Unplanned Extubations: Rare Event Monitoring for Continuous Quality Improvement

Mechanical ventilation is the most common technological support provided to ICU patients, and unplanned extubation is a well-known, associated life-threatening event. Unplanned extubation is generally expressed as the number of unplanned extubations per patient on ventilation or as the number of unplanned extubations per 100 d of mechanical ventilation. The prevalence of unplanned extubations varies widely in the literature. A 2012 systematic review reported rates that ranged from 0.5 to 35.8 (median, 7.3) or from 0.1 to 4.2 unplanned extubations per 100 ventilator days (median, 0.9).¹ A 2022 systematic review reported a similar pooled prevalence of 6.7% and incidence density of 1.06 events per 100 ventilator days.

In this issue of Respiratory Care, Harbrecht et al² describe a decade of analysis of unplanned extubation etiology in trauma patients, including the impact of the COVID pandemic. In 2012, the institution identified a “high rate” of unplanned extubation in trauma patients (8.0%; 1.20 per 100 ventilator days).² In 2013, in response, the authors began to track these events, the reasons for the unplanned extubation, and initiated interventions to mitigate unplanned extubation. They implemented a multidisciplinary multifaceted education and direct feedback program. They continued to track the incidence of unplanned extubation and causes, and incorporated data review into their educational programs and presentations to physician, nurse, and respiratory therapist leadership. In examining the etiology of unplanned extubation, the authors categorized unplanned extubation related to patient factors, mechanical factors, or provider factors. As with previous studies, most cases of unplanned extubation were associated with self-extubation (a patient factor). Patient factor–related unplanned extubations were associated with the lowest rates of re-intubation and more favorable outcomes than other factors.

Overall, their program was successful, which demonstrated a reduction of unplanned extubation from their pre-implementation (2013) baseline rate of 8.0% with 1.15 unplanned extubations per 100 ventilator days to a post-intervention rate of 4.0% with 0.60 unplanned extubations per 100 ventilator days. Bolstering their conclusions based on the before and after pooled unplanned extubation numbers is the included rare-event control chart, a more effective tool for ongoing monitoring of rare events. Statistical process control (SPC) methods trace back to their development by Walter A. Shewart at Bell Laboratories in 1920 and were incorporated more broadly into quality improvement methods by W. Edwards Deming. SPC charts (control or Shewhart charts) contribute to quality improvement in health care through providing visualization and monitoring of health-care processes and outcomes. Data are plotted on the y axis and time is plotted on the x axis. There are multiple types of SPC charts, and the most appropriate choice of chart relates to the type of measure (count versus attribute data) frequency and/or amount of data and the goals of the measurement. Control chart limits are calculated for each data point and provide a reference for application of established rules that determine if the data point or points represent random versus non-random change. Proper chart selection and sampling strategy is critical to monitoring for and detecting change. Control charts are frequently used to demonstrate a process or outcome shift after a quality intervention.³

Whereas pooled before and after intervention summary statistics may accurately reflect change, they can be misleading. Data graphed over time on either run charts or control charts by using their respective interpretive rules can provide more insight into whether the observed variation is either “common cause” (random) or “special cause” (non-random) variation. Common cause variation is the inherent variability of a system, whereas special cause variation corresponds to a change. There are numerous resources available that summarize these points, including the United Kingdom National Health Service's A Guide to Creating and Interpreting Run and Control Charts and Public Health Scotland's SPC Tutorial Guide.^4,5

Rare events, for example, unplanned extubation, pose a problem with respect to data analysis. Many intervals, depending on the time period chosen, will have few or no events to count. Dichotomizing quality improvement measures and attempting to analyze them in conventional attribute or variable SPC charts may make detecting change difficult or may even be misleading. This scenario is best addressed through application of rare-event control charts, with g-charts and t-charts being two commonly used SPC charts.^6,7 Exponentially weighted moving average charts and the cumulative sum charts, although more complex, may be more sensitive at detecting early change.

Rare-event SPC charts provide a continuous data set and allow for more informative analysis of the system. G-charts are intended for scenarios in which the time between events is a count (number of opportunities for the event) and t-charts are used when the time between events is a measurement. G-charts could be applied for measuring the number of intubations between events of unrecognized esophageal intubation. Each point on the chart would represent the number of intubations (y axis) that occurred between esophageal intubations. To use a g-chart, the number of opportunities for the event (ie, intubations) must be known. When the number of opportunities for the event is unknowable or unknown, a t-chart can be used to plot the elapsed time interval between events. Rare event charts require fewer data to effectively detect non-random change.

Harbrecht et al² summarized a reduction of their unplanned extubation rate from 8.0% pre-intervention to 4.0% (pooled average) post-intervention; impressive indeed. Their t-chart (manuscript Fig. 1 A) of days between each unplanned extubation supports their conclusion of improvement after intervention.² The chart demonstrates an overall shift toward greater time intervals between unplanned extubations post-intervention (represented by the arrow). All 6 of the “astronomical points” (an indicator of non-random change), those above the upper control limit, occurred after the intervention, and, although the figure is compact, it appears that, more points are above the center line post-intervention. The use of a g-chart by Harbrecht et al² is commendable. Although rare-event charts are used in health care, given the expansive body of health-care–related rare-event–related quality improvement literature, they are underutilized. In a 2022 paper that evaluated the limitations of dichotomizing quality improvement measures and the merits of using g-charts, the researchers found only 14 results in PubMed when searching for “G chart.”⁷

However, the same g-chart by Harbrecht et al² also demonstrates a considerable number of post-intervention points below the average, similar to the pre-intervention state. Thus, the pooled average pre- and post-intervention and the g-chart tell slightly differing stories. Harbecht et al² also strove to tease out the etiologies that drive the unplanned extubation rate and found, post-intervention, a decrease in those unplanned extubations related to patient factors but no change in those related to either mechanical or provider factors. So, what is the point of all this discussion about unplanned extubations, their etiology, and rare-event charts, and how do we tie these all together?

Unplanned extubations are rare events with potential serious adverse consequences in some but not all patients. There is opportunity for institutions to monitor unplanned extubations continuously via rare-event reporting. An argument could be made to treat each rare adverse health-care event as a serious event to be investigated for assignable cause and abandon the concept that there is an acceptable baseline and/or common cause incidence.⁸ However, we are still optimizing our understanding of the etiology of unplanned extubation and its relationships to other processes and outcomes. Unplanned extubations related to “patient factors” (primarily manifesting as self extubation) may have more drivers involved to examine and potential issues to address. Unintended consequences, such as deeper sedation, limiting sedation holidays, or burdening staff, that may also have risk to reduce self-extubation must also be considered. These potential consequences were not explored in the current study.

A system of identifying, counting, analyzing, and reporting and of incident review and closed loop feedback seems the most effective strategy in the current state and that is exactly what Harbrecht et al² describe. There are a host of potential process factors that could relate to self extubation. As the authors hypothesize, perhaps some unplanned extubation events involve patients who should have been intentionally extubated sooner. Tracking and investigating these events in detail and those patients who did not require re-intubation could help build further knowledge and methods to more accurately identify patients who are candidates for “earlier” extubation and potentially reduce the risks of prolonged mechanical ventilation, resource utilization, and unplanned extubation events.

Simultaneously, striving for zero unplanned extubations related to mechanical and provider factors seems a reasonable goal. As the current study by Harbrecht et al² suggests, this population of unplanned extubations related to mechanical and provider factors is associated with a higher rate of complication. Interestingly, the frequency of unplanned extubations related to mechanical and provider factors did not change over the study period. Quality improvement projects to reduce unplanned extubation can be fruitful, but it is not enough. It is imperative that we use a strategy of continuous quality improvement that involves ongoing surveillance by using tools that detect early change. As systems mature and etiologies for unplanned extubation are better understood, adding surveillance tools to monitor related drivers such as nursing staffing ratios, time interval from extubation readiness testing to extubation, intensivist staffing, sedation and/or agitation scores could assist in optimizing processes that are drivers for unplanned extubation.

Using multiple SPC charting tools can provide different insights on the same data, detect and confirm special-cause variation earlier shine light on a potentially correctible cause of the defects. For example, in 2014, one medical center formed a team to decrease its rate of unplanned extubation in its neonatal ICU. Over the years, as their knowledge grew, they evolved to simultaneously use u-charts, g-charts, and exponentially weighted moving average charts, each with its own attributes. In 2016, a cluster of unplanned extubations was noted and special cause was first detected on the exponentially weighted moving average chart. The signal prompted an investigation by the lead respiratory therapist who discovered that the endotracheal tube tape had been replaced by a different, similar appearing product because the old tape had been discontinued by the manufacturer. The team chose a replacement tape by using an in-depth quantitative and qualitative evaluation process that involved multiple staff and simultaneous unplanned extubation surveillance and unplanned extubation performance returned to previous baseline.⁹

The current study by Harbrecht et al² offers valuable insight into a single center's experience in reducing unplanned extubation events. Because patient factors and self extubation are the primary drivers of unplanned extubation rates, reduction may be intertwined with improvements in extubation readiness assessment and execution. The future of predicting successful extubation in individual patients will likely use machine learning. More realistic for the near term, tools can be built into the electronic health record to facilitate digitally capturing the unplanned extubation via structured fields in the electronic health record, an affordance to the health-care worker documenting and simultaneously to reporting. The categorization of patient-related factors, mechanical factors, or provider factors may be useful in designing interventions. Real-time up-to-date reporting that uses a combination of SPC charts, timely review, and feedback can improve unplanned extubation rates.