Introduction

Ineffective efforts (IEs) during mechanical ventilation, defined as inspiratory efforts unable to trigger a ventilator-delivered breath, are a common type of patient–ventilator asynchrony [1, 2]. The incidence of IEs varies, depending on patient sample, ventilator settings, sedation level and wakefulness/sleep state [1–6]. In unselected, critically ill patients, an IEs index (IEs expressed as percentage of the total number of breaths, effective and ineffective) of 1–3 % has been reported [1, 7, 8]. The recent study by Blanch et al. [8], showed that the incidence of IEs varies over time in the same patient, suggesting that the incidence of IEs also depends on the observation period.

It has been shown that, in critically ill patients, the presence of asynchronies, including IEs, is associated with worse patient outcomes [9]. Using a threshold of asynchrony, or IEs index greater than 10 %, studies have shown an association with increased duration of mechanical ventilation [1, 7], and with higher patient mortality in intensive care unit (ICU) [8]. However, it has been observed that IEs tend to occur in clusters, between often prolonged uneventful periods [8, 10], and indexing IEs over time obviously obscures the presence of clusters.

The aim of this study was to investigate the role of IEs in critically ill patients, focusing specifically on clusters of IEs. The hypothesis was that clusters of IEs would have a stronger correlation with patient outcome than sporadic IEs and the global IEs index. For this purpose, continuous recordings obtained from a validated research prototype monitor, capable of identifying with high accuracy IEs in this patient population, were used [11], and a mathematical model to describe clusters was developed. Part of this study has been presented as an abstract at the European Society of Intensive Care Medicine International Conference [10].

Methods

Data collection

This study was conducted in a medical–surgical ICU between January 2010 and July 2011. The study was approved by the Hospital Ethics Committee and, since there was no interference with patient management, signed informed consent was waived.

To detect IEs, a research prototype monitor (PVI Monitor; YRT, Winnipeg, Canada) [11, 12] was used as detailed in the electronic supplementary material. The PVI monitor, using continuous measurements of airway pressure and flow, estimates a patient’s total inspiratory muscle pressure via the equation of motion, using improvised values of resistance and elastance of the respiratory system, obtained without additional interventions in mechanically-ventilated patients [11, 12].

Patients under controlled mechanical ventilation for at least 12 h were screened for eligibility. The patients were studied when they were first placed on assisted mode of support (pressure-support or proportional-assist ventilation mode), after remaining stable for >1 h, and if the primary physician estimated that they would remain on assisted mechanical ventilation for the next 24 h. Patients were excluded only when the PVI monitor was unavailable, or unable to detect initiation of patient effort (see electronic supplementary material), usually occurring when patients were on continuous positive airway pressure or very low assist levels (<5 cmH2O on pressure support) [11].

The recording period was 24 h, unless the patient was either disconnected earlier from the PVI, or placed on controlled modes of ventilation, low assist level or continuous positive airway pressure. Patients that remained on assisted ventilation were studied again on the 3rd and 6th days, if the PVI monitor was available.

Data analysis

The output of PVI monitor data was processed before analysis to optimize data quality (e.g., artifact rejection due to disconnection, suction, cough). To facilitate analysis, the initial sequence of IEs was re-sampled to a time-series with the number of IEs calculated in uniform intervals of 30 s, while preserving the total number of IEs.

The IEs index was calculated as previously described [1]. To describe the clusters of IEs, the concept of an event of IEs was developed (see also electronic supplementary material). Events of IEs were defined as periods of time containing more than 30 IEs in a 3-min period, which, assuming a breathing rate of 20 per minute, corresponds to IEs being 50 % of breaths. To identify events, the IEs signal (number of IEs in 30-s intervals) was smoothed by a moving average window of 6. Using this filtered signal, values above a threshold of at least 30 IEs in a 3-min period were selected. Neighboring values were grouped in larger areas until a zero value was reached. The maximum value was located and the limits were set at the points where this maximum value dropped by 80 %. These areas were identified as events, and characterized by their duration (time), and power, that is the number of IEs they contained (Fig. 1; Fig. S1). All analysis was performed using the R programming language and software environment.

Fig. 1
figure 1

Representative view of events of ineffective efforts (IEs) in one patient, in a selected time-frame, generated via an R-Shiny application built for this purpose. The green line shows the processed number of IEs per 30-s period (unfiltered signal). The purple line represents the calculated average value of IEs after smoothing, and the teal blue horizontal line represents the set threshold for events (30 IEs/3-min period). The event is the green area. The start and end of the event are at the point of 80 % drop from the maximal value of the smoothed signal. The duration of this event is 14 min, while its power is 172 IEs. At the bottom, the selected time-frame is highlighted against the complete recording. See text for further explanation

Full size image

The evaluated outcome measures included ICU and hospital mortality, the length of stay in ICU, and the duration of mechanical ventilation. For the duration of mechanical ventilation as outcome, both the total time of invasive ventilatory support and the time after the first recording were analyzed, the latter because any effect of the observed IEs could not affect the time patients had already spent on controlled (passive) mechanical ventilation before inclusion.

Due to study design and instability of critically ill patients, the recording periods were expected to vary between patients. To overcome the issue of potential selection bias due to either short or long recordings, apart from the analysis of all the patients’ data, an additional analysis was performed. This analysis, which aimed at patients with relatively similar observation periods, included data obtained only during the 1st day, from patients having a recording period of at least 16 h (1st day group).

Statistical analysis

Continuous variables are reported as means and standard deviation for normally distributed data, and medians and interquartile ranges (IQR) for non-normally distributed data. Categorical variables are presented as percentages, and compared using Fisher’s exact test. Between-group differences in continuous variables were compared using the Mann–Whitney U test. Spearman’s rho was used to evaluate correlations between continuous variables. Multivariate logistic regression analysis was used to evaluate the adjusted contribution of IEs on survival and duration of mechanical ventilation. A p value of <0.05 was considered significant. We used IBM SPSS-Statistics for Windows v.22 (Armonk, NY, USA) for analysis.

Results

A total of 111 patients were studied. The electronic file of PVI data was corrupted in 1 patient, so 110 patients were included in the analysis. Overall, 160 recordings were available for analysis, corresponding to 2931 h of assisted mechanical ventilation, and 4,456,537 breaths, free of artifacts. All patients had a first recording (total 2028 h), a second recording was obtained in 37 patients (646 h), and a third in 13 patients (257 h). The median, artifact-free recorded time in all patients was 23 h (IQR = 19–73 h). The 1st day group included 79 patients, in whom the total recorded time was 1768 h, and the median recorded time was 22 h (IQR = 20–24 h). Table 1 shows the patients’ demographic data, comorbidities, admission diagnosis, and parameters of mechanical ventilation of the 1st day group and of all patients. The major outcome measures are presented in Table 2.

Table 1 Patient’s characteristics
Full size table
Table 2 Main outcomes in 1st day group (n = 79) and all patients (n = 110)
Full size table

Analysis of IEs index

The median IEs index was 2.43 (IQR 1.1–5.1). No statistically significant association of IEs index with outcomes was found (Table 3; and Table S3). An IEs index greater than 10 % was found in 13 patients (12 %). Again, no statistically significant association of IEs index greater than 10 % with any outcome was observed (Table S4).

Table 3 Correlations of IEs index and events characteristics, power and duration, with length of ICU stay, and duration of mechanical ventilation in total, and after first recording in 1st day group (n = 79) and in all patients (n = 110)
Full size table

Analysis of events of IEs in 1st day group

Events were identified in 24 out of 79 patients (30.4 %), and the total number of events was 80. In patients who had events, the median number of events was 2 (1–3), the median event duration was 21 (12–57) min., and median event power was 215 (101–488) IEs. There were no differences in sex, age, APACHE II and mode of ventilation between patients with or without events (Table S5). Patients with an admission diagnosis of sepsis had events more commonly than patients with any other admission diagnosis (p = 0.03).

The presence of events in the 1st day group was associated with significantly longer duration of mechanical ventilation after first recording (Fig. 2a) and in logistic regression analysis, with increased risk: (1) of being on mechanical ventilation for more than 8 days after the first recording (a cut-off point corresponding to the median duration of mechanical ventilation after the first recording); and (2) of hospital mortality (Table 4). The event’s characteristics, power and duration, were also correlated with the duration of mechanical ventilation after the first recording (Table 3).

Fig. 2
figure 2

Duration in days of ICU stay, and mechanical ventilation, in total (MV-total), and after the first recording (MV-post), for patients with events (teal), and without (orange). a 1st day group (n = 79), b all patients (n = 110), c patients with an IEs index less than 10 % (n = 97), *p < 0.05, **p < 0.01, box and whiskers at 10–90 %, line at median

Full size image
Table 4 1st day group (n = 79)
Full size table

Analysis of events of IEs in all patients

All recordings from the same patient were combined, after confirming that none of the variables differed significantly between recordings (Table S2).

Events were identified in 42 out of 110 patients (38.2 %). Event characteristics, median power and duration, were not different from those observed in the 1st day group. There was also no difference in the presence and characteristics of events between the morning, afternoon and evening (Fig. S2).

The duration of ICU stay and mechanical ventilation (in total and after 1st recording) were significantly longer for patients with events than for those without (Fig. 2b). The event characteristics, power and duration, correlated with ICU length of stay and duration of mechanical ventilation (Table 3). In logistic regression analysis, the presence of events was associated with increased risk of being on mechanical ventilation for more than 8 days after the first recording, and with hospital mortality (Tables S6, S7).

Finally, among the 97 patients who had IEs index less than 10 %, events were present in 29 (29.9 %). The duration of mechanical ventilation after the first recording and of ICU stay was on average 8 days longer in patients with events than in those without (p < 0.05; Fig. 2c).

Discussion

The main findings of this study in critically ill patients were: (1) the IEs index, or a value above 10 %, had no correlation with patient outcome; (2) the presence of clusters of IEs, described as events of IEs, was associated with longer duration of mechanical ventilation and higher hospital mortality.

Three studies over the last decade have focused on the incidence of asynchronies and specifically IEs and their effect on outcome of unselected, critically ill patients [1, 7, 8]. In an attempt to identify a threshold of IEs associated with poor outcome, the value of an IEs or asynchronies index greater than 10 % has been used. Using this cut-off value, two of the studies showed an association with increased duration of mechanical ventilation but not mortality [1, 7], while the other study found higher ICU mortality, and a trend towards longer duration of ventilation [8]. An important observation from this last study was the significant variability of IEs over time in the same patient, highlighting the need for continuous recordings to monitor IEs.

While undoubtedly a high incidence of IEs is associated with poor patient outcome, the choice of the 10 % cut-off value has certain limitations. Firstly, it is not derived from a representative sample of patients [13]. Secondly, indexing over time cannot reveal the presence of clusters of IEs. More importantly, the presence of IEs index more than 10 % can only be identified retrospectively, and therefore cannot be used as an alarm on a ventilator.

Indeed, it has been shown that IEs tend to occur in clusters, between often prolonged uneventful periods [8]. A possible explanation for this observation is that the factors affecting the presence of IEs, such as sedation, wakefulness/sleep state, level of assist and ventilatory drive, may vary substantially during the course of mechanical ventilation. We hypothesized that the presence of such clusters could have a stronger link to patient outcome than sporadic, scattered IEs, as biological phenomena are often non-linear. Clearly, when the number of IEs is indexed over a prolonged time period, the presence of such clusters can be missed. We therefore sought to develop a mathematical model to describe those clusters, and thus developed the concept of an event of IEs. Given the lack of previous data, the choice of cutoff values defining an event had to be rather heuristic, and further prospective studies with larger numbers of patients are needed for fine tuning of these thresholds.

In the patient population studied, almost all patients had some IEs, but only 30 % had events. The characteristics of the events, power and duration, remained relatively stable during the observation period. The presence of events, as well as their power and duration, were associated with prolonged duration of mechanical ventilation, even in patients with IEs index less than 10 %. An increased risk of hospital mortality was observed, although a larger cohort may be needed to confirm this.

The presence of clusters of IEs in two different studies [8] highlights the utility of the concept of event of IEs. More importantly, events could be prospectively identified by appropriate software on the ventilator, like IEs [14, 15], and used as alarms. Indeed, in the process of selecting the event definition, the potential use as an alarm was paramount. An event can be identified, by definition, after a 3-min period, while the observed median event duration in our patients was 21 min. Thus, although the presence of events was associated with adverse outcomes in our study, this significant time difference suggests that, if IEs are correctable, an alarm at 3 min could potentially prevent the evolution of an event. Nonetheless, the efficiency of such intervention should be prospectively evaluated.

Some further aspects and limitations of our study need to be detailed. Firstly, we did not study all forms of asynchronies, but focused only on IEs, which is the most common major asynchrony [1]. Moreover, we did not study the whole duration that patients were on ventilators, nor all modes of ventilation. We only studied patients on assisted modes of ventilation, pressure support and proportional assist, as IEs occurring in spontaneously breathing patients are probably different in pathophysiology and effects from those occurring in patients ventilated passively in controlled modes. Indeed, it is increasingly being recognized that IEs observed during inspiration in controlled modes of ventilation often represent reverse-triggering (entrainment) [16–18]. Although all ‘control’ modes in modern ventilators allow assisted breaths, use of those modes in spontaneously breathing patients varies in everyday practice, and is very limited in our ICU. Thus, it should be emphasized that the observed results are derived from the specific population studied, and cannot be generalized without further studies.

The value of the obtained results also relies on the method used to identify IEs. The reported sensitivity of the PVI monitor in identifying IEs is 87 % [11], with most cases of missed IEs occurring in patients with very severe flow limitation. In our study, 24 patients (22 %) had a diagnosis of COPD, and only 11 (10 %) were admitted for exacerbation of COPD, suggesting that at least a similar sensitivity could be expected. A similar accuracy in identification of IEs was reported for the software used in the Blanch et al. study [14]. Furthermore, the main results of the study were the same in an analysis of a subgroup of patients, excluding those with COPD (Table S8).

A rather prolonged duration of ventilation and ICU stay was observed in our patients, which could be attributed to the exclusion of patients on CPAP or low assist, and those expected to proceed to a T-piece trial within 24 h of initiation of assisted ventilation. This could be regarded as one of the strengths of the study, as monitoring of events would be implemented in everyday practice in patients expected to have a relatively long weaning period. However, in our study, the patients were not rigorously classified into weaning category [19].

Finally, this work was not designed to study the cause of IEs or events, and cannot clarify to what extent the presence of events has a causal relationship with patient outcome. It is reasonable to assume that more severely ill patients having ICU-acquired weakness would have more IEs, and would also require prolonged mechanical ventilation [20]. Yet, there are other possible mechanisms by which IEs, and particularly events, could be associated with adverse patient outcome. For example, the presence of IEs during expiration would cause pleiometric contraction to the diaphragm, damaging muscle fibers [21, 22]; discomfort could induce stress [13, 23]; and unrecognized IEs could lead to mistakes in decision-making during weaning [24, 25]. Whether and to what extent IEs are correctable, and whether that would affect patient outcome, were not examined in this study, nor, to our knowledge, in any other past study. However, it is reasonable to assume that appropriate ventilator alarms would significantly facilitate research in this important issue.

In conclusion, this study introduces the concept of events to describe the clusters of ineffective efforts. Notwithstanding that the thresholds used for event definition were rather arbitrary, the presence of events, as well as their power and duration, are associated with prolonged duration of mechanical ventilation and higher hospital mortality. As the computation of an event can be performed in real time, through the use of appropriate software, events could be introduced as alarms on ventilators to facilitate the management of ineffective efforts and improve patient–ventilator interaction.