The Effect of Emergency Department Crowding on Lung-Protective Ventilation Utilization for Critically Ill Patients

Abstract

OBJECTIVE:

To measure effects of ED crowding on lung-protective ventilation (LPV) utilization in critically ill ED patients.

METHODS:

This is a retrospective cohort study of adult mechanically ventilated ED patients admitted to the medical intensive care unit (MICU), over a 3.5-year period at a single academic tertiary care hospital. Clinical data, including reason for intubation, severity of illness (MPM_0-III), acute respiratory distress syndrome (ARDS) risk score (EDLIPS), and ventilator settings were extracted via electronic query of electronic health record and standardized chart abstraction. Crowding metrics were obtained at 5-minute intervals and averaged over the ED stay, stratified by acuity and disposition. Multivariate logistic regression was used to predict likelihood of LPV prior to ED departure.

RESULTS:

Mechanical ventilation was used in 446 patients for a median ED duration of 3.7 hours (interquartile ratio, IQR, 2.3, 5.6). Mean MPM₀-III score was 32.5±22.7, with high risk for ARDS (EDLIPS ≥ 5) seen in 373 (82%) patients. Initial and final ED ventilator settings differed in 134 (30.0%) patients, of which only 47 (35.1%) involved tidal volume changes. Higher percentages of active ED patients (workup in-progress) and those requiring eventual admission were associated with lower odds of LPV utilization by ED departure (OR 0.97, 95%CI 0.94–1.00; OR 0.97, 95%CI 0.94–1.00, respectively). In periods of high volume, ventilator adjustments to settings other than the tidal volume were associated with higher odds of LPV utilization. Reason for intubation, MPM₀-III, and EDLIPS were not associated with LPV utilization, with no interactions detected in times of crowding.

CONCLUSIONS:

ED patients remain on suboptimal tidal volume settings with infrequent ventilator adjustments during the ED stay. Hospitals should focus on both systemic factors and bedside physician and/or respiratory therapist interventions to increase LPV utilization in times of ED boarding and crowding for all patients.

INTRODUCTION

Lung-protective ventilation (LPV), defined as less than 8 cc/kg of ideal body weight (IBW), is safe and remains the standard of care for mechanical ventilation in patients with ARDS.^1,2 Recent epidemiological work has shown the prevalence of ARDS in emergency department populations to range from 7–9% with significant associated mortality.^3–5 Despite best practice recommendations, utilization of LPV in both the emergency department (ED) and intensive care unit (ICU) is often infrequent in the early phase of critical illness.^6,7 The concept of ventilator inertia has emerged providing evidence that early critical care interventions and ventilator settings are often carried forward into the patient’s hospital and ICU stay.⁸ Newer studies on ED-based LPV mechanical ventilation bundles have shown correlation with improved mortality and decreased ventilator-associated complications.⁹ However, perceived barriers to LPV bundle implementation include higher labor intensity compared to conventional ventilation, delays in diagnostic testing, and variable attitudes toward LPV.^10,11 We aim to investigate the systems factors which serve as obstacles in the implementation of standard-of-care LPV in critically ill ED patients admitted to the ICU, especially for those at risk for ARDS.

While LPV bundles drive improvements in bedside ARDS care, emerging work on ICU triaging and resource management have shown the complexity of care delivery to critically ill patients in the ED.¹² Even after disposition to ICU is established, patients often experience prolonged boarding, typically defined as the period after admission request to physical transfer to the ICU, which has been associated with worse outcomes for critically ill patients.^12,13 Emergency departments in the United States have experienced a 55% increase in critically ill patient presentations in recent decades,¹⁴ as well as a 7% increase in ED length-of-stay (LOS) between 2001 to 2005.¹⁵ ED crowding has been associated with delays in triage, diagnosis, and treatment, as well as increased morbidity and mortality for critically ill patients.^13,16–20 Within an increasingly cost-conscious US healthcare system, various systems factors, patient characteristics, and provider behaviors influence LPV implementation in the ED-to-ICU transition.

Literature on critical care services in the ED has focused on medical education/training, the effects of overcrowding on ICU outcomes, and the epidemiology of ARDS management in initial ICU care.^7,21,22 Yet to be investigated is the implementation science behind critical care services at both a systems and provider level (i.e. physicians and respiratory therapists) in the emerging models for delivering critical care in the ED.

Understanding mechanical ventilation practices during periods of ED capacity strain may play a central role in improving health care delivery within the evolving ED-ICU interface, especially for those patients at higher risk for ARDS.²³ In contrast, to previous work on ARDS in ED and early ICU stay, our work aims to characterize the larger, hospital systems factors that influence the management of those at risk for ARDS. We sought to investigate the complex interactions between patients at risk for ARDS, critical illness, and the systems factors that influence critical care delivery in a large academic medical center. The study’s primary objectives were to 1) characterize LPV utilization for critically ill ED patients requiring mechanical ventilation prior to ICU admission, 2) evaluate the effect of prolonged boarding and ED crowding on LPV utilization, and 3) determine if certain patient risk profiles change the likelihood of LPV utilization. We hypothesize that LPV utilization in the ED decreases in times of increased ED crowding and associated high physician workload. We also hypothesize that patients with higher acuity or higher risk for ARDS will be more likely to be placed on LPV settings by ED departure.

METHODS

Study Setting and Population

This is a single institution retrospective study at a large urban academic tertiary care center with a high-volume Emergency Department (ED) serving more than 100,000 visits annually and an average of 300 visits per day. The ED is a 61-bed unit with a five-bed area designated for high-acuity patients upon arrival. The five-bed area is also used for ongoing management of patients who clinically deteriorate while boarding or while actively undergoing a workup in other sections of the ED. The majority of ICU admissions and/or patients requiring mechanical ventilation are cohorted to these five beds. This area is staffed by an emergency medicine (EM)-trained attending physician and an upper-level EM resident, who also simultaneously care for patients in other sections of the ED. The high acuity section has a typical nursing-to-patient ratio of 1:3 and a shared respiratory therapist (RT) for the entire ED.

Mechanical ventilation settings are managed at the bedside, with both live-streamed data collection into the electronic health record (EHR) respiratory flowsheet and manual documentation of ventilator settings in nursing and RT notes. EHR system orders for mechanical ventilation settings are also entered by ED providers. The tidal volume setting for all new invasive mechanical ventilation orders within the EHR were set for 500 ml, as a default setting, though able to be overwritten with alternative settings by the ordering ED provider.

Study Design and Measurements

This was a retrospective, observational cohort study of consecutive adult ED patients (18 years or older) from January 2012 to June 2015, generated through the hospital bed management system (Cerner ADT), with cross-validation with the hospital billing system via EPIC EHR. Patients included were those intubated by emergency medical services in the field or by EM providers with ongoing mechanical ventilation in the ED and eventual admission to the MICU. Patients intubated on the inpatient service; those intubated for a procedure or in the operating room; those on chronic mechanical ventilation (e.g., via tracheostomy); and those who expired in the ED prior to ICU admission were excluded from the study. Patients without a recorded height as well as those on a non-volume control mode of ventilation (e.g., pressure control ventilation) were also excluded. Manual chart review of the institution’s EPIC EHR was employed to validate cohort inclusion and confirm absence of exclusion criteria.

Patient data were captured from the electronic health records (EHR), including clinical presentation, reason for intubation, and ED course (initial and final ED vitals; ED laboratory values; and initial and final ED ventilator variables). Baseline patient characteristics including age, gender, race, height, weight, code status, comorbidities, and primary reason for intubation were manually abstracted from the medical chart. Clinical time stamps were collected including the following: time of hospital presentation, time of ED provider assignment, intubation time, admit time, time transported out of ED, and ICU admit time. All chart abstraction was performed in a standardized fashion, using iteratively trained reviewers (CO, JLK) with overlapping review of the entire cohort, a standardized case report form with specified protocols for data types and EHR locations, secondary validation by a third reviewer (KSM) of a five percent sample of the abstracted fields, and adjudication of conflicting review in the data collection phase (KSM). Flowsheet data (auto-populated in the EHR through device integration), MD and RN documentation, and RT charting were utilized as data sources. Mechanical ventilation orders entered by ED and ICU providers and mechanical ventilator settings found in the EPIC EHR via both manual entry by staff respiratory therapists and through device integration (continuous data streaming), as well as patient throughput data for the hospitalization, were collected via structured electronic query of the EPIC EHR using the institution’s Data Warehouse. Electronically captured data was validated with manual chart abstraction.

Severity of illness was calculated for all patients using the Mortality Probability Model III on admission (MPM₀-III), which predicts ICU mortality using 16 variables obtained within 1 hour of ICU admission.²⁴ In contrast other SOI tools, MPM₀-III includes variables typically available and/or ordered during the ED stay and readily available in the EHR. Risk for ARDS was captured with calculation of the Emergency Department Lung Injury Prevention Score (EDLIPS).¹⁴ The various clinical and laboratory data required for EDLIPS (shock, aspiration, sepsis, pneumonia, high-risk surgery, high-risk trauma, alcohol use, BMI, albumin, chemotherapy, FiO₂ respiratory rate, SpO₂, pH, and diabetes) were manually abstracted from the EHR. Diagnoses and conditions as specified in MPM₀-III, EDLIPS, as well as confirmed diagnoses of ARDS were captured from provider notes in the ED as well as progress and admission notes from the ICU attending physician, reflecting patients for whom ARDS was quickly recognized by the treating team or eventually worked up as an inpatient.

Hospital course data including ED length of stay, ED boarding time, ICU length of stay, and hospital length of stay were calculated from electronically collected time stamp and disposition data. ED length of stay is the time from hospital presentation to time transported out of ED; boarding time includes time from admission to time transported out of ED. Using process metrics and subcycle intervals,²⁵ total ED census data were obtained per 5-minute intervals, with calculation of the volume of patients being actively managed by the ED team (currently undergoing initial diagnostic workup) and volume of patients who end up requiring an inpatient admission. Metrics were averaged over the patients’ ED length-of-stay.

This project was approved by the institutional review board at the study site under expedited review procedure, with a waiver of informed consent.

Statistical Methods

The primary outcome was the final ED mechanical ventilator tidal volume settings of LPV (≤8cc/kg) or non-LPV (>8cc/kg). Univariate statistics were conducted to characterize the central tendencies and frequency distributions. T-testing, Chi-square testing, analysis of variance (ANOVA), and/or non-parametric testing was used to test for differences between baseline characteristics and final ventilator settings, as appropriate. Bivariate analysis using simple logistic regression was performed to determine crude associations between the outcome, predictors of interest and covariates. EDLIPS was treated as a categorical variable (< 5 or ≥ 5) in accordance with previous research defining optimal test characteristics for ARDS risk.¹⁴

Multivariable logistic regression was utilized to determine the odds of being placed on LPV by the time of ED departure by patient- and hospital-related characteristics, with particular attention to predictors related to boarding, patient volume, and clinical risk factors (i.e. EDLIPS variables) for ARDS. We reviewed EDLIPS variables as candidate standalone predictors in our model.^24,26 Our model also examined patient variables of age, gender, race and BMI in the probability of LPV by ED departure (Table 3). Possible interactions between patient volume and ARDS risk factors, severity of illness, and reason for intubation were also tested, as patients with higher clinical acuity or specific diagnoses may affect how providers both set and adjust the ventilator during periods of ED crowding.

TABLE 3.

Multivariate regression model examining probability of LPV utilization by ED departure

Predictors	Model* (n=444) Odds Ratio (95%CI)	p-value
Age, year	0.974 (0.959–0.989)	<0.001
Gender	10.929 (6.582–18.147)	<0.001
BMI	0.944 (0.917–0.972)	<0.001
Reason for intubation (ref: primary hypoxemia)
Primary hypercapnia	1.786 (0.850–3.749)	0.126
Airway protection	1.276 (0.663–2.456)	0.465
Cardiac arrest, shock, sepsis	1.664 (0.768–3.601)	0.197
Other, unclear	2.247 (0.897–5.628)	0.084
Mechanical ventilation duration in ED	0.942 (0.882–1.005)	0.072
ED Ventilator adjustments (non-TV related)	2.07 (1.098–3.905)	0.025
Percentage of active ED patients (workup in-progress)	0.968 (0.939–0.997)	0.033
Percentage of higher acuity patients (eventual inpatient admission)	0.967 (0.935–1.000)	0.049

^*Model adjusted for race/ethnicity

AIC = 482.573; Hosmer-Lemeshow goodness of fit = 0.794

Abbreviations: 95%CI = 95% confidence interval; AIC = Akaike Information Criterion; BMI = body mass index; ED = emergency department; MPM₀-III = Mortality Probability Model III score on admission; TV = tidal volume.

All analyses were conducted using SAS 9.3 software (SAS Institute).

RESULTS

Cohort Characteristics

During the study period, 1,117 patients admitted from the ED, 44.3% (n=495) of whom required mechanical ventilation during their ED stay. Of these patient encounters, 446 met inclusion and exclusion criteria. (See Appendix Figure 1.) Patient characteristics, stratified by use of LPV, are detailed in Table 1. Mean Mortality Probability Model III score at ICU admission (MPM₀-III) was 32.2% ± 22.4%, with actual in-hospital mortality or hospice discharge in 34.8% (n=155) of patients. High risk for ARDS (EDLIPS ≥ 5) was seen in 373 (82%) patients, with no significant differences between groups placed on LPV or non-LPV settings. Thirty-four patients (7.6%) were recognized as having ARDS by ICU admission.

TABLE 1.

Baseline cohort characteristics

Characteristics	Total N = 446	LPV by ED Departure N = 256	Non-LPV by ED Departure N = 190
Patient characteristics
Age, years (mean, SD)**	62.2 (16.0)	59.9 (15.8)	65.3 (15.8)
Female Gender (n, %)**	229 (51.3)	80 (31.3)	149 (78.4)
Race/ethnicity, (n, %)
White, Non-Hispanic	108 (23.7)	60 (23.4)	47 (24.7)
Black, Non-Hispanic	149 (33.3)	94 (36.7)	55 (28.9)
Hispanic/Latino	168 (37.5)	86 (33.6)	82 (43.2)
Other/unknown	22 (4.9)	16 (6.3)	6 (3.2)
Body mass index (median, IQR)**	26.4 (22.5, 31.3)	25.3 (21.7, 29.7)	28.2 (24.2, 33.4)
Mortality Probability Model III score on admission (mean, SD)*	32.2 (22.4)	29.8 (21.2)	35.4 (23.6)
Reason for intubation, n (%)
Primary Hypoxemia	93 (20.9)	55 (21.5)	38 (20.0)
Primary Hypercapnia	93 (20.9)	53 (20.7)	40 (21.1)
Airway protection	138 (30.9)	77 (30.1)	61 (32.1)
Cardiac arrest, shock, sepsis	72 (16.1)	43 (16.8)	29 (15.3)
Other/unclear	50 (11.2)	28 (10.9)	22 (11.6)
ED Lung Injury Prevention Score (LIPS) (mean, SD)	7.2 (2.6)	7.2 (2.6)	7.2 (2.6)
ED LIPS ≥ 5 (n, %)	336 (82.1)	208 (81.3)	158 (83.2)
ARDS diagnosed on admission, n (%)	34 (7.6)	17 (6.6)	17 (8.9)
Hospital course
ED length of stay, hrs (median, IQR)	7.6 (5.3. 11.0)	7.8 (5.4, 11.1)	8.8 (5.2, 10.9)
ED Boarding time, hrs (median, IQR)	3.4 (1.8, 6.2)	3.7 (1.6, 6.6)	3.3 (1.9, 6.1)
ICU length of stay, days (median, IQR)	5.5 (2.8, 9.6)	5.2 (2.6, 9.8)	6.0 (3.0, 9.6)
Hospital length of stay, days (median, IQR)	9.6 (5.4, 19.4)	8.9 (5.2, 20.1)	9.8 (5.9, 19.1)
In-hospital mortality or hospice discharge (n, %)	155 (34.8)	85 (33.2)	70 (36.8)
Average ED volume metrics during patient visit
Number of ED patients (mean, SD)	73.8 (22.9)	74.1 (22.7)	73.4 (23.3)
Active patients (workup in-progress), as a percentage of total census (mean %, SD)	66.2 (9.6)	65.8 (9.6)	66.7 (9.7)
Patients with eventual inpatient admission, as a percentage of total census (mean %, SD)	46.8 (8.4)	46.6 (8.3)	47.0 (8.5)

^*p<0.01;

^**p<0.001

Abbreviations: ARDS = acute respiratory distress syndrome; ED = Emergency Department; hrs = hours; IQR = interquartile range; LPV = lung protective ventilation; n = number; SD = standard deviation

Ventilator Management in the ED

LPV was utilized for final ED ventilator settings prior to departure for ICU in 57.4% of the final cohort available for analysis. LPV was less commonly utilized in female patients (31.3% vs. male 68.8%, p<0.001) and those with higher BMIs (p<0.001). Other clinical characteristics were similar between groups. ED and hospital lengths-of-stay were also similar for the LPV and non-LPV groups. While boarding time did not differ between groups, median ED boarding time was 3.4 hours (IQR 1.8, 6.2), with ranges of ICU wait time between 0.0 and 29.8 hrs.

Two-thirds of patients departed the ED on tidal volume settings of 450 ml (n=136, 30.5%) or 500 ml (n=161, 36.1%)—the latter setting being the default value for tidal volume setting on the mechanical ventilation order within the institution’s electronic health record system. Overall, less than one-third (n=134) of the patient cohort had any ventilator adjustments during the ED stay, with the majority of patients remaining on similar settings on ICU admission (Figure 1). Of these patients with any ventilator adjustment, TV changes were made for only 47 (35.1%) patients, of which only half (n=24) were to LPV settings (See Table 2).

FIGURE 1. Tidal volume from Initial ED settings, ED departure, and ICU Arrival.

The range of tidal volume settings, in cc/kg predicted body weight, in the ED and on ICU arrival are shown in this figure. Patients remain on similar settings throughout their ED stay and after ICU admission, with only 57% placed on LPV strategy.

TABLE 2.

Ventilator duration & management in the ED and on ICU arrival

Characteristics	Total N = 446	LPV by ED Departure N = 256	Non-LPV by ED Departure N = 190
Ventilator duration
Total in-hospital mechanical ventilation duration, days (median, IQR)	3.8 (1.6, 9.1)	3.3 (1.4, 7.9)	4.2 (1.7, 9.5)
ED mechanical ventilation duration, hrs (median, IQR)	3.7 (2.3. 5.6)	3.7 (2.3, 5.8)	3.7 (2.1, 6.1)
Ventilator dependence on discharge (n, %)	27 (6.1)	16 (6.3)	11 (5.8)
Ventilator management
Initial ED TV settings , cc/kg PBW (median, IQR)	7.8 (6.8, 8.8)	7.1 (6.4, 7.6)	9.0 (8.4, 10.0)
Final ED TV settings, cc/kg PBW (median, IQR)	7.7 (6.8, 8.8)	7.0 (6.4, 7.6)	9.0 (8.4, 10.0)
TV settings on ICU arrival, cc/kg PBW (median, IQR)	7.8 (6.9, 8.8)	7.1 (6.4, 7.6)	9.0 (8.4, 10.0)
Any ventilator change in ED, n (%)	134 (30.0)	86 (33.6)	48 (25.3)
TV change in ED, n (%)	47 (10.5)	24 (9.4)	23 (12.1)
Non-TV change in ED, n (%)	124 (27.8)	81 (31.6)	43 (22.6)
Any ventilator setting change on ICU arrival (n, %)	304 (68.2)	181 (70.7)	123 (64.7)
TV change in ICU, n (%)	123 (27.6)	78 (30.5)	45 (23.7)
Non-tidal volume change in ICU, n (%)	288 (64.6)	171 (66.8)	117 (61.6)

Abbreviations: cc/kg PBW= cubic centimeters per kilogram predicted body weight; ED = Emergency Department; ICU = Intensive Care Unit; IQR = interquartile range; LPV = lung protective ventilation; n = number; TV = tidal volume.

Crowding and LPV Utilization

In the multivariate model controlling for age, gender, and body mass index, higher percentages of active ED patients and those with higher acuity (eventual inpatient admission) were associated with lower odds of LPV utilization by ED departure (OR, 0.968, 95% CI [0.939–0.997], p=0.03; OR 0.967 [0.935–1.000, p=0.049], respectively) (See Table 3 and Appendix Figure 2). Higher severity of illness as measured by the MPM₀-III score, ARDS risk as measured by EDLIPS, and reason for intubation did not contribute to the likelihood of LPV by ED departure, nor was there any evidence of effect modification in times of high volume (data not shown).

For patients initially not on LPV, having active ventilator management with some type of adjustments, other than to tidal volume settings, was associated with high proportions of patients on LPV by ED departure (11.9 vs. 2.0%, p=0.014). In the adjusted model, ventilator adjustments to non-TV settings were associated with improved odds of LPV adherence (OR 2.07, [1.098–3.905], p=0.03), even with increased ED patient volume (See Figure 2).

FIGURE 2. Probability of LPV settings at ED departure, with increasing ED patient volume, stratified by presence or absence of documented ventilator adjustments.

Though increased ED patient volume is associated with decreasing odds of LPV settings, documented ventilator adjustments, other than TV setting, mitigates this decline.

DISCUSSION

Our study addresses the expanding demands for critical care in the ED by analysis of care delivery in a large, urban, tertiary care academic center. Our observational analysis provides new data on the optimization of critical care in the ED in the era of multidisciplinary critical care and an ICU without boundaries. Mechanical ventilation in the ED and early ICU course has several important factors which transcend traditionally investigated patient characteristics like BMI and gender in critically ill populations. Our study has brought to light the process metrics and clinician behaviors that drive LPV adherence in emergency department patients prior to ICU admission.

The findings of our study support existing data reflecting the infrequency of lung protective ventilation in the ED even for those patients with high severity of illness or at risk for ARDS. In addition to finding similar effects of female gender and high BMI on LPV as prior studies, we have added to the ARDS and critical care literature by performing a novel analysis of systems, patient, and provider factors affecting implementation of LPV in the unique ED environment. Many other studies have previously investigated crowding, ED boarding and the ED’s use of LPV.²⁷ We have taken a different approach than previous literature which has described ED strain in terms of annual census or shift-level crowding. Our novel approach to the analysis of ED strain on critical care services focuses on the cognitive workload influencing bedside management by critical care physicians or respiratory therapists.²² We chose the incorporation of MPM₀-III and EDLIPS because these populations would be expected to receive higher levels of care. Importantly, our findings are counterintuitive as LPV adherence by ED departure was not increased despite patients having increased risk for ARDS or increased illness severity.

As unique measures of mental bandwidth, the percentages of active ED patients requiring high clinician attention during initial diagnostic workup and the fraction of higher acuity patients (defined as those requiring eventual inpatient admissions) serve as truer measures of cognitive workload. These metrics of ED crowding were averaged over an individual patient’s entire ED stay to encapsulate the dynamic effects of crowding expanding upon previously identified measures for overall ED crowding.^25,28 The increased requirement for decision making in these patient populations can interfere with the clinician’s ability to perform multiple subsequent tasks like adjusting the ventilator.²⁹ While pre-existing literature has investigated the under recognition of ARDS by critical care clinicians, our aim was to investigate the relationship between ED crowding, cognitive burden and risk for ARDS as they influence critical care providers in the ED-ICU interface.

Electronically capturing sequential 5-minute data points, we are able to more accurately account for the cognitive burden placed upon critical care providers in the ED than prior literature. To our knowledge, we are the first to show how our unique measures of cognitive workload and ED crowding contribute to the challenge of using LPV in the ED. The pre-existing literature on ED crowding has been variable in defining high strain or high capacity (e.g., high occupancy rate, waiting time, or total ED volume).^19,30 In our multivariate model, higher concentrations of actively worked up patients and patients with eventual admission reflect a higher task burden which creates a significant barrier to critical care delivery. Because we used novel crowding metrics, we treated them as continuous variables. Our study’s dynamic measures of ED crowding and ED staff workload are more clinically meaningful metrics than static census counts.

Part of our hypothesis was correct in that increased ED crowding correlated with decreased likelihood of LPV by ED departure. Interestingly, our hypothesis positing that LPV adherence would improve with higher severity of illness or higher risk for ARDS was incorrect. This finding was counterintuitive, as we expected patients with more prominent phenotypes for critical illness or ARDS risk would receive more diligent attention in the ED’s dedicated resuscitation area. In our multivariate model, there was no increased LPV adherence for critically ill patients who appeared to benefit the most from diligent ARDS care. The likelihood of departing the ED on LPV settings was not affected by severity of illness (i.e. MPM₀_III) nor the risk for ARDS (i.e. EDLIPS ≥ 5)—both of which unfortunately portend worse outcomes when LPV is not utilized.³¹ Similar to previous literature on mechanical ventilation in the ED, LPV was only implemented in just over half the cohort (57.4%).⁶ Interestingly, there were no differences in ventilator practices for patients with primary hypoxemic respiratory failure as their reason for intubation, despite known benefit of LPV in these populations for risk mitigation for ARDS, pneumonia, and mortality.^32–34 Despite these patients having conditions warranting increased clinical attention, it is possible that physicians focused on other aspects of their management. More protocolized or third party support could ensure that high risk patients receive diligent attention despite varying clinician tendencies.

Though crowding is often beyond the control of the ED-ICU physician, ventilator setting adjustment was infrequent in our study cohort similar to previous studies on critical care delivery in the ED.^27,35 The majority of patients remained on the same ventilator settings from ED departure to ICU arrival. When adjustments were made to tidal volume settings, only half of these adjustments were to an LPV-compliant setting. The large fraction (66.6%) of the cohort with tidal volume settings that were either 450 ml or 500 ml (the default electronic health record setting) suggests that personalizing the ventilator settings to patient’s ideal body weight (i.e. height) specific for their gender during the ED stay rarely occurred. Interestingly, we found that ventilator adjustments to other settings (i.e. non-tidal volume settings), like respiratory rate or positive end expiratory pressure, were associated with increased likelihood of having LPV settings by ED departure. The correlation of non-tidal volume adjustments with increased LPV utilization suggests physicians or RTs who are comfortable managing a ventilator (i.e. those comfortable changing other variables beyond tidal volume) are more adept at providing critical care and implementing LPV in the ED-ICU interface. This correlative finding, while hypothesis-generating in our work, creates avenue for further study in optimizing critical care delivery for ED patients en route to ICU. Increasing the numbers of critical care-trained physicians in the ED-ICU interface and creating ventilator bundles could elevate critical care delivery regardless of patient location.³⁶ This further adds to the discussion for multidisciplinary critical care to cross geographic boundaries and in-hospital cultural barriers to provide intensive care to patients throughout their hospital course. More importantly, it provides an opportunity to make targeted interventions for high risk, critically ill patients in a high volume, high acuity clinical environment.

Prior work showed ED crowding to have a differential impact on various critical care interventions in septic, traumatic, and cardiac ED patient cohorts.^19,37,38 This work is a timely addition to the emerging literature on care delivery in the ED-ICU interface as lengths of stay and critically ill patient presentations continue to increase in our nation’s EDs.¹⁵ Emergency medicine and critical care medicine continue to evolve as hospitals create ED-ICU care delivery areas; however, there is ongoing debate on the integration of the two within multidisciplinary critical care.³⁹ Critical care delivery—specifically, mechanical ventilation practices—in the emergency department is disproportionately affected by boarding and crowding leading to substandard care for vulnerable populations.^15,40

Critical care delivery models to target vulnerable populations is an area for further study. Emergency medicine resident and attending physicians have reported mechanical ventilation education is often sparse and disproportionate to their exposure to critically ill patients in the ED.^21,41 The under-recognition of the importance of LPV in patients with higher severity of illness, primary hypoxemic respiratory failure, and higher risk for ARDS could be improved with increased specialized training for critically ill ventilated patients. However, our results support that overreliance on default settings, lower prioritization of ventilator management in times of high patient volume, and an absence of protocolized ventilator adjustment may contribute to the lower adherence to LPV in the ED. Best practice alerts, electronic health record user interface changes, and targeted deployment of hospital resources are all avenues by which LPV adherence rates, and thereby outcomes for high risk patients, can be improved.^9,42

Our study is limited in its generalizability as it is a single center study. Examination of ventilator practice within this large academic tertiary care center with faculty assigned to care for critically ill patients cohorted to one area of the ED, affords an opportunity to identify both provider practice and system-level targets for intervention to improve LPV adherence. A common potential source of bias in the electronic data capture is overreliance on an automated process like electronic data capture, leading to misclassification bias, though we attempted to mitigate this with cross-validation with physician and RT documentation. We are also limited in that initial ventilator settings and each subsequent adjustment cannot be accurately assigned to a provider, whether it be the physician, nurse, or RT. While our institution maintains the ED provider as the primary team for any critically ill patients still physically in the emergency department, it is also unclear if ventilator adjustments in our cohort were at the direction of the ED provider, as opposed to the ICU team in a consultative role. Our findings still support the need for protocolized approaches to ventilator management, as significant variability exists for the primary provider responsible for ventilator management in the ED.²¹ Finally, the diagnostic variability and under recognition of ARDS has been well-documented in the literature.⁷ With the understanding that these limitations have affected even large scale data on ARDS, we attempted to maintain validity by using ICU attending notes as the standard for ARDS in our cohort. In effort to include all those at risk for ARDS, we guided our inclusion criteria by ED and ICU attending notes. While we do acknowledge that our study does address the under-recognition of ARDS, our study goal was to describe systems issues precluding prompt ARDS care in the EDICU interface for patients at risk for ARDS. The under-recognition of ARDS by clinicians is well-described but does not change our conclusions on critical care delivery. Our main finding that LPV utilization is not improved in a crowded ED when patients are at increased risk of ARDS or more severely ill establishes an important role for further research on critical care delivery in the ED.

Ensuring accurate initial TV settings or increasing post-intubation adjustments to the ventilator during the ED stay can improve the odds of departing the ED with LPV. As the fields of emergency medicine and critical care continue to overlap and expand, both providers and hospital systems are pressured to establish how best to provide critical care interventions to those at risk in the ED-ICU interface.³⁹ Understanding the effect of ED crowding on the utilization of evidence-based critical care practices opens avenues for further study departmentally from the ED as well as the ICU. Our study identifies current practice patterns for providers in the ED-ICU interface and provides specific avenues for improvement in critical care delivery. Whereas preexisting data has shown shortcomings in LPV to be associated with patient-specific characteristics (e.g., patients of female gender or with higher BMI), our work incorporates system-level variables to identify how mechanical ventilation and critical care practices can be improved in transitions of care from the ED to the ICU.

• In this retrospective cohort study, we investigated the differential effect of various ED crowding metrics on lung-protective ventilation (LPV) practices for patients at risk for ARDS using multivariate regression modeling

• LPV adherence decreased as the ED became increasingly crowded

• Despite stratifying by illness severity or by risk for ARDS, ED ventilation practices were shown to often fall short of standard LPV settings

• Risk for non-LPV settings was mitigated by any non-tidal volume setting change prior to ED departure, suggesting that increased attention to the ventilator is associated with implementation of known LPV standards

• We investigated how mechanical ventilation practices in the ED-ICU interface are affected by measures of cognitive workload

• This opens further discussion into systems-level changes that may introduce high-yield interventions for at-risk critically ill populations