of prognostic factor.
c Downgraded one point for inconsistency.
d Downgraded one point as imprecise (confidence interval).
* The corresponding risk (95% CI) is based on the assumed risk in the comparison group and the relative effect (95% CI) estimated for the intervention group. Assumed risk was estimated from the meta-analysis of risks in the control.
Three studies reported the hospital LOS, and four studies reported the ICU LOS (Table 1). No statistically significant differences in the pooled hospital and ICU LOS were noted between patients undergoing unplanned extubation and the controls; the mean differences were 1.71 days (3 trials, n = 753; 95% CI: −7.68 to 11.69, p = 0.61; I2 = 74%) and 1.26 days (4 trials, n = 1,177; 95% CI: −3.58 to 6.10, p = 0.81; I2 = 86%), respectively (Fig. 2c, d). Seven studies reported the total duration of mechanical ventilation in the 2 study groups (Table 1). The pooled duration of mechanical ventilation demonstrated no statistically significant difference between the 2 groups, with a mean difference of 0.44 days (7 trials, n = 1,934; 95% CI: −2.68 to 3.56, p = 0.78; I2 = 86%; Fig. 2e). The certainty of evidence for hospital and ICU LOS and duration of mechanical ventilation was “very low” (Table 2). Only 1 study compared the proportion of re-intubation between the 2 groups. There was a statistically significant increase in the proportion of reintubation in patients undergoing unplanned extubation compared with controls (1 trial, n = 330, RR: 6.17, 95% CI: 4.05–9.41, p < 0.01; Fig. 2f). The certainty of evidence for the proportion of reintubation was “high” (Table 2). None of the included studies reported any differences in costs.
The findings of this review, which included 11 studies and 2,532 patients, suggested that self-extubation was significantly related to lower ICU and unplanned extubation is associated with lower hospital mortality. The certainty of evidence for each factor was moderate. However, unplanned extubation was neither significantly associated with the ICU or hospital LOS, nor the duration of mechanical ventilation among patients undergoing invasive mechanical ventilation; the certainty of evidence for these findings was very low. Unplanned extubation was related to reintubation, and the certainty of evidence for this finding was high; however, the proportions of reintubation in patients with unplanned extubation were extremely varied across the included studies, ranging from 10.3% to 76.6%.
We found that unplanned extubation could probably be related to lower ICU, and self-extubation was associated with lower hospital mortality. These findings are consistent with those of previous studies that reported unplanned extubation to be associated with lower mortality [14–17]. De Groot et al. [15] reported that unplanned extubation was an independent factor for hospital mortality after adjustment for the APACHE (Acute Physiology And Chronic Health Evaluation) II score, age, and type of admission. A previous study indicated that most incidences of unplanned extubation occurred during weaning from mechanical ventilation [14]. Moreover, a high Glasgow Coma Scale score and agitation were both significantly more likely in patients with unplanned extubation than those without unplanned extubation [18, 19]. In our review, self-extubation was the exposure variable in the two studies analyzed for hospital mortality [15, 17]. This could imply that the self-extubation was due to light sedation of the patients enabling them to move actively, hence the low risk of death. Therefore, our findings indicate that patients who experience unplanned extubation have a lower risk of death than those who do not. In both the included studies, the APACHE II score was higher in patients with unplanned extubation than in controls. Contrary to our findings, Yoon et al. [20] reported that unplanned extubation was related to higher hospital mortality in patients in a surgical ICU; however, the control group in the study comprised patients who underwent planned extubation and, therefore, excluded patients who died during mechanical ventilation. Larger cohort studies are needed to confirm the association between unplanned extubation and mortality.
Our study suggests that unplanned extubation is not a prognostic factor for lengths of hospital and ICU stay, or duration of mechanical ventilation, although the evidence is very uncertain. A high level of heterogeneity (Fig. 2) in the hospital and ICU LOS and duration of mechanical ventilation was noted between studies. One explanation for this was a difference in the reintubation rates after unplanned extubation among the studies. A previous study suggested that reintubation after unplanned extubation was associated with the length of mechanical ventilation, and ICU and hospital LOS [16]. Most of the studies we analyzed did not have specified criteria for ICU discharge or weaning protocols. These factors may have affected our findings. Therefore, further studies with defined criteria for discharge from the ICU and weaning from mechanical ventilation are required.
We defined unplanned extubation, including self-extubation and accidental extubation; however, self-extubation and accidental extubation were completely different events in terms of risk factor and outcome. Previous studies indicated that the proportion of re-intubation was higher in accidental extubation [14], and re-intubation led to higher mortality [20]. In this study, we assessed ICU mortality based on studies with only self-extubation and both self-extubation and accidental extubation, and this may have contributed to our results. Our sensitivity analysis of studies limited to self-extubation for ICU mortality showed no significant differences in ICU mortality. We do not have a valid explanation for this finding; however, it is possible that the proportion of accidental extubation was lower in unplanned extubation and did not affect the results of the analysis [14, 21]. Additionally, the sample size was insufficient; thus, random error may contribute to these results. Further studies that distinguish self-extubation and accidental extubation are needed.
This systematic review and meta-analysis has several strengths. First, this is the first systematic review to investigate the relationship between unplanned extubation and patient outcomes with quantitative synthesis. Second, the novelty of our review is clear since we considered duplicate assessments of eligibility, risk of bias, and data extraction, and used the GRADE approach [9, 10] to assess the certainty of evidence. Third, our review used a rigorous methodology that followed a written a priori registered protocol developed according to the PRISMA statement [4].
However, our study had several limitations. First, the number of pooled studies was minimal; sufficient subgroup and sensitivity analysis could not, therefore, be conducted. Second, in the analysis of secondary outcomes, relatively old studies were included, which may have led to publication bias [22]; however, we were unable to assess bias with a funnel plot because of the insufficient number of studies. Third, we could not assess the effects of accidental extubation on hospital mortality because all studies assessed for hospital mortality included self-extubation only. Fourth, the reintubation rates varied considerably between the included studies, and the proportion of reintubation in each study was a possible confounder in our review. Fifth, 3 of the 9 studies included in the meta-analysis lacked or had inadequate monitoring for unplanned extubation (e.g., from an incident reporting system); alternatively, a method for detecting unplanned extubation was missing. Future studies are required to develop methods for the systematic detection of unplanned extubation.
The aim of this systematic review was to evaluate whether unplanned extubation was a prognostic factor for short-term outcomes but not to assess the causal relationship between unplanned extubation and short-term outcomes. We did not suggest that unplanned extubation directly affect mortality; however, we noted that unplanned extubation could be a prognostic factor for lower ICU and hospital mortality. The causal relationship between unplanned extubation and short-term outcomes is likely to be influenced by potential confounders such as severity of illness, depth of sedation, and phase of disconnection of mechanical ventilation process; therefore, future cohort studies that account for these confounders are warranted.
Our findings suggest that many ready-to-extubate patients remain intubated. Previous studies suggested that self-extubation frequently occurred during weaning from the mechanical ventilation process [1, 23]. Generally, patients undertaking self-extubation are stable and could be expected to have lower mortality. This hypothesis was supported by the relatively low re-intubation rate following self-extubation [14].
We expect our findings to contribute to the medical profession by adding knowledge about the possibility of a duration of unnecessary mechanical ventilation. To minimize time on unnecessary ventilation, effective strategies to facilitate the weaning process, such as the ABCDEF bundle, should be implemented. A previous study indicated that the implementation of the ABCDEF bundle was associated with a lower likelihood of next-day mechanical ventilation [24]. Finally, we emphasize that unplanned extubation is not always a prognostic factor for lower mortality of the mechanically ventilated patient. For patients with conditions such as severe hypoxemia and neurological respiratory depression and deficit, unplanned extubation could lead to severe complications and higher mortality.
Our study has suggested that self-extubation could probably be related to lower hospital mortality, but a definitive conclusion cannot be made as only two studies were included in the analysis. Further large-sample well-designed cohort studies are required to evaluate the prognostic effect of unplanned extubation.
None.
This work was supported by the Sapporo City University.
We would like to thank Dr. Ayman El-Menyar, Dr. Robert Hyzy, and Dr. Christophe Girault for providing us with additional information regarding their studies. This work was supported by the research grant from Sapporo City University.
