Association of an Emergency Department–embedded Critical Care Unit with Hospital Outcomes and Intensive Care Unit Use

Abstract

Rationale: A small but growing number of hospitals are experimenting with emergency department–embedded critical care units (CCUs) in an effort to improve the quality of care for critically ill patients with sepsis and acute respiratory failure (ARF).

Objectives: To evaluate the potential impact of an emergency department–embedded CCU at the Hospital of the University of Pennsylvania among patients with sepsis and ARF admitted from the emergency department to a medical ward or intensive care unit (ICU) from January 2016 to December 2017.

Methods: The exposure was eligibility for admission to the emergency department–embedded CCU, which was defined as meeting a clinical definition for sepsis or ARF and admission to the emergency department during the intervention period on a weekday. The primary outcome was hospital length of stay (LOS); secondary outcomes included total emergency department plus ICU LOS, hospital survival, direct admission to the ICU, and unplanned ICU admission. Primary interrupted time series analyses were performed using ordinary least squares regression comparing monthly means. Secondary retrospective cohort and before–after analyses used multivariable Cox proportional hazard and logistic regression.

Results: In the baseline and intervention periods, 3,897 patients met the inclusion criteria for sepsis and 1,865 patients met the criteria for ARF. Among patients admitted with sepsis, opening of the emergency department–embedded CCU was not associated with hospital LOS (β = −1.82 d; 95% confidence interval [CI], −4.50 to 0.87; P = 0.17 for the first month after emergency department–embedded CCU opening compared with baseline; β = −0.26 d; 95% CI, −0.58 to 0.06; P = 0.10 for subsequent months). Among patients admitted with ARF, the emergency department–embedded CCU was not associated with a significant change in hospital LOS for the first month after emergency department–embedded CCU opening (β = −3.25 d; 95% CI, −7.86 to 1.36; P = 0.15) but was associated with a 0.64 d/mo shorter hospital LOS for subsequent months (β = −0.64 d; 95% CI, −1.12 to −0.17; P = 0.01). This result persisted among higher acuity patients requiring ventilatory support but was not supported by alternative analytic approaches. Among patients admitted with sepsis who did not require mechanical ventilation or vasopressors in the emergency department, the emergency department–embedded CCU was associated with an initial 9.9% reduction in direct ICU admissions in the first month (β = −0.099; 95% CI, −0.153 to −0.044; P = 0.002), followed by a 1.1% per month increase back toward baseline in subsequent months (β = 0.011; 95% CI, 0.003–0.019; P = 0.009). This relationship was supported by alternative analytic approaches and was not seen in ARF. No associations with emergency department plus ICU LOS, hospital survival, or unplanned ICU admission were observed among patients with sepsis or ARF.

Conclusions: The emergency department–embedded CCU was not associated with clinical outcomes among patients admitted with sepsis or ARF. Among less sick patients with sepsis, the emergency department–embedded CCU was initially associated with reduced rates of direct ICU admission from the emergency department. Additional research is necessary to further evaluate the impact and utility of the emergency department–embedded CCU model.

Approximately 3 million patients with sepsis and/or acute respiratory failure (ARF) are admitted to U.S. hospitals via emergency departments annually (1–3). These patients are often admitted to intensive care units (ICUs), are at high risk for death and long hospital stays, and are among the costliest patients to the U.S. healthcare system (4–6). Despite these risks, however, care provided to patients with sepsis and ARF is heterogeneous and often deviates from evidence-based practice (7–15).

Emergency departments care for the majority of patients with sepsis and AR upon disease presentation, when management decisions are critical. They simultaneously face a burden of increased patient visits, heightened patient acuity, and strained ICU and ward bed capacity, which can lead to delays in ICU and ward admission and in the implementation of evidence-based practices. These admission delays are associated with adverse outcomes (10, 11, 13).

To address these challenges, the Hospital of the University of Pennsylvania established the Resuscitation and Critical Care Unit, a multidisciplinary emergency department–embedded critical care unit (CCU) similar to those at a small number of other U.S. hospitals (16, 17). The goals of this unit were to 1) improve clinical outcomes by optimizing the quality and timeliness of critical care delivery in the early hours of disease presentation and 2) reduce the volume of inpatient ICU admissions for patients who improve sufficiently with early critical care. We sought to evaluate, through a single-hospital, retrospective study, the association of an emergency department–embedded CCU with hospital outcomes and ICU use for patients admitted with sepsis and/or ARF.

Methods

Study Setting and Study Period

The study site is the Hospital of the University of Pennsylvania. The study emergency department–embedded CCU is a five-bed multidisciplinary unit staffed by emergency medicine (EM) physicians, EM-trained intensivists, EM house staff, emergency department– or ICU-trained nurses, respiratory therapists, and clinical pharmacists. Emergency department–embedded CCU physician and nursing staff were a separate team working in addition to, not pulled from, that staffing the main emergency department. Compared with a recent single-center study (18), the patient:physician ratio was lower and the patient:nurse ratio was similar at 2:1.

The study period spanned two calendar years (2016–2017) and included a baseline period from January 1, 2016, to November 14, 2016 (when rollout of the emergency department–embedded CCU began), followed by the intervention period from February 27, 2017, (when the emergency department–embedded CCU rollout period was complete) to December 31, 2017. The approximately 4-month stepwise rollout period was excluded from the analysis. In the intervention period, the emergency department–embedded CCU was open and continuously staffed on weekdays (i.e., Monday 7:00 a.m. to Saturday 7:00 a.m.) to support peak emergency department census hours but was closed on weekends (i.e., Saturday 7:00 a.m. to Monday 7:00 a.m.). In both the baseline and the intervention periods, emergency department patients were initially evaluated and managed by the main emergency department team. Throughout the baseline period and in the intervention period when the emergency department–embedded CCU was closed, the main emergency department team was responsible for all clinical care of critically ill patients until an inpatient ICU bed was available. In the intervention period when the emergency department–embedded CCU was open, patients requiring more than 1 hour of ongoing critical care could be rapidly transferred to the emergency department–embedded CCU for initial and ongoing management, even if inpatient ICU beds were available, based on a protocol and clinician judgment. Any patients requiring medical, surgical, or neurologic critical care were eligible for emergency department–embedded CCU transfer. The study hospital is a quaternary center that provides the full spectrum of care with the exception of trauma patients, who are directed when possible to a different nearby Penn Medicine hospital at which trauma care is consolidated.

Study Design

Because of the challenges in assessing the impact of an organizational intervention of this nature, we performed one set of primary analyses and three sets of secondary analyses (Figure 1). For our primary analyses, we performed interrupted time series analyses that compared outcomes trends among weekday admissions before and after the opening of the emergency department–embedded CCU. For our secondary analyses, we 1) performed retrospective cohort analyses during the intervention period between weekday emergency department presentations (when the emergency department–embedded CCU was open) and weekend emergency department presentations (when the emergency department–embedded CCU was closed), 2) performed before–after analyses comparing outcomes of weekday emergency department presentations between the baseline period (before the emergency department–embedded CCU was open) and the intervention period (when the emergency department–embedded CCU was open), and 3) evaluated difference-in-difference analyses comparing the baseline period–intervention period differences between weekday emergency department presentations (emergency department–embedded CCU not yet open during baseline period and emergency department–embedded CCU open during intervention period) and weekend emergency department presentations (emergency department–embedded CCU closed during both periods). Although the weekday–weekend comparison is limited by comparing potentially dissimilar weekday versus weekend patients, it may surmount secular trends, such as respiratory viral season, that could limit before–after approaches, and weekday–weekend effects may be less profound among higher acuity critically ill patients whose ICU admission decisions may be less discretionary.

Figure 1.

Study design and analytic plan. We evaluated 1) interrupted time series and before–after analyses comparing outcomes of weekday emergency department presentations between the baseline period (before the emergency department–embedded critical care unit [CCU] was open) and the intervention period (when the emergency department–embedded CCU was open) (A vs. B), 2) retrospective cohort analyses during the intervention period between weekday emergency department presentations (when the emergency department–embedded CCU was open) and weekend emergency department presentations (when the emergency department–embedded CCU was closed) (B vs. D), and 3) difference-in-difference analyses in which we compared the baseline period–intervention period differences between weekday emergency department presentations (emergency department–embedded CCU not yet open during baseline period and open during intervention period) and weekend emergency department presentations (emergency department–embedded CCU not yet open during baseline period and closed during intervention period) ([A vs. B] vs. [C vs. D]).

Study Population

Patients were eligible for inclusion if they were adult patients (18 yr or older) admitted from the emergency department to a medical ward, step-down unit, or ICU at the study hospital during the baseline or intervention period and met clinical diagnostic criteria for sepsis and/or ARF based on physiologic data collected in the emergency department.

Patients with sepsis and ARF were identified using the clinical criteria used in previous studies that included the study hospital (19). Sepsis was defined as suspected or confirmed infection (requiring at least one antibiotic order and microbiologic culture order) meeting at least one end-organ dysfunction physiologic criterion in the emergency department, anchored on the Sepsis-3 consensus definition of Sepsis-related Organ Failure Assessment (SOFA) score of 2 or greater (20, 21). Because of limitations in the availability and reliability of all SOFA data elements in the emergency department (for example, partial pressure of oxygen (Pa_O2):fraction of inspired oxygen [Fi_O2] [P:F] ratio) as well as recent concerns that this definition excludes many infected patients at risk for poor outcomes (22–24), we expanded the criteria to include patients who did not meet SOFA criteria but had a quick SOFA (qSOFA) score of 2 or greater (20, 21, 25, 26), serum lactate of 4 mmol/L or greater, a single oxygen saturation as measured by pulse oximetry (Sp_O2) of 85% or less while requiring any supplemental oxygen, receipt of Fi_O2 of 60% or greater or via a nonrebreather mask for at least two measurements at least 2 hours apart, or receipt of any noninvasive ventilation or invasive mechanical ventilation.

ARF was defined as meeting any one of the following at any point in the emergency department: 1) a single Sp_O2 of 85% or less while requiring any supplemental oxygen; 2) receipt of supplemental oxygen of 6 L per minute or greater or Fi_O2 of 40% or greater for at least two measurements at least 2 hours apart; 3) Pa_CO2 of greater than 45 mm Hg or mixed venous CO₂ partial pressure (Pv_CO2) of greater than 50 mm Hg and respiratory rate of 22 breaths per minute or greater; 4) Pa_CO2 of greater than 60 mm Hg or Pv_CO2 of greater than 65 mm Hg and pH of 7.3 or less on a single blood gas; or 5) receipt of any noninvasive ventilation (including noninvasive bilevel positive airway pressure and continuous positive airway pressure) or invasive mechanical ventilation.

Patients were excluded if they were admitted to an exclusively nonmedical inpatient location (e.g., a surgical-only location). Patients with a do-not-resuscitate/do-not-intubate status in the emergency department were included because they remain eligible to otherwise receive the full range of critical care interventions, including treatment in the emergency department–embedded CCU and admission to the ICU, but patients with higher levels of restriction on life-sustaining therapy, such as do-not-escalate status, comfort measures only, or hospice status, were excluded.

Exposure

The exposure (independent variable) was eligibility for admission to the emergency department–embedded CCU, which was defined as admission to the emergency department during the intervention period on a weekday (i.e., between Monday at 7:00 a.m. and Saturday at 7:00 a.m.) when the emergency department–embedded CCU was open. To avoid selection bias because of emergency department–embedded CCU-admitted patients being of greater acuity and complexity than non–emergency department–embedded CCU-admitted patients (especially when the emergency department–embedded CCU was operating at full capacity), patients who met inclusion criteria when the emergency department–embedded CCU was open were categorized as exposed regardless of whether they were actually admitted to the emergency department–embedded CCU or received care in the normal emergency department workflow. The small number of patients (n = 23) who presented to the emergency department when the emergency department–embedded CCU was closed and who were later transferred to the emergency department–embedded CCU were considered to be unexposed.

Outcomes

The primary outcome was hospital length of stay (LOS), which was defined as the continuous time in days from emergency department registration to hospital discharge. Secondary outcomes included total emergency department plus ICU LOS (emergency department + ICU LOS; the sum of emergency department LOS and total time spent in the ICU between hospital admission and discharge in hours), hospital survival (with hospice status considered an in-hospital death), direct ICU admission (defined as admission from the emergency department directly to an ICU), and unplanned ICU admission (defined as transfer to the ICU within 48 h after initial admission to the ward).

Hospital LOS was selected as the primary outcome because reducing LOS is important for all stakeholders, represents a critical operational and financial metric for hospitals, and enables improved access to care for patients (27, 28). Reducing LOS is also commonly considered to be a patient- and family-centered outcome because of these groups’ clear interests in leaving the hospital quickly to reduce the stresses, discomforts, and risks of complications that accrue with acute inpatient care and to return to their lives at home in a timely manner (27). The emergency department + ICU LOS outcome was selected for the following two reasons: 1) to be able to evaluate whether the emergency department–embedded CCU reduced overall ICU LOS (a desired outcome) versus simply changing the location of critical care delivery (a less beneficial outcome) and 2) to be able to accurately assess LOS in the situation in which the emergency department LOS was intentionally prolonged in the emergency department–embedded CCU to enable patients believed to need some critical care to ultimately be triaged to wards. The analysis of direct ICU admission excluded patients who received invasive mechanical ventilation or vasopressors in the emergency department because these patients were believed to have a near-absolute indication for ICU admission and were unlikely to be downgraded to a lower level of care before hospital admission. Step-down unit admissions were considered to be ward admissions because the step-down unit at the study hospital is a specialized cardiac unit that admits patients requiring inotropic support but does not routinely admit patients from the emergency department requiring vasopressors or ventilatory support, and the classification of this step-down unit as a ward has not influenced results in prior studies among a similar patient population at the study hospital (26).

Adjustment Variables

Interrupted time series analyses, which cannot adjust for covariates on a patient level, were adjusted for monthly means of the Laboratory-based Acute Physiology Score version 2 (LAPS2) and Comorbidity Point Score version 2 (COPS2) (29, 30). Weekday–weekend cohort and before–after models were adjusted for a priori patient-level covariates including age, sex, race, LAPS2, and COPS2. Ethnicity was not included in the models because there was greater than 97% non-Hispanic ethnicity in the study population. LAPS2 scores include 24 variable inputs, including vital signs, neurologic assessments, laboratory values, and basic demographics and were calculated based on all data collected during the emergency department stay. COPS2 is a comorbidity burden score based on 12 months preceding International Classification of Diseases diagnosis coding.

Statistical Analysis

The primary interrupted time series analyses were performed using ordinary least squares regression with Newey-West standard errors (31). Outcomes were measured as monthly means of hospital LOS (continuous in days) and emergency department + ICU LOS (continuous in hours) and monthly means of probabilities of the binary outcomes (i.e., hospital survival, direct ICU admission, and unplanned ICU admission). Lag was set to 0 because we had already removed the 4-month rollout period. Interrupted time series results have the following two components: 1) the change compared with preintervention in the first postintervention time interval (e.g., a step or level change) and 2) the change compared with preintervention in all subsequent postintervention time intervals (e.g., a slope change).

In the secondary weekday–weekend cohort and before–after analyses, hospital LOS and emergency department + ICU LOS were analyzed using multivariable Cox proportional hazard models with death as a censoring event, in which an increased hazard ratio represents a shorter LOS (i.e., an increased hazard of discharge). The binary outcomes of hospital survival, direct ICU admission, and unplanned ICU admission were analyzed using multivariable logistic regression. For difference-in-difference analyses, the parallel trend assumption was assessed by plotting outcome measures by week in the baseline and intervention periods, and because of violations of this assumption (see Figures E1 and E2 in the online supplement), difference-in-differences analysis results are not reported. All analyses were stratified into sepsis and ARF groups, and patients who met both sets of diagnostic criteria were eligible for inclusion in both analyses. We considered multiple hospital admissions by the same patient during the study period to be unique patients (19, 26). We performed complete case analyses because data were missing in less than 3% of hospitalizations.

Sensitivity Analyses

Our choice to categorize as exposed patients who were eligible for but not admitted to the emergency department–embedded CCU (e.g., patients with lower acuity) when the emergency department–embedded CCU was open risks bias toward the null. To explore this and, relatedly, to investigate the hypothesis that improved clinical outcomes due to early critical care might differ by acuity, we repeated the hospital LOS, emergency department + ICU LOS, and hospital survival analyses stratifying patients with sepsis by requiring or not requiring vasopressors and stratifying patients with ARF by requiring or not requiring ventilatory support in the emergency department (defined as treatment with noninvasive ventilation or invasive mechanical ventilation). To evaluate for bias in the hospital LOS and emergency department + ICU LOS analyses because of informative censoring from in-hospital deaths, we repeated these analyses using a “placement of death” approach, in which deaths are assigned to an undesirable LOS (e.g., the 99th percentile LOS) (19, 32, 33).

P values of less than 0.05 were considered statistically significant, but all results are reported with 95% confidence intervals (CIs). Analyses were conducted using Stata (StataCorp LP). The study protocol was approved by the institutional review board of the University of Pennsylvania.

Results

Patient and Emergency Department–embedded CCU Characteristics

Emergency department patient volume during the baseline and intervention periods was similar; mean daily inpatient admissions were 35.1 (standard deviation [SD], 1.3) during the baseline period and 34.8 (SD, 1.4) during the intervention period. Overall, mean emergency department LOS in the baseline and intervention periods was 12.3 (SD, 12.4) and 12.4 (SD, 10.3) hours, respectively. During the intervention period, the emergency department–embedded CCU admitted 5.4 (SD, 0.9) new patients per 24 hours on average and had a mean LOS of 11.3 (SD, 9.4) hours. The emergency department–embedded CCU occupancy was 50.5% on average but had periods of high and low occupancy that tracked with the overall emergency department census. Median emergency department to emergency department–embedded CCU transfer time was 4.0 (interquartile range, 1.9–6.9) hours. A total of 8.7% of emergency department–embedded CCU patients were treated, observed, and discharged.

During the study period, 15,579 patients were admitted from the emergency department to a medical ward, step-down unit, or ICU at the study hospital and screened for inclusion. Among those, 3,897 patients met the inclusion criteria for sepsis (1,524 in the baseline period and 1,499 in the intervention period) and 1,865 met the criteria for ARF (787 in the baseline period and 634 in the intervention period). A total of 873 patients met inclusion criteria for both cohorts. Sixteen patients with sepsis and 18 patients with ARF were excluded because of restrictions on life-sustaining therapy beyond do-not-resuscitate/do-not-intubate status in the emergency department. Among patients eligible for emergency department–embedded CCU admission, 362 (24.2%) patients with sepsis and 251 (39.4%) patients with ARF were admitted to the emergency department–embedded CCU. Among patients who met the inclusion criteria, illness acuity was similar between the baseline and intervention periods; the mean LAPS2 score was 107.4 (SD, 38.6) versus 101.4 (SD, 40.5) for patients with sepsis and 108.5 (SD, 40.7) versus 106.7 (SD, 45.3) for patients with ARF, respectively. Table 1 reports patient characteristics by diagnostic group and study period.

Table 1.

Patient characteristics

Characteristics	Patients with Sepsis (n = 3,897)		Patients with ARF (n = 1,865)
	Baseline	Intervention	Baseline	Intervention
Patients, n (%)	1,952 (50.1)	1,945 (49.9)	1,025 (55.0)	840 (45.0)
Weekday admissions, n (%)	1,524 (78.1)	1,499 (77.1)	787 (76.8)	634 (75.5)
Admitted to emergency department–embedded CCU, n (%)	n/a	362 (24.2)	n/a	251 (39.4)
Age, yr, mean (SD)	59.3 (16.2)	60.1 (16.8)	61.3 (15.5)	61.5 (16.4)
Female, n (%)	865 (44.3)	903 (46.5)	482 (48.4)	403 (48.8)
Race, n (%)
White	926 (47.4)	954 (49.1)	367 (35.8)	317 (37.7)
Black	862 (44.2)	842 (43.3)	567 (55.3)	458 (54.5)
Other*	164 (8.4)	149 (7.7)	91 (8.9)	65 (7.7)
Non-Hispanic ethnicity, n (%)	1,874 (96.7)	1.876 (96.7)	964 (97.4)	804 (97.3)
SOFA score,† mean (SD)	2.8 (2.8)	3.3 (2.1)	2.2 (2.3)	2.8 (2.8)
LAPS2 score,‡ mean (SD)	107.4 (38.6)	101.4 (40.5)	108.5 (40.7)	106.7 (45.3)
COPS2 score,§ mean (SD)	138.6 (75.8)	147.5 (75.9)	133.7 (74.8)	145.9 (75.3)

Definition of abbreviations: ARF = acute respiratory failure; CCU =  critical care unit; COPS2 = Comorbidity Point Score version 2; LAPS2 = Laboratory-based Acute Physiology Score version 2; n/a = not available; SOFA = Sepsis-related Organ Failure Assessment; SD = standard deviation.

^*Includes Asian, Hawaiian/Pacific Islander, American Indian/Native American, self-reported race as multiple or other, or unknown.

^†Because of unreliability of vasopressor titration dosing in the emergency department, SOFA scores for patients receiving vasopressors are underestimates. The SOFA renal function subscore relied on serum creatinine alone because urine output is not routinely recorded in the emergency department.

^‡LAPS2 possible range 0–414 and univariate relationship with hospital mortality across pooled cohorts: 0–49, 0.4%; 50–99, 3.8%; ≥100, 19.3%.

^§COPS2 possible range 0–1,014 and univariate relationship with hospital mortality across pooled cohorts: 0–64, 0.5%; ≥65, 12.9%.

Outcomes Trends

Figures E1 and E2 display observed unadjusted trends in the outcome measures over the duration of the study period across the baseline and intervention periods for sepsis and ARF, respectively.

Association of the Emergency Department–embedded CCU with LOS and Hospital Survival

In the primary interrupted time series analysis, among patients with sepsis (Table 2), the opening of the emergency department–embedded CCU was not associated with hospital LOS (β = −1.82 d; 95% CI, −4.50 to 0.87; P = 0.17 for the first month after emergency department–embedded CCU opening compared with baseline and β = −0.26 d; 95% CI, −0.58 to 0.06; P = 0.10 for subsequent months after emergency department–embedded CCU opening compared with baseline) (Figure 2A), emergency department + ICU LOS (β = −24.25 h; 95% CI, −78.71 to 30.20; P = 0.36 for the first month and β = −4.53 h; 95% CI, −9.61 to 0.55; P = 0.08 for subsequent months) (Figure 2B), or probability of hospital survival (β = 0.030; 95% CI, −0.022 to 0.079; P = 0.25 for the first month; β = −0.006; 95% CI, −0.015 to 0.003; P = 0.17 for subsequent months) (Figure 2C).

Table 2.

Interrupted time series analyses of the association of emergency department–embedded CCU opening with hospital outcomes and ICU use among patients with sepsis and acute respiratory failure

	Estimates [β-Coefficient (95% CI, P Value)] for Change in Mean Outcome by Month
	Baseline Period	First Month after Emergency Department–embedded CCU Opening Compared with Baseline	Subsequent Months after Emergency Department–embedded CCU Opening Compared with Baseline
Sepsis
Hospital LOS, d (1° outcome)	0.17 (−0.10 to 0.44, 0.19)	−1.82 (−4.50 to 0.87, 0.17)	−0.26 (−0.58 to 0.06, 0.10)
Emergency department + ICU LOS, h	3.74 (−0.65 to 8.14, 0.09)	−24.25 (−78.71 to 30.20, 0.36)	−4.53 (−9.61 to 0.55, 0.08)
Hospital survival probability	−0.002 (−0.005 to 0.002, 0.34)	0.030 (−0.022 to 0.079, 0.25)	−0.006 (−0.015 to 0.003, 0.17)
Direct ICU admission probability*	−0.003 (−0.009 to 0.003, 0.30)	−0.099 (−0.153 to 0.044, 0.002)†	0.011 (0.003 to 0.019, 0.009)†
Unplanned ICU admission probability‡	−0.002 (−0.006 to 0.002, 0.38)	0.002 (−0.023 to 0.028, 0.85)	0.0001 (−0.004 to 0.004, 0.96)
Acute respiratory failure
Hospital LOS, d (1° outcome)	0.50 (0.01 to 0.98, 0.046)†	−3.25 (−7.86 to 1.36, 0.15)	−0.64 (−1.12 to 0.17, 0.01)†
Emergency department + ICU LOS, h	7.48 (−3.18 to 18.14, 0.16)	−17.65 (−124.06 to 88.75, 0.73)	−10.69 (−21.64 to 0.25, 0.06)
Hospital survival probability	0.002 (−0.007 to 0.010, 0.66)	0.007 (−0.061 to 0.075, 0.84)	−0.010 (−0.026 to 0.006, 0.22)
Direct ICU admission probability*	−0.005 (−0.016 to 0.005, 0.32)	0.046 (−0.049 to 0.141, 0.32)	0.006 (−0.017 to 0.029, 0.56)
Unplanned ICU admission probability‡	−0.001 (−0.010 to 0.007, 0.74)	−0.008 (−0.075 to 0.060, 0.82)	0.002 (−0.020 to 0.011, 0.92)

Definition of abbreviations: CI = confidence interval; CCU = critical care unit; ICU = intensive care unit; LOS = length of stay.

^*Among patients not requiring mechanical ventilation or vasopressors in the emergency department.

^†P < 0.05.

^‡Among patients initially admitted to the ward.

Figure 2.

Interrupted time series analyses of the potential impact of opening the emergency department–embedded critical care unit among patients with sepsis. No impact is seen for the (A) hospital length of stay (LOS), (B) emergency department + intensive care unit (ICU) LOS, (C) hospital survival, or (E) unplanned ICU admission outcomes. (D) The opening of the emergency department–embedded critical care unit was associated with a transient decrease and then subsequent increase back toward baseline in direct ICU admission. *P < 0.05 for first month and **P < 0.05 for subsequent months.

Among patients with ARF (Table 2), the opening of the emergency department–embedded CCU was not associated with hospital LOS for the first month after the emergency department–embedded CCU opening compared with baseline (β = −3.25 d; 95% CI, −7.86 to 1.36; P = 0.15; i.e., no level change) but was associated with a 0.64 d/mo shorter mean hospital LOS for subsequent months after emergency department–embedded CCU opening compared with baseline (β = −0.64 d; 95% CI, −1.12 to 0.17; P = 0.01; i.e., a slope change) (Figure 3A). In this ARF population, the opening of the emergency department–embedded CCU was not associated with emergency department + ICU LOS (β = −17.65 h; 95% CI, −124.06 to 88.75; P = 0.73 for the first month and β = −10.69 h; 95% CI, −21.64 to 0.25; P = 0.06 for subsequent months) (Figure 3B) or with probability of hospital survival (β = 0.007; 95% CI, −0.061 to 0.075; P = 0.84 for the first month and β = −0.010; 95% CI, −0.026 to 0.006; P = 0.22 for subsequent months) (Figure 3C).

Figure 3.

Interrupted time series analyses of the potential impact of opening the emergency department–embedded critical care unit among patients with acute respiratory failure. (A) Opening of the emergency department–embedded critical care unit was associated with a reduced hospital length of stay (LOS). No impact is seen for the (B) emergency department + intensive care unit (ICU) LOS, (C) hospital survival, (D) direct ICU admission, or (E) unplanned ICU admission outcomes. *P < 0.05 for first month and **P < 0.05 for subsequent months.

Among patients with sepsis or ARF, the availability of the emergency department–embedded CCU was not associated with hospital LOS, emergency department + ICU LOS, or probability of hospital survival in the secondary weekday–weekend cohort or before–after analyses (Tables E1 and E2).

Association of the Emergency Department–embedded CCU with Direct ICU Admission

Among patients admitted with sepsis who did not require mechanical ventilation or vasopressors in the emergency department, the opening of the emergency department–embedded CCU was associated with an initial 9.9% reduction in direct ICU admissions in the first month (β = −0.099; 95% CI, −0.153 to 0.044; P = 0.002) and then a 1.1% per month increase back toward baseline in subsequent months (β = 0.011; 95% CI, 0.003–0.019; P = 0.009) (Figure 2D and Table 2). A similar finding was observed in the secondary weekday–weekend cohort (odds ratio [OR], 0.40; 95% CI, 0.28–0.57; P < 0.001) and before–after (OR, 0.59; 95% CI, 0.45–0.78; P < 0.001) analyses (Table E1).

Among patients admitted with ARF who did not require mechanical ventilation or vasopressors in the emergency department, the opening of the emergency department–embedded CCU was not associated with the probability of direct ICU admission in the primary interrupted time series analysis (β = 0.046; 95% CI, −0.049 to 0.141; P = 0.32 in the first month and β = 0.006; 95% CI, −0.017 to 0.029; P = 0.56 in subsequent months) (Table 2 and Figure 3D) or in the secondary before–after analysis (OR, 1.16; 95% CI, 0.82–1.63; P = 0.42) but was associated with reduced odds of direct ICU admission in the weekday–weekend cohort analysis (OR, 0.46; 95% CI, 0.29–0.72; P = 0.001) (Table E3).

Association of the Emergency Department–embedded CCU with Unplanned ICU Admission

Among patients admitted with sepsis who were initially admitted to the ward, the opening of the emergency department–embedded CCU was not associated with the probability of unplanned ICU admission in the primary interrupted time series analysis (β = 0.002; 95% CI, −0.023 to 0.028; P = 0.85 in the first month and β = 0.0001; 95% CI, −0.004 to 0.004; P = 0.96 in subsequent months) (Table 2 and Figure 2E) or in the secondary weekday–weekend cohort and before–after analyses (Table E1).

Similarly, among patients admitted with ARF who were initially directly admitted to the ward, the opening of the emergency department–embedded CCU was not associated with the probability of unplanned ICU admission in the primary interrupted time series analysis (β = −0.008; 95% CI, −0.075 to 0.060; P = 0.82 in the first month and β = 0.002; 95% CI, −0.020 to 0.011; P = 0.92 in subsequent months) (Table 2 and Figure 3E) or in the secondary weekday–weekend cohort and before–after analyses (Table E2).

Sensitivity Analyses

During the baseline and intervention periods, 327 (7.2%) patients admitted with sepsis required vasopressors in the emergency department, and 976 (44.3%) patients admitted with ARF required ventilatory support in the emergency department. Patients with higher acuity (sepsis requiring vasopressors and ARF requiring ventilatory support) were admitted to the emergency department–embedded CCU at higher rates than patients not requiring vasopressor or ventilatory support in the emergency department (80.3% and 59.5%, respectively). The supplemental Appendix A, Figures E3 and E4 and Tables E3–E7 report results from the interrupted time series, weekday–weekend cohort, and before–after analyses stratified by acuity. Notably, in the interrupted time series sensitivity analyses stratified by acuity (Table E3), the hospital LOS results for the higher acuity strata (Figures E3B and E4B), in which greater proportions of patients were admitted to the emergency department–embedded CCU, were consistent with the primary unstratified results.

In interrupted time series analyses, in which in-hospital deaths were assigned to the 99th percentile hospital LOS (48.1 d for sepsis and 53.6 d for ARF) and emergency department + ICU LOS (501.7 h for sepsis and 712.9 h for ARF), the opening of the emergency department–embedded CCU was not associated with either LOS outcome (Table E8).

Discussion

Emergency department overcrowding and early adherence to evidence-based care for patients with or at risk for critical illness remain important and dynamic issues. Embedding critical care within the emergency department has face validity as a potential countermeasure. This organizational intervention, however, remains challenging to study, and our results, which sought for greater robustness with multiple analytic approaches, are by extension difficult to interpret.

Our primary interrupted time series finding of an association between the emergency department–embedded CCU opening and reduced LOS among patients with ARF, perhaps concentrated to those with high acuity requiring ventilatory support, was not replicated in other analytic approaches and may be susceptive to informative censoring bias because of in-hospital deaths, as the relationship appears null using a “placement of death” approach. Hospital LOS analyses were null in the sepsis cohort, and hospital survival analyses were null in both sepsis and ARF cohorts. In this single-hospital, retrospective study, the availability of an emergency department–embedded CCU was therefore not associated with improved clinical outcomes among patients admitted with sepsis or ARF. These findings are in conflict with a recent large single-center study suggesting improved mortality with an emergency department–embedded CCU among a heterogeneous all-comer emergency department patient population (18). This observation that emergency department–embedded CCU availability was not associated with improved clinical outcomes among patients with sepsis and ARF could be attributable to the misclassification of exposure status, an insufficient follow-up that did not allow observation of more mature emergency department–embedded CCU use and care patterns, or a lack of true benefit of the emergency department–embedded CCU, although such a declaration cannot be made from a single-center study. Overall, the emergency department–embedded CCU concept remains heterogeneous, and its effectiveness, if present, will be heavily reliant on the surrounding factors of the implementation of evidence-based practices, patient selection, handoffs, and other processes of care that span the spectrum from outpatient medicine to inpatient critical care.

Among patients admitted with sepsis who were less sick, the availability of the emergency department–embedded CCU was associated with the secondary outcome of reduced ICU use, at least initially. This finding was observed at another center (18), and similar findings were reported in a different single-center study, which found that an emergency department–embedded CCU was associated with increased ICU access and throughput for critically ill transfer patients (34). In our study, among patients admitted with sepsis who did not require mechanical ventilation or vasopressors in the emergency department, the availability of the emergency department–embedded CCU was associated with reduced direct ICU admission from the emergency department, although this association was transient in the interrupted time series analysis. It is unclear if the transient nature of this association was due to practice changes in the emergency department–embedded CCU or to adjustments made to ICU and hospital patient flow and disposition decisions in response. This overall result was replicated in the complementary weekday–weekend cohort and before–after analyses and corresponds with a potential association between the emergency department–embedded CCU and shorter emergency department + ICU LOS in the subgroup of patients with lower acuity sepsis who did not require vasopressors, although this result is present in only some secondary analyses. This could represent improved early critical care delivery that, although not changing overall survival, does avert some ICU admissions and may reduce overall time spent in an ICU. Patients who present with rapidly reversible conditions or those with highly predictable clinical trajectories, such as those due to diabetic ketoacidosis or toxin ingestions, may specifically benefit from emergency department-based critical care delivery models that avert inpatient ICU admissions (35, 36). Conversely, if only the location but not the total duration of critical care is modified, the emergency department–embedded CCU intervention would be far less, or not at all, beneficial, serving in that scenario simply as an expansion of overall ICU capacity.

Preventing unnecessary or avoidable ICU stays is an important, patient-centered, and potentially cost-saving outcome (37). The latter would need to be put in context with the cost of the intervention, which was not assessed in this study. Furthermore, the sustainability of such a model would depend on adequate cost-sharing between the hospital or health system, which stands to benefit from more cost-effective acute and critical care delivery, and the emergency department, which stands to increase spending on more intensive upfront care (e.g., lower patient-to-clinician ratios, physical critical care resources, etc.) that yields benefits only later in a hospital stay.

There were a number of organizational and clinical challenges faced when developing the emergency department–embedded CCU at the Hospital of the University of Pennsylvania. Emergency department nurses without significant prior ICU experience required cross-training, which was accomplished through a novel module-based learning curriculum that included physician faculty lectures; a 1-month multidisciplinary clinical rotation through neurological, surgical, and cardiac surgery ICUs; and a comprehensive nursing-directed orientation to the unit. Emergency department and emergency department–trained critical care physicians who provided clinical services in the emergency department–embedded CCU all participated in monthly quality improvement, morbidity and mortality, and educational conferences. Most critically for sustainability, despite thorough critical care billing and value-based initiatives with the goal to reduce the need for inpatient ICU services, financing longitudinal dedicated physician and nursing staff at ICU care ratios was challenging.

Strengths and Limitations

The primary strengths of this study include 1) the evaluation of a novel and innovative organizational intervention and 2) the deployment of multiple analytic approaches that took advantage of organizational features of the intervention to counteract typical analytic barriers to a study of this type.

The findings of this study should be interpreted in the context of a number of important limitations. First, despite our efforts to increase the rigor of the results by performing multiple corresponding study designs, the study overall still relies on nonoptimal retrospective approaches in examining this question. Perhaps most importantly, our primary analysis is biased toward the null because of classifying patients who were eligible for but not admitted to the emergency department–embedded CCU as exposed. This challenge was only partially addressed by sensitivity analyses showing similar results, which were restricted to higher acuity patients who were admitted to the emergency department–embedded CCU at higher rates, reducing but not eliminating misclassification. Second, the intensity of patient exposure to the emergency department–embedded CCU care delivery model was variable because of a number of factors not fully captured in our analyses (emergency department–embedded CCU occupancy fluctuated; patients were not managed for a fixed duration; patients with higher acuity were often transferred to ICUs more quickly than less acute, borderline patients; triage patterns differed near emergency department–embedded CCU opening and closing; and emergency department–embedded CCU attending staffing varied with time of day). Third, this was a single-hospital study, and results should not be generalized without additional supporting research; consideration of a resource-intensive emergency department–embedded CCU may not be feasible or practical for low-volume centers. Fourth, despite excluding a 4-month rollout period and including almost 1 year of intervention period, it is possible that the emergency department–embedded CCU had not yet matured to reach its optimal impact in this timeframe, and results might have differed with continued longitudinal observation. Fifth, sepsis and ARF are clinical syndromes with many diverse etiologies, among which there may be unmeasured heterogeneity of treatment effects. Finally, this study did not evaluate specific processes of care delivery (e.g., time to antibiotics or microbiologic cultures) or costs, the latter of which would be necessary for any cost effectiveness evaluations and required by hospital administrators considering similar organizational interventions.

Conclusions

In this single-hospital, retrospective study, the availability of an emergency department–embedded CCU was not associated with clinical outcomes among patients admitted with sepsis or ARF. Among less sick patients admitted with sepsis, the availability of an emergency department–embedded CCU was initially associated with reduced rates of direct ICU admission from the emergency department. Additional research is necessary to further evaluate the impact and utility of the emergency department–embedded CCU model.