Impact of performance improvement strategies on unplanned extubation in an inner-city intensive care unit

Abstract

Background:

Unplanned extubation (UE) in intensive care units (ICUs) is a significant patient safety concern, associated with increased morbidity and healthcare utilization; the reported rates of UE vary from 1% to 15%. There is sparse data on the effects of multiple performance improvement (PI) strategies to decrease the rate of UE, particularly in inner-city ICU populations. This study evaluates the impact of PI strategies on UE rates and associated patient outcomes in an adult ICU.

Objectives:

To determine the impact of performance improvement (PI) strategies on rates of unplanned extubation (UE), reintubation, tracheostomy, mortality, and length of hospital stay in ICU patients.

Design:

Retrospective cohort study

Methods:

This retrospective observational study included 6,397 mechanically ventilated patients admitted to a single tertiary ICU between 2015 and 2023. Three distinct time periods were compared: Period 1 (2015–2017, pre-PI), Period 2 (2018–2020, early-PI), and Period 3 (2021–2023, sustained-PI). Demographics, sedation practices, UE characteristics, and outcomes were analyzed using logistic regression.

Results:

UE incidence declined significantly from 3.79% in Period 1 to 2.17% in Period 3 (p = 0.002). Reintubation rates dropped from 45.2% to 26.7% (p = 0.011), and tracheostomy rates from 19.0% to 2.2% (p < 0.001). Multivariate analysis showed reduced odds of reintubation in Periods 2 (OR = 0.219, p = 0.001) and 3 (OR = 0.345, p = 0.021) and reduced odds of tracheostomy in Period 3 (OR = 0.011, p = 0.016). Risk factors for reintubation included the absence of prior intubation history and not undergoing spontaneous breathing trials. Older age (⩾71 years) and positive urine toxicology for opiates were strongly associated with tracheostomy.

Conclusion:

Implementation of PI strategies significantly reduced rates of unplanned extubation, reintubation, and tracheostomy. These findings support continued quality improvement initiatives in ICU airway management.

Plain language summary

Challenges in a urban city ICU-unplanned removal of endotracheal tube and strategies to prevent it

Endotracheal tubes (ET tubes) are breathing tubes that are used to provide respiratory support to a patient from a ventilator. Since unintentional removal of ET tubes burdens both patients and the patient care team, several studies have been done to look for strategies to prevent it. There are various reasons why a ET tube removal can happen in an unplanned manner. This usually depends on risk factors in the patient which puts them at a higher risk of either self removal of the tube or it could be due to environmental factors such as different methods to secure the tube and different medications used to keep the patient comfortable while breathing with a ventilator. Regardless, every unplanned removal of ET tube carries a risk to the patient and can potentially lead to more complications. Urban city ICUs cater to a especially high risk patient population who are more critically ill and have unique disease pathologies. At our inner city hospital ICU, we explored and refined strategies to prevent these unplanned episodes and studied their effectiveness over almost 10 years. We divided our study period into 3 phases- before any strategy was implemented, after the 1st round of intervention and after the 2nd round of interventions. We found that a systemic way of implementing specific strategies has shown to reduce both the chances of unplanned removal event occurring and the complications occurring from it, specifically the need of replacing a removed ET tube and the need of long term ventilator. We did not find any change in death rate or how long the patients spends in the ICU even after these strategies were implemented. We recommend that a team approach including different levels of medical staff, nurses and respiratory staff to improve outcomes with unplanned extubations. More research is needed to further understand even better strategies in urban centers. We hope our study serves as a direction in which inner city hospital like ours can improve patient outcomes

Introduction

Unplanned extubation (UE) is the unintended and premature removal of an endotracheal tube, recognized as a serious and frequent adverse event in the intensive care unit (ICU). UE rates among intubated ICU patients vary widely, from less than 1% to as high as 15%, depending on the patient population, ICU setting, and sedation practices.^1–3 These rates tend to be higher in urban hospitals, which often manage more critically ill and complex patient populations.

UE is associated with multiple risk factors. Patient-related contributors include agitation, delirium, and suboptimal sedation, while clinical management factors encompass inconsistent sedation protocols and variable use of physical restraints. In addition, system-level issues such as inadequate staffing and suboptimal bedside monitoring may further increase the risk of UE.^4,5 The clinical consequences of UE are substantial, often leading to emergency reintubation, ventilator-associated complications, prolonged ICU length of stay, and increased healthcare costs.

To address these risks, many ICUs have adopted performance improvement (PI) initiatives aimed at reducing UE incidence. These strategies typically involve multidisciplinary education, standardized checklists, protocol-driven sedation management, and enhanced ICU workflow optimization.^6–8 While these interventions have demonstrated efficacy in reducing UE rates, inner-city hospitals face unique challenges, such as higher patient acuity, resource limitations, and increased variability in staff experience. As a result, UE remains a persistent challenge in these settings.

In this study, we conducted a retrospective analysis of UE incidents in an inner-city ICU in New York City over a 10-year period. We examined trends in UE incidence, associated risk factors, and clinical outcomes across three distinct time periods, each marked by the implementation of specific PI interventions. By evaluating changes over time, our goal was to assess the effectiveness of these UE reduction strategies and to identify evolving risk factor profiles and patient outcomes in this high-risk setting.

Objectives

1. To evaluate the impact of performance improvement (PI) strategies on the incidence of unplanned extubation (UE), reintubation, and tracheostomy among ICU patients.

2. To assess the effect of PI strategies on ICU mortality rates and length of hospital stay.

Methods

Definition

Unplanned extubation was defined as the unintended, premature removal of an endotracheal tube, including deliberate self-extubation by the patient or accidental extubation.

Reintubation was defined as the need for re-institution of mechanical ventilation within 48 h following extubation.

Study design

This was a retrospective, observational, single-center cohort study conducted at BronxCare Hospital Center, an 859-bed community teaching hospital serving the South and Central Bronx. The reporting of this study conforms to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement. ⁹

The study included all adult ICU patients who experienced an unplanned extubation between January 1, 2015, and December 31, 2023. The study period was divided into three distinct time periods, each spanning 3 years:

• Period 1 (2015–2017): Pre-implementation of PI initiatives

• Period 2 (2018–2020): Post-initial PI implementation

• Period 3 (2021–2023): Post-additional PI implementation

The performance improvement initiatives were implemented in 2 steps: at the start of period 2 in 2018 and at the start of period 3 in 2021. Time intervals of 3 years were chosen to allow for an even distribution of the number of patients in each interval.

Setting

The hospital includes a 36-bed adult ICU that provides medical, cardiac, and surgical critical care. The ICU is staffed by a multidisciplinary team consisting of critical care nurses, respiratory therapists, senior medical and surgical residents, pulmonary and cardiology fellows, and attending physicians specializing in surgical, cardiology, and pulmonary critical care. Staffing workload in the ICU includes two respiratory therapists available 24 h a day and an average nurse-to-patient ratio of 1:2. The average occupancy rate for our ICU ranged between 80% and 95% during the study period.

Participants

Inclusion criteria

• all adult patients (⩾18 years) in the ICU who were intubated and mechanically ventilated for >24 h between 2015 and 2023

Exclusion criteria

• Patients who were extubated electively

• Patients who required mechanical ventilation for less than 24 h

• Patients who expired in the emergency department (ED) prior to ICU admission

• Patients admitted to other hospital units outside the ICU

• Patients who were transferred to another facility while intubated

Variables

Unplanned extubation rates were calculated as the number of UEs divided by the total number of mechanically ventilated ICU patients within each study period.

Reintubation rate was defined as the proportion of UE events that resulted in the patient being re-intubated within 48 h.

Data sources

Data were obtained from a retrospective chart review covering the period from January 2015 through December 2023.

Bias

No sources of bias were identified

Study size

A study size of 6397 mechanically ventilated patients was used for analysis of the rate of unplanned extubation. Analysis of all other outcomes was done on a study size of 179 patients who underwent unplanned extubation (Figure 1).

Figure 1.

Flowsheet of study methodology and interventions.

The timeline represents the duration of data collection. The timeline is divided into three periods as indicated on the left of the timeline. Significant steps in the study are described on the right side of the timeline.

UE, unplanned extubation.

Performance improvement initiatives

Sedation level was assessed at least every 4 h using the Riker Sedation-Agitation Scale (SAS), targeting a SAS of 2–4 (very sedated or light sedation). UE events were identified by ICU staff and managed promptly per institutional protocol. Initial response involved immediate airway assessment and noninvasive respiratory support by a team consisting of an ICU physician, nurse, and respiratory therapist. If the patient was in respiratory distress, unable to oxygenate, ventilate, or protect the airway, reintubation was performed. Otherwise, patients were managed noninvasively and reassessed periodically over the next 48 h.

PI interventions were implemented in two phases, demarcating three distinct study periods.

Phase I interventions (implemented January 2018; start of period 2)

Following the establishment of a multidisciplinary Unplanned Extubation Committee in November 2016—including the ICU Director, Nurse Manager, Director of Respiratory Therapy, and a Pulmonary Fellow—key PI initiatives were identified and implemented after review and approval by the ICU PI Committee.

• (A) Standardized Endotracheal Tube Securing Method

• ● Transition from tape-based fixation to ETT holders to enhance tube stability and reduce accidental dislodgement and skin injury.

• (B) Revision of Spontaneous Awakening and Breathing Trials (SAT/SBT) Protocols

• ● Sedation tapering began daily at 6:00 a.m., following a multidisciplinary assessment by the on-call pulmonary fellow, RT, and ICU charge nurse.

•  SAT/SBT assessments were standardized and scheduled, in contrast to prior unstructured post-round decisions.

• (C) Sedation practice modifications

• ● Benzodiazepines (e.g., midazolam) were discouraged unless clinically necessary.

• ● Dexmedetomidine and other short-acting sedatives were promoted to reduce the risk of oversedation and UE.

Note: Sedation and restraint data were abstracted from the 24-h period preceding the UE event to preserve temporal validity. Seasonal and staffing variability were not modeled due to data limitations, although ICU workflows and staff ratios remained stable throughout the study.

Phase II interventions (implemented January 2021; start of period 3)

Based on continued data review from Period 2, additional strategies were introduced to further reduce UE rates and improve safety.

• (A) Implementation of Visual Communication for Difficult Airways

• ● Patients with high-risk airway features (e.g., documented difficult intubation, macroglossia, BMI > 35, Mallampati class III–IV, neck immobility) were flagged using a visual bedside indicator.

• ● The system aimed to alert staff and prompt heightened vigilance.

• (B) Modification of Sedation Guidelines

• ● Reduction in fentanyl use was emphasized, with ongoing use of SAS-guided sedation assessments and individualized management strategies.

• (C) Quarterly Nursing Education

• ● Structured in-service training focused on:

  • ○ UE risk factors and prevention
  • ○ Sedation and restraint practices
  • ○ Proper application of SAT/SBT protocols

• (D) Multidisciplinary House Staff Education

• ● At the start of each ICU rotation, house staff received training modules on:

  • ○ UE prevention strategies
  • ○ Weaning practices and sedation management

• ● Quarterly sessions were extended to pulmonary fellows, attending physicians, and RT staff.

• (E) Expansion of the ICU UE Committee

• ● The committee expanded to include additional pulmonary fellows and appointed an intensivist champion to:

  • ○ Review UE cases in real time
  • ○ Ensure protocol adherence
  • ○ Promote continuous quality improvement

Outcomes

The primary outcomes of the study were the rate of unplanned extubation, the rate of reintubation within 48 h following a UE event, and the incidence of tracheostomy among affected patients. These outcomes were selected to evaluate the direct impact of PI interventions on airway management and procedural escalation in the ICU.

Secondary outcomes included ICU mortality and hospital length of stay, which were analyzed to assess the broader clinical consequences of UE and the effectiveness of the implemented strategies in improving overall patient outcomes. All outcomes were compared across the three study periods to determine temporal trends and intervention impact.

Statistical analysis

All statistical analyses were conducted using SPSS version 29 (IBM Corp., Armonk, NY, USA), with a two-tailed p-value < 0.05 considered statistically significant. Comparisons of categorical variables across the three study periods were performed using the Chi-square test (χ² test). Length of stay (LOS), a continuous variable, was analyzed using one-way analysis of variance (ANOVA, F-test) to assess differences across time periods. To identify predictors of the primary outcome (unplanned extubation) and secondary outcomes (reintubation, tracheostomy, and mortality), logistic regression analysis was employed, with results reported as adjusted odds ratios (OR) with 95% confidence intervals (CI). Linear regression analysis was used to model LOS and evaluate associations with potential predictors. Multicollinearity among independent variables was assessed using variance inflation factors (VIF), and all variables included in the multivariate models demonstrated VIF values less than 2.0, indicating no significant collinearity. Model calibration for logistic regression was evaluated using the Hosmer–Lemeshow goodness-of-fit test, which confirmed acceptable model performance.

Given that LOS data were right-skewed, a nonparametric test was also applied to verify findings, yielding similar significance levels. In addition, sensitivity analyses using a log-transformed LOS variable were conducted to confirm the robustness of linear regression results. Both the raw and log-transformed models demonstrated consistent findings, and no violations of linear regression assumptions were observed.

A significance level (α) of 0.05 was uniformly applied across all analyses.

Where an overall omnibus test for proportions was significant across periods, we performed post‑hoc pairwise comparisons of column proportions. Period‑level differences are displayed in tables using superscript letters; periods sharing a letter are not significantly different at α = 0.05. No a priori sample‑size or power calculation was performed. This retrospective study included all eligible mechanically ventilated ICU patients between January 2015 and December 2023; precision is conveyed via 95% CI.

Results

A total of 6397 mechanically ventilated patients were included in the study across the three time periods. Of these, 179 patients (2.79%) experienced UE (Table 1). The incidence of UE was highest in period 1 (2015–2017), with 84 patients (3.79%) experiencing UE before the implementation of PI strategies. Following the introduction of PI initiatives, the number of UEs decreased to 50 patients (2.37%) in period 2 (2018–2020) and further declined to 45 patients (2.17%) in period 3 (2021–2023). This reduction in UE rates over time was statistically significant (p = 0.002). Pairwise testing indicated higher UE rates in Period 1 versus Period 2 and Period 3 (Holm‑adjusted p = 0.0298 and 0.0057, respectively), with no difference between Periods 2 and 3 (adjusted p = 0.51).

Table 1.

Rate of unplanned extubation in different time periods.

Period	Total ventilated patients	Unplanned extubation	Percentage	p-Value
1. 2015–2017	2216	84	3.79	0.002
2. 2018–2020	2017	50	2.37
3. 2021–2023	2074	45	2.17

Within row, proportions that do not share a superscript letter differ at α = 0.05 by pairwise two‑proportion z‑tests with Holm–Bonferroni adjustment (P1 vs P2 adj‑p = 0.0298; P1 vs P3 adj‑p = 0.0057; P2 vs P3 adj‑p = 0.51). Cells sharing a letter do not differ (p ⩾ 0.05, Holm-adjusted); cells with different letters differ (p < 0.05).

Comparison of demographic variables

Table 2 presents the demographic characteristics of patients who experienced UE across the three study periods. Over time, the proportion of patients undergoing UE who were above 70 years of age decreased from 28.6% in period 1 (2015–2017) to 22% in period 2 (2018–2020) and 11% in period 3 (2021–2023). Conversely, the percentage of patients aged 46–70 years increased from 44% in period 1 to 56% in period 2 and 62.2% in period 3.

Table 2.

Comparison of demographic variables.

Variables	Categories	Period 1 2015–2017		Period 2 2018–2020		Period 3 2021–2023		Total		p-Value
		N	%	N	%	N	%	N	%
Age years	20–45	23	27.4	11	22.0	12	26.7	46	25.7	0.167
	46–70	37	44.0	28	56.0	28	62.2	93	52.0
	71 and above	24	28.6	11	22.0	5	11.1	40	22.3
Gender	Female	33	39.3	22	44.0	12	26.7	67	37.4	0.195
	Male	51	60.7	28	56.0	33	73.3	112	62.6
Admissions	Emergency department	59	70.2	30	60.0	30	66.7	119	66.5	0.314
	Medical-surgical units	15	17.9	15	30.0	13	28.9	43	24.0
	Operating room	10	11.9	5	10.0	2	4.4	17	9.5
Reason for Intubation	Respiratory failure	52	61.9	26	52.0	26	57.8	104	58.1	0.434
	Cardiac arrest	3	3.6	3	6.0	2	4.4	8	4.5
	Sepsis/metabolic	0	0.0	1	2.0	0	0.0	1	0.6
	Airway protection	19	22.6	14	28.0	12	26.7	45	25.1
	Substance abuse	0	0.0	1	2.0	2	4.4	3	1.7
	Post operative	10	11.9	5	10.0	2	4.4	17	9.5
	Others	0	0.0	0	0.0	1	2.2	1	0.6

Across all study periods, UE was more common in male patients than in female patients. The majority of admissions occurred through the emergency department (ED), reflecting the high acuity of the patient population.

The leading indication for intubation was respiratory failure, accounting for 58.1% of cases, followed by airway protection in 25.1% of patients.

Risk factors for unplanned extubation

Table 3 presents a comparison of risk factors for UE across the three study periods. History of previous intubation and the number of days on MV were not significantly different among the periods. However, there was a significant increase in the proportion of patients with positive urine drug screens for benzodiazepines, opioids, and tetrahydrocannabinol (THC) over time (p < 0.05). The increases were most notable between Period 1 and Periods 2 and 3.

Table 3.

Comparison of risk factors for unplanned extubation.

Variables	Categories	Period 1 2015–2017		Period 2 2018–2020		Period 3 2021–2023		Total		p-Value
		N	%	N	%	N	%	N	%
History of previous intubation		19	22.6	13	26.0	5	11.1	37	20.7	0.168
Days on mechanical ventilation	<2	43	51.2	21	42.0	21	46.7	85	47.5	0.472
	2–5	27	32.1	17	34.0	11	24.4	55	30.7
	>5	14	16.7	12	24.0	13	28.9	39	21.8
Urine drug screen	Benzodiazepine	1	1.3%	8	16.0	12	26.7	21	12.4	<0.001
	Opiates	2	2.6	9	18.0	9	20.0	20	11.6	0.004
	Phencyclidine	2	2.6	2	4.0	1	2.2	5	2.9	0.855
	Tetrahydrocannabinol	1	1.3	7	14.0	8	17.8	16	9.3	0.004
	Cocaine	6	7.1	6	12.0	7	15.6	19	10.6	0.313
History of alcohol use		3	3.6	5	10.0	2	4.4	10	5.6	0.272
Use of mechanical restraints		62	73.8	33	66.0	35	81.4	130	73.4	0.272
Sedation agents	Propofol	26	31.0	19	38.0	15	33.3	60	33.5	0.705
	Fentanyl	29	34.5	24	48.0	15	33.3	68	38.0	0.227
	Dexmedetomidine	4	4.8	7	14.0	15	33.3	26	14.6	<0.001
	Midazolam	12	14.3	1	2.0	2	4.4	15	8.4	0.025
Actively undergoing weaning trial		36	42.9	17	34.0	23	51.1	76	42.5	0.241
Time of unplanned extubation	8 a.m.–2 p.m.	24	28.6	16	32.0	15	33.3	55	30.7	0.023
	2 p.m.–8 p.m.	20	23.8	15	30.0	9	20.0	44	24.6
	8 p.m.–2 a.m.	22	26.2	4	8.0	3	6.7	29	16.2
	2 a.m.–8 a.m.	18	21.4	15	30.0	18	40.0	51	28.5

Rates of alcohol use and utilization of mechanical restraints remained stable across the study periods. Notably, the use of dexmedetomidine for sedation increased markedly from 4.8% in Period 1 to 14.0% in Period 2, and further to 33.3% in Period 3 (p < 0.001). Post-hoc analysis revealed that the greatest difference occurred between Period 1 and Period 3. In contrast, the use of continuous midazolam infusion was most prevalent in Period 1 and declined thereafter.

While the proportion of UE events occurring during weaning trials remained consistent across all three periods, a significant temporal shift was observed in the timing of UE events. The proportion of UEs occurring between 8:00 p.m. and 2:00 a.m. decreased significantly over time, from 26.2% in Period 1 to 8.0% in Period 2 and 6.7% in Period 3 (p = 0.023). This suggests that the timing-related changes were most prominent between Period 1 and the subsequent periods.

Table 4 summarizes the primary and secondary outcomes across the three study periods. The rate of reintubation in period 1 (45.2%) was significantly higher than in the other periods (22.0% and 26.7%, p-value = 0.011). Pairwise comparisons revealed significant differences between Period 1 and Periods 2 and 3, with no significant difference between Periods 2 and 3.

Table 4.

Comparison of outcomes.

Variables	Categories	Period 1 2015–2017		Period 2 2018–2020		Period 3 2021–2023		Total		p-Value
		N	%	N	%	N	%	N	%
Reintubation		38	45.2	11	22.0	12	26.7	61	34.1	0.011
Tracheostomy		16	19.0	0	0.0	1	2.2	17	9.5	<0.001
Length of hospital stay (Mean, SD)		24.49	48.65	20.30	17.96	22.53	17.08	22.83	35.60	0.805
Disposition	Home	46	54.8	23	47.9	24	53.3	93	52.5	0.138
	Nursing home	23	27.4	20	41.7	10	22.2	53	29.9
	Death	9	10.7	5	10.4	5	11.1	19	10.7
	Interinstitutional transfer	6	7.1	0	0.0	6	13.3	12	6.8
Mortality		8	9.5	5	10.0	5	11.1	18	10.1	0.960

Similarly, the incidence of tracheostomy following UE was highest in Period 1 (19.0%) and declined sharply in subsequent periods, with no tracheostomies reported in Period 2 and a rate of 2.2% in Period 3 (p < 0.001). This change was statistically significant when comparing Period 1 to both Period 2 and Period 3.

There were no statistically significant differences in average LOS, ICU mortality, or discharge disposition across the three time periods, indicating that while airway-related complications improved, broader clinical outcomes remained stable.

Predictors of reintubation following unplanned extubation

Logistic regression analysis identifying factors associated with reduced odds of reintubation following unplanned extubation is presented in Table 5. Patients in Periods 2 and 3 demonstrated significantly lower odds of reintubation compared to those in Period 1. In univariate analysis, period 2 had a 65.9% lower odds of reintubation (OR = 0.341, 95% CI: 0.15–0.76, p = 0.008), while period 3 had a 56% lower odds (OR = 0.440, 95% CI: 0.20–0.97, p = 0.041). These associations remained significant in multivariate analysis, with adjusted ORs of 0.219 (95% CI: 0.09–0.55, p = 0.001) for period 2 and 0.345 (95% CI: 0.14–0.85, p = 0.021) for period 3.

Table 5.

Logistic regression results with reintubation as the dependent variable.

Variables	Univariate			Multivariate
	OR	95% CI	p-Value	OR	95% CI	p-Value
Period 2 – 2018–2020	0.341	(0.15–0.76)	0.008	0.219	(0.09, 0.55)	0.001
Period 3 – 2021–2023	0.440	(0.20–0.97)	0.041	0.345	(0.14, 0.85)	0.021
History of previous intubation	0.306	(0.12–0.78)	0.013	0.234	(0.08, 0.68)	0.007
Use of mechanical restraints	0.531	(0.27, 1.05)	0.070	0.360	(0.16, 0.82)	0.016
Actively undergoing weaning trials	0.484	(0.25, 0.93)	0.029	0.347	(0.16, 0.74)	0.006

A history of prior intubation was also associated with a significantly lower risk of reintubation (OR = 0.306, 95% CI: 0.12–0.78, p = 0.013). The odds of reintubation were 69.4% lower in patients with prior intubation history in univariate analysis (OR = 0.306, 95% CI: 0.12–0.78, p = 0.013), and this association persisted after adjustment for covariates (OR = 0.234, 95% CI: 0.08–0.68, p = 0.007).

Although the use of mechanical restraints was not significantly associated with reintubation in univariate analysis (OR = 0.531, 95% CI: 0.27–1.05, p = 0.070), it reached statistical significance in the multivariate model, where restraint use was associated with a 64.0% reduction in the odds of reintubation (OR = 0.360, 95% CI: 0.16–0.82, p = 0.016).

Participation in a spontaneous breathing trial (SBT) prior to UE was also protective. Patients undergoing SBTs had 51.6% lower odds of reintubation in univariate analysis (OR = 0.484, 95% CI: 0.25–0.93, p = 0.029), and this association was strengthened in the multivariate model (OR = 0.347, 95% CI: 0.16–0.74, p = 0.006).

Predictors of tracheostomy following unplanned extubation

Logistic regression analysis evaluating factors associated with tracheostomy in patients who experienced unplanned extubation is presented in Table 6.

Table 6.

Logistic regression results with tracheostomy as the dependent variable.

Variables	Univariate			Multivariate
	OR	95% CI	p-Value	OR	95% CI	p-Value
Period 3 (2021–2023)	0.097	(0.01, 0.75)	0.026	0.011	(0.0, 0.43)	0.016
Age – 71 year-age or older	3.882	(0.74, 20.45)	0.110	14.007	(1.41, 139.43)	0.024
Urine toxicology positive for opiates	1.015	(0.21, 4.81)	0.985	45.253	(1.89, 1.42)	0.019
Fentanyl use for sedation/analgesia	1.964	(0.72, 5.36)	0.188	3.340	(0.98, 11.36)	0.053

In univariate analysis, period 3 (2021–2023) was associated with a 90.3% lower odds of tracheostomy compared to period 1 (2015–2017; OR = 0.097, 95% CI: 0.01–0.75, p = 0.026). Use of continuous midazolam infusion was also significantly associated with increased odds of tracheostomy (OR = 4.22, 95% CI: 1.18–15.1, p = 0.027). Other variables, like age and urine toxicology results, were not significant in the univariate model.

In the multivariate logistic regression model, four variables remained independently associated with tracheostomy. Period 3 continued to demonstrate a strong protective effect, with an adjusted odds ratio of 0.011 (95% CI: 0.00–0.43, p = 0.016), indicating a 98.9% reduction in the odds of tracheostomy relative to Period 1.

Age ⩾71 years emerged as a significant risk factor, with these patients exhibiting markedly increased odds of tracheostomy (adjusted OR = 14.007, 95% CI: 1.41–139.43, p = 0.024), suggesting that advancing age contributes to prolonged mechanical ventilation and airway dependency.

Additionally, a positive urine toxicology screen for opiates was strongly associated with the need for tracheostomy. The adjusted odds ratio was 45.253 (p = 0.019), indicating a significant association between opioid use and increased risk of prolonged ventilation following UE.

Discussion

The incidence of UE varies widely across ICUs, with reported rates ranging from 0.3% to 14%, reflecting differences in patient populations, care practices, and institutional resources.^10–12 In our study, the overall UE incidence among patients on MV decreased significantly with the implementation of PI initiative, from 3.79% in 2015–2017, the baseline period, to 2.17% after implementation of PI initiatives, representing a relative reduction of approximately 43% (p = 0.002). These findings place our UE rates (2%–4%) at the lower end of reported ranges. A 2022 meta-analysis by Li et al., encompassing 38 studies, estimated a pooled UE prevalence of 6.7% (95% CI: 5.3%–8.2%) in adult ICUs. ³

We attribute the observed downward trend in UE incidence in our study to the PI interventions implemented in our ICU. Previous research has similarly demonstrated that protocolized interventions—particularly those involving standardized sedation, structured weaning practices, and staff education—can significantly reduce UE incidence. ¹³ However, much of the existing literature has focused on pediatric and neonatal ICUs,^14–18 with relatively few studies conducted in adult populations. A systematic review of adult ICU interventions reported reductions in UE rates of 22% to 53% following implementation of structured protocols and multidisciplinary education programs. ¹⁹ Our findings reinforce the value of evidence-based, system-level changes in adult critical care environments.

We observed a shift in the timing of UE events, with significantly fewer events occurring during late-night hours (8 p.m.–2 a.m.) in later periods. The proportion of UEs occurring late at night decreased from 26.2% in 2015–2017 to 7% in 2021–2023. Prior studies have suggested that UEs frequently occur overnight or during shift changes, when monitoring and staff-to-patient ratios may be suboptimal and patients are more vulnerable to delirium or circadian rhythm disturbances.^20–22 Our results suggest that improvements in sedation and structured early morning weaning protocols may have helped reduce late-night UE events.

Sedation practices evolved considerably over the study period. Midazolam use decreased from 14.3% of UE cases in Period 1 to approximately 4% in Period 3, while dexmedetomidine use rose to 33.3% during the same interval. These trends align with existing evidence that benzodiazepines, especially midazolam, are associated with paradoxical agitation,^23,24 increased delirium risk, ²⁵ and higher UE rates.^6,23,26 Our PI strategy explicitly discouraged benzodiazepine use, promoting alternative agents such as dexmedetomidine, which is associated with lighter, more controllable sedation profiles and fewer adverse effects.

Among the primary outcomes assessed, reintubation rates declined significantly across the three study periods (p = 0.011). While the overall reintubation rate was 34%, the rate dropped from 45.2% in Period 1 to 22.0% and 26.7% in Periods 2 and 3, respectively. This trend is consistent with published data, although wide variability exists in reported rates (15–88%).^19,27–30 A meta-analysis estimated that 50.2% of UE patients require reintubation within 48 h. ³ Although reintubation is strongly associated with increased mortality and prolonged ICU stays,^12,27 we did not observe significant differences in either ICU mortality or LOS over time.

The rise in opioid-related overdoses in our region—particularly involving fentanyl—adds important epidemiologic context. Between 2010 and 2020, drug overdose death rates in the Bronx rose from 8.7% to 38.7%. ³¹ Opioid toxicity complicates sedation management and may contribute to both UE risk and reintubation. In one study of overdose patients in a toxicology ICU, the UE rate was 12.1%, with a 50% reintubation rate and 15% mortality. ³² These findings exceed rates typically seen in general ICU populations. Although our study was underpowered to detect statistical differences related to opioid use, logistic regression showed that a positive urine toxicology screen for opiates was strongly associated with increased odds of tracheostomy, warranting further investigation in larger cohorts.

Logistic regression analysis revealed that a history of prior intubation, unplanned extubation during a weaning trial, and the use of mechanical restraints were associated with a lower likelihood of reintubation. The protective effect of prior intubation is novel and may reflect reduced patient anxiety or increased clinician confidence in airway management. It is also possible that this finding reflects unmeasured confounders, such as a lower burden of chronic illness in certain subgroups (e.g., drug overdose patients). This hypothesis remains speculative and should be validated in future research.

The association between UE during an SBT and reduced reintubation risk is well-supported in the literature.^33,34 These events may represent functionally successful extubations that occurred slightly ahead of schedule. Similarly, the observed association between mechanical restraint use and lower reintubation risk contrasts with some prior studies that have linked restraint use with increased UE risk.^19,35,36 One possible explanation is that patients who are restrained may be more alert and physiologically prepared for weaning, reducing the likelihood of respiratory failure after UE.

Our analysis also showed a significant decline in tracheostomy rates over time, which parallels evolving sedation strategies and weaning protocols. Notably, midazolam use was associated with increased tracheostomy rates. While causality cannot be established, this finding supports the trend in critical care practice that discourages benzodiazepine use due to its association with prolonged ventilation and poor outcomes.

Despite these favorable trends in primary outcomes, we found no significant differences in mortality or hospital LOS across the three study periods. Prior studies have consistently reported prolonged ICU and hospital stays among patients experiencing UE.^37,38 A recent 2023 study found that patients with UE had a median ICU stay of 12 days, compared to 10 days in matched controls. ³⁹ Our findings suggest a non-significant trend toward shorter hospital stays in later periods, but further research is needed.

The relationship between UE and mortality is complex. While reintubation following UE has been linked to increased mortality in some studies, overall mortality rates among patients experiencing UE do not appear consistently higher than those without UE. Our ICU’s standardized protocol of early assessment and prompt reintubation may have mitigated adverse outcomes. This is consistent with data showing no mortality difference unless reintubation is required.^10,40,41 A recent multicenter study by Guillemin et al., involving 47 French ICUs, also found no significant difference in mortality between planned and unplanned extubations. ¹

Limitations

This study has several limitations. First, as a single-center retrospective analysis, our findings may not be generalizable to ICUs with different patient populations, staffing models, or levels of specialization. Second, the observational nature of the study limits causal inference. Third, unmeasured confounding variables such as seasonal illness patterns, ICU occupancy rates, and concurrent quality improvement initiatives may have influenced outcomes. Fourth, we did not incorporate standardized severity-of-illness scores (e.g., APACHE II, SOFA), although baseline demographics and risk factors remained consistent across study periods (Tables 2 and 3), suggesting comparable patient acuity. Finally, because no a priori power calculation was performed and several subgroup events were infrequent (e.g., tracheostomy in later periods), the study may be underpowered to detect modest between‑period differences in some outcomes; estimates should be interpreted with their CI.

Interpretation

Our findings support that PI strategies involving standardized sedation practices, structured weaning protocols, and multidisciplinary education are associated with reductions in UE, reintubation, and tracheostomy among mechanically ventilated ICU patients.

Generalizability

This study was conducted at a large urban community hospital serving a high-density population with a substantial burden of substance use disorders. The literature remains sparse regarding UE prevention in such high-risk populations. Despite this, the strategies implemented in our ICU are broadly applicable and likely feasible in other secondary and tertiary care settings. The wide CI for certain variables (e.g., opioid toxicology) reflect limited sample size and low event frequency; thus, findings should be interpreted cautiously and validated in larger, multicenter cohorts.

Conclusion

Our longitudinal analysis highlights the impact of a sustained, iterative PI process in mitigating unplanned extubation and improving patient outcomes. A multidisciplinary, protocol-driven approach plays a critical role in improving airway safety and optimizing outcomes for mechanically ventilated ICU patients. As unplanned extubation remains a persistent challenge, especially in resource-constrained urban settings, these findings offer actionable insights for other institutions seeking to reduce UE and enhance care for critically ill patients. Further research is warranted to explore the mechanistic underpinnings of identified risk factors and to define best practices for long-term prevention in diverse ICU populations.