Outcomes of emergency medical patients admitted directly to an intermediate care unit with detailed admission guidelines

Abstract

Background

An important population receiving intermediate care, but not well characterized, is that composed of medical patients admitted directly from the emergency department.

Objective

Characterize emergency medical patients and their outcomes when admitted to an intermediate care unit with clearly defined admission guidelines.

Methods

Electronic health records of emergency medical patients admitted directly to an established intermediate care unit from July to December 2012 were examined. Demographic data, admitting diagnoses, lengths of stay, severity of illness, comorbidity, and hospital mortality were assessed.

Results

Three hundred and seventeen unique patients were admitted with a mean age of 54 (SD 16). The most common diagnoses at admission were sepsis (14%) and diabetic ketoacidosis (12%). By organ system, most patients were admitted with respiratory (27%) or cardiac (17%) syndromes. The mean Acute Physiology and Chronic Health Evaluation score version 2, Simplified Acute Physiology Score version 2, and Charlson Comorbidity Index were 15.6(SD 6.5), 20.7(SD 11.8), and 2.7(SD 2.3), respectively. Overall hospital mortality was 4.4% (14 deaths). Severity of illness and length of stay were significantly different for patients that required intensive care within 24 hours of admission (n = 16) or later (n = 25), those that continued with intermediate care for more than 24 hours (n = 247), and those that were downgraded or discharged in less than 24 hours (n = 29).

Conclusions

Emergency medical patients with moderate severity of illness and comorbidity can be admitted to an intermediate level of care with relatively infrequent ICU transfer and relatively low mortality.

Introduction

Intensive care resources are limited while the number of patients needing intensive care is increasing^1,2. It was previously recognized that among patients admitted to intensive care units (ICUs), many do not require intensive care and are admitted for close monitoring^3–6. Intermediate care units (IMCUs), also known as high dependency units, step-down units, or progressive care units, were created to accommodate patients whose needs do not require intensive care but surpass the care and monitoring feasible on a general ward^3,7–11. Patients may be transitioned to an IMCU after being stabilized in an ICU, having worsened on a general ward, or may be directly admitted from the emergency department (ED) or post-anesthesia care unit. Over the last 20 years, billing for intermediate care and the prevalence of IMCUs have increased^12–14. However, the optimal staffing structure, physical layout, and admission guidelines for these units are not well-defined. This is complicated by regional needs, institutional missions, clinical expertise, and physical resources. The result is a marked heterogeneity of IMCUs and the characteristics of the patients they serve¹⁵.

This heterogeneity makes the interpretation of IMCU patient outcomes and cost-effectiveness studies difficult^15,16. Though professional societies and governmental groups have published IMCU admission guidelines^8,17, they are broad and used in only a few studies^18–22. Other studies specify no, or very limited admission guidelines^23–28. Further, much of the IMCU literature is old and focused on the use of IMCUs as an alternative to postoperative ICU monitoring¹⁵. As a result, a contemporary “benchmark” of IMCU organizational structure paired with patient characteristics and outcomes is lacking^15,29,30. This has generated interest in more focused assessments of intermediate care^15,28,31.

An important population considered for intermediate care is that composed of medical patients admitted directly from the ED^{15,18,19,26,28,32}. Importantly, ED patients admitted to IMCUs who require ICU transfer soon after admission have higher mortality than similar patients admitted directly to ICUs³³. Appropriate initial triage may therefore prevent irreversible deterioration. The purposes of this report are to characterize the organizational structure of a well-established IMCU in a large urban academic medical center, and the diagnoses, severities of illness, and short-term outcomes of medical patients admitted directly from the ED in the context of detailed admission guidelines. This characterization may be helpful to institutions contemplating the development of an IMCU and to those interested in comparing outcomes to peer institutions.

Methods

The study protocol was approved by the Institutional Review Board of Johns Hopkins University (Protocol Number: NA_00083663).

Study Population

This is a retrospective study of patients admitted directly to the medical IMCU from the ED of an urban academic medical center between July 1^st, 2012 and December 31^st, 2012. All patients age 18 or greater admitted from the ED were included. Only data from each patient’s first IMCU admission were included.

IMCU Admission and Transfer Guidelines

Admission guidelines were developed from published guidelines with added detail (Table 1), including the intensity of monitoring and nursing care feasible on the unit^8,34. Note, code status is not addressed by the guidelines. For patients admitted to the IMCU, “triggers” for critical care consultation and ICU transfer were delineated. To preserve continuity, admitted patients could receive short intervals of intense care on the IMCU (i.e.: hourly vitals & 1:1 nursing for up to 4 hours), though these needs would have precluded initial admission.

Table 1.

IMCU Guidelines

Category	IMC Admission Guidelines	Trigger for ICU Consult/Transfer
Monitoring/General	Q2h vital signs or less frequent Labs Q2h (Q1h glucose) or less frequent Continuous pulse oximetry & cardiac monitor Arterial/venous pressure	Q1h vitals needed for > 4h (one 4h interval of Q1h vitals within 24h acceptable)
Respiratory	PaO2 ≥ 60 mmHg or SpO2 ≥ 90% Suctioning Q2h or less frequently Nebulizer treatment q2h or less frequently NC, Hi-Flow NC, FM O2 BIPAP/CPAP (new or chronic) Patient > 24h from tracheostomy Prostacyclin infusion for pulmonary hypertension	FiO2 1.0 for > 24h Respiratory rate > 35, accessory muscle use Suctioning Q1h or more often > 8h Continuous nebulizer treatment
CV – HTN	Hypertensive urgency Intravenous push anti-hypertensive Labetalol, nicardipine, nitroglycerine infusion	Hypertensive emergency Frequent titration of infusions (more often than Q2h)
CV – CHF, MI, Sepsis	Hemodynamically Stable NSTEMI Dopamine ≤ 10 mcg/kg/min (≤3 titrations/day) Dobutamine ≤ 10 mcg/kg/min (≤3 titrations/day)	STEMI Shock* Vasopressor for sepsis
CV – Arrhythmia	Risk of life threatening arrhythmia IVP adenosine, diltiazem, labetalol, metoprolol Diltiazem, labetalol, & amiodarone infusion	Need for temporary pacer (TC or TV) Bedside cardioversion Frequent titration of infusions (more often than Q2h)
Gastrointestinal	GI bleeding with orthostasis but not shock ≤ 10 point drop from baseline hematocrit Acute liver failure Hepatic encephalopathy ≤ grade III S/P uncomplicated TIPSS	GI bleeding with shock/need for venous sheath (Cordis) > 10 point drop from baseline hematocrit Grade IV encephalopathy, hepatic coma
Renal	Bedside intermittent hemodialysis Acute hemodialysis for drug intoxication Electrolyte abnormalities at risk for arrhythmias Electrolyte abnormalities requiring frequent labs	Hemodynamic intolerance of IHD
Metabolic	Metabolic disorders requiring frequent labs (i.e., DKA, non-ketotic hyperglycemia)
Hematologic	Thrombolytic infusion (no bolus) for stable patients	Bolus or infusion of thrombolytics for unstable patients (i.e. submassive PE)
Neurological	Neuro checks Q2h or less frequently High aspiration risk due to impaired mental status Alcohol WD (benzodiazepine infusions permitted) Opiate overdose (naloxone infusion permitted) PCA and epidural PCA pumps	Sustained Glasgow Coma Score < 9 Neuro checks more often than Q2h for > 8h Uncontrolled alcohol WD & frequent titration of benzodiazepine infusion
Miscellaneous	Endoscopy without sedation Venous sheaths permitted (non-hemorrhage patients) Arterial sheaths permitted × 4h	1:1 nursing care > 4h
Prohibited	Temporary pacing, PA catheters, lumbar drains, intrapleural bupivacaine, bladder pressure Non-emergent cardioversion, continuous nebulizer treatments, intermittent hemodialysis on vasopressors, continuous renal replacement therapy, procedural sedation

CHF: Congestive Heart Failure

CV: Cardiovascular

DKA: Diabetic Ketoacidosis

GI: Gastrointestinal

IHD: Intermittent Hemodialysis

HTN: Hypertension

MAP: Mean Arterial Pressure

MI: Myocardial Infarction

NC: Nasal Cannula

NSTEMI: Non-ST-Elevation MI

PA: Pulmonary Artery

PCA: Patient Controlled Analgesia

PE: Pulmonary Embolus

Q#h: Every # hours

SBP: Systolic Blood Pressure

TC: Transcutaneous

TIPSS: Transhepatic intravenous porto-systemic shunt

TV: Transvenous

WD: Withdrawal

^*Shock: SBP < 90 mm Hg or MAP < 60 mm Hg with end organ dysfunction

Triage to the IMCU

Within the limits of the admission guidelines, triage to the IMCU was determined by the ED attending physician in conjunction with the admitting medical resident and/or ICU resident as needed. This allowed clinicians to admit patients who objectively met criteria for IMCU admission to the ICU when there was concern for impending clinical deterioration. Patients with care needs exceeding IMCU standards remained in the ED until an ICU bed could be obtained or their condition improved. For patients triaged to intermediate care, the IMCU charge nurse reviewed ED data and the inpatient care plan to ensure compliance with admission guidelines.

IMCU Setting

The IMCU is an “open” unit, limited to medical patients¹⁵. It is in close proximity to the medical wards. All patients receive continuous pulse oximetry and cardiac monitoring (12-lead). During the first 3 months of data collection, the IMCU was a 15-bed unit and admitted an average of 119 (SD 7.5) patients/month. During the second 3 months of the study, the IMCU moved to an 18-bed unit and admitted an average of 136 (SD 12) patients/month. The average daily census before and after the move was 12.5 and 14.9, respectively. During the study period, the hospital staffed 980 beds of which 230 were controlled by the Department of Medicine. This included a 12-bed Cardiac ICU (CCU), a 12-bed Cardiac IMCU, an 18-bed Medical ICU (MICU), and the 15 to 18-bed Medical IMCU that is the focus of this report. The remaining 170 beds provided routine ward care with vital sign assessment every four hours and the option of telemetry. Of note, the IMCU and ICUs are located in different buildings with different nursing and physician staff.

IMCU Unit Staffing

Nursing shifts were staffed by a charge nurse without primary patient care responsibilities, one or two support associates, a unit clerk, and a nurse:patient ratio of 1:3. Nursing shifts were 12-hours with staff changes at 7 a.m. and 7 p.m. An additional 8-hour rotation (7a-3p, 3p-11p, 11p-7a) was staffed by a nurse who supported those on the 12-hour rotations and covered any unexpected absences. This was achieved with 29 full time nurse positions during the first 3 months and 34.5 positions during the second 3 months of the study. To work independently, each nurse completed an 11-week IMCU orientation program of classroom teaching, self-education, supervised bedside learning, and exams covering unit specific pharmacology, arrhythmia detection, and general knowledge. A critical care nursing credential was not required. Unit operations were further overseen by a physician medical director, a full time nurse manager, a clinical nurse specialist (0.2 Full Time Equivalents [FTEs]), and a unit safety officer (0.2 FTEs).

Ancillary Services

The unit was supported with a unit-dedicated respiratory therapist 24 hours per day. A full time inventory management clerk maintained stock of all supplies typically used on the unit (1 FTE). Physical, occupational, and speech therapy, wound care services, and nutrition staff were available from a hospital-wide pool. Social workers staffing the IMCU also staffed the MICU. Patient transport was supervised by an inpatient critical care transport service so that IMCU nurses were infrequently removed from bedside care.

Physician Providers

Patients admitted to the IMCU were managed by one of eight physician teams. Five were general medicine teams each with five intern housestaff supervised by two resident housestaff and one attending physician. The other three were subspecialty teams. One was dedicated to HIV patients, and the other two cared for patients with other subspecialty problems (i.e. pulmonary, gastrointestinal, renal, rheum, cardiomyopathy). Subspecialty fellows and attendings supervised the housestaff caring for their patients on these services. Patients were assigned to an intern or resident from one of these teams at the time of hospital admission or transfer from an ICU. This physician and team (daytime) and on call team members (overnight coverage) cared for their patients in the IMCU or ward (after transfer) until hospital discharge. This included daily bedside assessments and clinical decision making by the assigned physician and team (rounds). For patients transferred to the ICU, all care responsibilities were assumed by the ICU physician and nursing teams.

Data Collection

Patients admitted from the ED to the IMCU were identified from unit admission logs. Data were recorded in duplicate by trained physician abstractors working independently. Inconsistencies were arbitrated by a third independent review. The Acute Physiology and Chronic Health Evaluation score version 2 (APACHE II), Simplified Acute Physiology Score version 2 (SAPS II), and Charlson Comorbidity Index (CCI) were calculated from patient data obtained in the first 24 hours of hospitalization^35–37. Missing data were assumed normal.

Outcomes

Study outcomes included the proportion of patients for whom an ICU consult was obtained in the ED, severity of illness, comorbidity, lengths of stay (LOS), and hospital mortality. Patients were further characterized by location 24 hours after admission.

Statistical Analysis

Clinical and administrative data were characterized with descriptive statistics. Severity of illness, comorbidity, and LOS among survivors were compared using the Wilcoxon Rank-Sum test. The Fisher’s exact test was used for comparisons of hospital mortality. Data are described as means (SD) or medians (IQR), as appropriate.

Results

Patient Characteristics

During the study, 649 unique patients were admitted to the IMCU. Of these, 317(49%) were admitted from the ED. The demographic and admitting characteristics are detailed in Table 2. Over half were considered for ICU admission. The average APACHE II, SAPS II, and CCI scores were 15.6(6.5), 20.7(11.8), and 2.7(2.3), respectively. Data components for the calculation of APACHE II and SAPS II were missing in 2.5% and 6% of patients, respectively. When diagnoses are assigned by organ system, most patients were admitted with respiratory (26.5%) and cardiac related syndromes (17%). However, when considering specific diagnoses, the most common were sepsis (13.6%) and diabetic ketoacidosis (12%).

Table 2.

IMCU Patient Characteristics

Unique Admissions	N = 317
Mean Age (SD)	54 (16.4)
Female %	51
Race %
African American	64
Caucasian	31
Asian	1
Other	4
ICU Consult in ED %	55
Median Days LOS (IQR)	4 (3 - 8)
Hospital Mortality %	4.4
Mean Severity/Comorbidity (SD)
APACHE II	15.6 (6.5)
SAPS II	20.7 (11.8)
CCI	2.7 (2.3)
Primary Diagnostic Category %
Respiratory*	26.5
Cardiac	17.0
Sepsis*	13.6
Gastrointestinal	13.6
Endocrine (DKA)	12.0
Neurological	10.7
Metabolic/Renal	6.6

^*Most often transferred to an ICU within 24 hours of IMC admission

SD: Standard Deviation

IQR: Interquartile Range

LOS: Length of Stay

Characteristics by Need for ICU

After admission, patients followed one of four pathways. Within 24 hours of admission, 16(5%) were transferred to the ICU and 29(9%) were downgraded to a ward or discharged (Table 3). Among those that remained on the IMCU at 24 hours, 25(8%) were ultimately transferred to an ICU a median of 2.7 days (IQR: 1.7–3.8 days) after IMCU admission. The remainder either died on the IMCU (5 patients) or were downgraded or discharged. Patients transferred to the ICU at any time had higher APACHE II (p<0.001) and SAPS II (p≤0.06) scores than those that remained on the IMCU at 24 hours and never required ICU transfer. Those downgraded or discharged within 24 hours of admission had significantly lower APACHE II and SAPS II scores (p<0.01) than any other group. Comorbidity did not differ. Of the patients transferred to the ICU in the 24 hours following admission, all but one received interventions only available in an ICU (Table 4).

Table 3.

Median Severity of Illness, Comorbidity, and Mortality by ICU Admission Status and Location at 24 Hours

	ICU Admission		No ICU Admission
	Before 24 H (n = 16)	After 24 H (n = 25)	IMC > 24 H (n = 247)	Ward/Home ≤24 H (n = 29)
APACHE II (IQR)	21.5 (18 - 28.5)*	19 (16 - 22)*	15 (10 - 19)	12 (7 - 15)‡
SAPS II (IQR)	26 (12.5 - 43)†£	26 (22 - 40)*	18 (12 - 26)	12 (7 - 20)‡
CCI (IQR)	2 (2 - 3.5)	3 (1 - 6)	2 (1 - 4)	2 (1 - 3)
Number died (%)	2 (12.5)†	7 (28)*¥	5 (2)	0 (0)

^*P < 0.001 vs. No ICU Admission Groups

^†P = 0.06 vs. No ICU Admission & IMC >24 H

^£P = 0.004 vs. Ward/Home < 24 H

^‡P < 0.01 vs. IMC > 24 H

^¥P = 0.44 vs. ICU Admission Before 24 H

Note: Severity of Illness and Comorbidity calculated from data obtained during the first 24 hours of hospital admission.

Table 4.

Description of patients requiring transfer to ICU within 24 hours of admission to IMCU

Patient	IMCU Admit Dx	Additional Clinical Picture	Trigger for ICU Transfer	ICU Care
1	Sepsis	Bacteremia hypoxic respiratory insufficiency with ALI vs. CHF	High nursing needs. Worsening hypoxemia.	Intubated
2	Alcohol Withdrawal	Hypoxemia. Mental status change.	High nursing needs. Worsening Hypoxemia. Not handling secretions.	Intubated
3	Hypertensive Urgency	Hypertensive urgency	Intraparenchymal hemorrhage reclassified patient as hypertensive emergency.	Frequent titration of intravenous anti-hypertensives.
4	COPD Exacerbation	Trialed on BIPAP	High nursing needs. Progressive hypercapnia.	Intubated
5	Pneumonia	Progressive hypoxemia	Frequent nursing needs. High oxygen requirement and high respiratory rate. Aggressive fluid resuscitation. Frequent nebulized bronchodilators.	Intubated
6	COPD Exacerbation	Persistent respiratory insufficiency and hypotension	Frequent nursing needs. High oxygen requirement, respiratory rate, aggressive fluid resuscitation. Frequent nebulized bronchodilators.	Hourly nebulized bronchodilators
7	Hypercapnia	Femur fracture, narcotic therapy for pain. Trialed on BIPAP.	Frequent nursing needs. Failed extubation post fixation of femur.	Intubated
8	DKA, Hypertensive Urgency	Hypertensive urgency	Labile blood pressure. Acute left iliac artery dissection reclassified patient as hypertensive emergency.	Frequent titration of intravenous anti-hypertensives.
9	Pneumonia	Volume overload, left ventricular ejection fraction 15%, tachycardia, progressive hypoxemia.	Frequent nursing needs, progressive hypoxemia, initiation of intravenous anti-hypertensives. BIPAP.	Diuresis, frequent titration of intravenous antihypertensives.
10	Alcohol Withdrawal	Escalation of intravenous benzodiazepines.	Frequent nursing needs and benzodiazepine titration. Delirium Tremens.	Deep sedation and Intubation.
11	COPD Exacerbation	Trialed on BIPAP	Frequent Nursing needs. Progressive hypercapnic respiratory failure.	Intubated. Placed on vasopressors.
12	COPD Exacerbation	Trialed on BIPAP	Low respiratory rate, progressive hypercapnia. Mental status change.	Cardiac arrest peri-intubation.
13	Sepsis	Progressive hypotension and acidosis despite fluid resuscitation and antibiotics. Trialed on BIPAP.	Frequent nursing needs, mental status change, progressive respiratory failure, and pH 7.10.	Intubation, vasopressors, inotropes.
14	Sepsis	Progressive hypotension, Ejection Fraction 15%	Frequent nursing needs. Hypotension unresponsive to fluid.	Vasopressors.
15	Sepsis	Progressive hypotension and acidosis despite fluid resuscitation and antibiotics.	Frequent nursing needs. Progressed to septic shock despite aggressive fluid resuscitation and antibiotics. PEA Arrest.	Intubated. Vasopressors.
16	NSTEMI	Dyspnea	Troponin > 30	None

ALI: Acute Lung Injury

BIPAP: Biphasic positive airway pressure

Dx: Diagnosis

CHF: Congestive Heart Failure

COPD: Chronic Obstructive Pulmonary Disease

DKA: Diabetic Ketoacidosis

NSTEMI: Non-ST-Elevation Myocardial Infarction

PEA: Pulseless Electrical Activity

Median hospital LOS for survivors was 4 days (IQR:3–8). Among patients that required ICU transfer at any time, hospital LOS was longer (12.5 days; IQR:8.5–22) than for patients that never required ICU transfer (4 days;IQR 2–7) (p<0.001). Hospital LOS was shorter (1 day; IQR:1–3) for patients downgraded within 24 hours compared to all other survivors (p<0.001). The majority of patients admitted remained on the IMCU at least 24 hours and never required ICU transfer (n = 242). Their median IMCU LOS of 2.4 days (IQR: 1.7 – 3.7).

Mortality

Hospital mortality was 4.4% overall (14 deaths) and was highest among patients transferred to the ICU more than 24 hours after IMCU admission (Table 3), two of whom had limitations in care (Do Not Intubate/Resuscitate). Though five patients died without ICU transfer, each was pursuing palliative, end-of-life care. No patients downgraded from the IMCU within 24 hours died or required ICU admission.

Discussion

We have characterized medical patients admitted directly from the ED to the IMCU of an urban academic medical center. The spectrum of diagnoses admitted, presented with severity of illness indices and LOS data, staffing and triage structure detail, and well-defined admission guidelines provide a point of reference that fills a gap in the literature.

Most prior studies of intermediate care focus on postoperative patients^{20,25,38–48}. Only a few studies report outcomes of predominantly medical patients, but they provide little characterization of patients admitted from the ED^{18,19,21,23,24,27,28}. Further, descriptions of admission guidelines, triage, and staffing structure are limited in these studies. For example, Franklin et al. described a 12-bed IMCU that provided “cardiorespiratory monitoring” and a nurse:patient ratio of 1:4²⁷. Patient source, severity of illness, and physician staffing models were not detailed. Auriant et al. described a 4-bed IMCU within an ED²⁴. Admission to, and discharge from the unit were at the discretion of an intensivist. With the exception of excluding patients in need of mechanical ventilation, hemodialysis, and invasive hemodynamic monitoring, admission guidelines were not detailed, nor was nurse:patient ratio. More recently, Torres et al. characterized 412 patients admitted to a 20-bed IMCU capable of electrocardiographic monitoring, non-invasive ventilation, invasive hemodynamic monitoring, and use of inotropic agents²³. Physician staffing and nurse:patient ratio were not reported. Lastly, Lucena et al. described a “closed” IMCU with up to 9 beds staffed by a single physician team composed of residents supervised by a hospitalist¹⁹. Nurse: patient ratio was 1:3 and admission guidelines were those of the Society of Critical Care Medicine, which are non-specific⁸. Though each study has improved our understanding of intermediate care, a more standardized definition and consistent admission guidelines would improve the interpretability of future research.

When considering admission guidelines for intermediate care, special attention should be given to the clinical trajectory of patients that more frequently fail outside of an ICU. In our study, 5% of patients required ICU transfer within 24 hours. These events may suggest undertriage and occurred most often in patients admitted with respiratory syndromes (ex: COPD exacerbation, pneumonia) or sepsis. Two other studies have addressed this concept^26,32. In one, patients admitted to mixed medical-surgical IMCUs were defined as undertriaged if they required “active life support therapy” during the first day of admission^3,26. This occurred in 8.3% of 8971 patients of whom 38% were admitted from an ED. Among nonsurgical patients, undertriage occurred most often in patients with pneumonia, sepsis, cardiac dysrhythmias, and gastrointestinal bleeding. The other study was composed of predominantly nonsurgical ED admissions to a ward or IMCU in 13 hospitals³². Again, the most common diagnoses transferred to an ICU within 24 hours of admission included respiratory, cardiac, and sepsis syndromes. Patients with these syndromes may therefore represent easily identifiable subpopulations that may benefit from further study to identify features that would trigger earlier ICU admission.

We have presented data on severity of illness, comorbidity, mortality, and LOS. Because IMCU structure and admission criteria of prior studies have varied, and due to our focus on medical patients admitted from the ED, it is not appropriate to compare our results to those previously published. However, severity of illness and hospital LOS differed significantly between patients who were transferred to the ICU, those that remained on the IMC at 24 hours and never required ICU care, and those downgraded or discharged within 24 hours of hospital admission. These findings are consistent with the presence of distinct sub-populations admitted from the ED to our IMCU. Importantly, median IMCU LOS for patients remaining in the IMCU at 24 hours who never required ICU transfer (76% of all patients) was of 2.4 days. These observations, and the fact that 55% of our patients were assessed by the ICU team before admission, suggests intermediate care has an established role in our management of moderately ill emergency medical patients.

There are important limitations of this study. First, our findings reflect the experience of a single center and are limited to medical IMCU patients admitted directly from the ED. The extent to which our guidelines would be acceptable and our results reproducible in other centers is not clear. However, because we have provided a detailed description of our unit staffing model (i.e., nurses, physicians, managers, other) and detailed admission guidelines, our results build upon recent multi-center epidemiologic studies^{14,21,26,28,32}. Though larger studies provide valuable insight into trends that may be present in many health care systems, they do not report the organizational detail we have provided. Such detail may be helpful to institutions contemplating the development of an IMCU, and to those interested in comparing outcomes to peer institutions.

Second, it has been well demonstrated that the presence of guidelines does not translate directly to adherence^49,50. Based on chart review, we have confidence that all but one patient that failed in the IMCU within 24 hours truly progressed to require intensive care (Table 4). However, it is likely that some patients received intensive care in the IMCU without ICU transfer. Indeed, to preserve continuity of care, short term intense care may be preferred to ICU transfer in some patients. Our unit guidelines address this by allowing a time-limited period of intense care (i.e.: hourly vital signs for up to 4 hours and hourly suctioning for up to 8 hours), even though these needs would preclude initial admission.

Third, we assumed missing data components in our calculation of APACHE II and SAPS II scores were normal. Though this occurred in only 2.5% and 6% of calculations, respectively, our scores could underestimate the true severity of illness of some patients. We chose APACHE II and SAPS II scores because they have previously performed well in the IMCU setting, and because they are well understood by the critical care community due to frequent use^19,22,24. Only recently was an IMCU-specific severity of illness and mortality prediction model developed (IMCU Severity Score)²². This new score was not available at the time of our data collection and should undergo external validation prior to broad use.

Lastly, though mortality was higher among patients admitted to the ICU, the study was not designed to assess the cause of this higher mortality. However, it is reasonable to consider that deaths of patients transferred to the ICU within 24 hours of admission may represent undertriage. Though deaths among those transferred to the ICU after 24 hours could reflect undertriage too, they may also suggest failure of therapy or limitations in ICU bed availability.

In conclusion, using detailed admission guidelines with triggers for ICU consultation and transfer, we have shown that emergency medical patients with moderate severities of illness and comorbidity, many of whom were considered for ICU admission in the ED, can be admitted to an IMCU with relatively infrequent ICU transfer and relatively low mortality. A more standardized definition of intermediate care and IMCU admission guidelines will improve the interpretability of future research in this area.