1354
CRITICAL CARE
Nguyen et al. • CRITICAL CARE IN ED
CLINICAL INVESTIGATIONS Critical Care in the Emergency Department: A Physiologic Assessment and Outcome Evaluation H. BRYANT NGUYEN, MD, MS, EMANUEL P. RIVERS, MD, MPH, SUZANNE HAVSTAD, MA, BERNHARD KNOBLICH, MD, JULIE A. RESSLER, BS, ALEXANDRIA M. MUZZIN, BS, MICHAEL C. TOMLANOVICH, MD
Abstract. Objectives: The changing landscape of health care in this country has seen an increase in the delivery of care to critically ill patients in the emergency department (ED). However, methodologies to assess care and outcomes similar to those used in the intensive care unit (ICU) are currently lacking in this setting. This study examined the impact of ED intervention on morbidity and mortality using the Acute Physiology and Chronic Health Evaluation (APACHE II), the Simplified Acute Physiology Score (SAPS II), and the Multiple Organ Dysfunction Score (MODS). Methods: This was a prospective, observational cohort study over a three-month period. Critically ill adult patients presenting to a large urban ED and requiring ICU admission were enrolled. APACHE II, SAPS II, and MODS scores and predicted mortality were obtained at ED admission, ED discharge, and 24, 48, and 72 hours in the ICU. In-hospital mortality was recorded. Results: Eighty-one patients aged 64 ⫾ 18 years were enrolled during the study period, with a 30.9% in-hospital mortality. The ED length of stay was 5.9 ⫾ 2.7 hours and the hospital length of stay was 12.2 ⫾ 16.6 days. Nine (11.1%) patients initially accepted for ICU admission were later admitted to the general ward after ED intervention. Septic shock was the predominant admitting diagnosis. At ED admission, there was a significantly higher APACHE II score in nonsurvivors (23.0 ⫾ 6.0) vs survivors (19.8 ⫾ 6.5, p = 0.04), while there was no significant difference in SAPS II or MODS scores. The APACHE II, SAPS II, and MODS scores were
From the Departments of Emergency Medicine (HBN, EPR, BK, JAR, AMM, MCT), Internal Medicine (HBN), Surgery (EPR), Biostatistics (SH), and Epidemiology (SH) Henry Ford Hospital/Case Western Reserve University, Detroit, MI. Received March 6, 2000; revision received June 28, 2000; accepted July 3, 2000. Presented at the SAEM annual meeting, Chicago, IL, May 1998, and the Society of Critical Care Medicine annual meeting, San Francisco, CA, January 1999. Supported by the Weatherby Healthcare Resuscitation Research Fellowship. Address for correspondence and reprints: Emanuel P. Rivers, MD, MPH, Henry Ford Hospital, Department of Emergency Medicine, 2799 West Grand Boulevard, Detroit, MI 48202. Fax: 313-916-8675; e-mail:
[email protected]
significantly lower in survivors than nonsurvivors throughout the hospital stay (p ⱕ 0.001). The hourly rates of change (decreases) in APACHE II, SAPS II, and MODS scores were significantly greater during the ED stay (⫺0.55 ⫾ 0.64, ⫺1.02 ⫾ 1.10, and ⫺0.16 ⫾ 0.43, respectively) than subsequent periods of hospitalization in survivors (p < 0.05). There was a significant decrease in APACHE II and SAPS II predicted mortality during the ED stay (⫺8.0 ⫾ 14.0% and ⫺6.0 ⫾ 14.0%, respectively, p < 0.001) and equally at 24 hours in the ICU (⫺7.0 ⫾ 13.0% and ⫺4.0 ⫾ 16.0%, respectively, p ⱕ 0.02). The APACHE II and SAPS II predicted mortality approached actual in-hospital mortality at approximately 12 hours and 36 hours after ED admission (in the ICU), respectively. Conclusions: The care provided during the ED stay for critically ill patients significantly impacts the progression of organ failure and mortality. Although this period is brief compared with the total length of hospitalization, physiologic determinants of outcome may be established before ICU admission. This study emphasizes the importance of ED intervention. It also suggests that unique physiologic assessment methodologies should be developed to examine the quality of patient care, improve the accuracy of prognostic decisions, and objectively measure the impact of clinical interventions and pathways in the ED setting. Key words: physiologic scoring; APACHE II; SAPS II; MODS; emergency department; critical care; outcome. ACADEMIC EMERGENCY MEDICINE 2000; 7:1354–1361
T
HE DELIVERY of critical care is considered synonymous with the intensive care unit (ICU); however, critically ill patients are being cared for in the emergency department (ED) with increasing frequency. In some hospitals, up to 8% of all patients presenting to the ED and more than 25% of those admitted to the hospital are critically ill patients1 with a duration of stay ranging from 2.5 to 18 hours.1–4 Furthermore, there has been a 152% increase in the number of patients with ED length of stay greater than six hours from 1988 to 1997.5 This reality necessitates provision of critical
ACADEMIC EMERGENCY MEDICINE • December 2000, Volume 7, Number 12
care in the ED prior to ICU admission. Outside of the ICU, the operating room (OR), and postoperative care areas, critical care is more frequently delivered in the ED than in any other area in the hospital. Critical care services represent approximately 10% of the total health cost, 1% of the gross national product,6,7 and 7% of the total number of hospital beds and consume 20–30% of hospital costs.8,9 Because the delivery of critical care requires extensive resource allocation, methodologies to assess the quantity and quality of care are continuously being developed and refined to assess cost vs benefit. While numerous and extensive methodologies have been developed for the traditional ICU environment, similar techniques are lacking for the ED setting.10–12 The purpose of this study was to assess the impact of ED intervention on the morbidity and mortality of critical ill patients using the physiologic scoring systems, Acute Physiology and Chronic Health Evaluation (APACHE II),13 the Simplified Acute Physiology Score (SAPS II),14 and the Multiple Organ Dysfunction Score (MODS).15 Given that critically ill patients can receive a continuum of care delivered in multiple in-hospital settings, the temporal comparisons of care provided in these locations using these scores independently may provide objective physiologic and outcome assessment information.
METHODS Study Design. This was a prospective observational study of a consecutive cohort of critically ill adult patients presenting to a large urban ED and admitted to the medical ICU over a three-month period. The study was approved by the Institutional Review Board for Human Research. Study Setting and Population. The study was conducted in a 750-bed urban tertiary care hospital with an adult ICU bed capacity of 113. The ED is a 70-bed, Level 1 trauma center that provides care to approximately 84,000 patients per year, and 2.2% of these patients are admitted to the ICU. Critically ill patients are triaged and treated in a nine-bed intensive care area that is equipped with full hemodynamic monitoring and life support capabilities in the ED. This ED intensive care area is staffed by a board-certified emergency physician (EP), two emergency medicine residents, and four emergency nurses 24 hours a day. Patients presenting to the ED are admitted to the ICU (medical, cardiac, neurosurgical, or surgical) from this area. Study Protocol. Adult patients with a severity of illness necessitating ICU admission were consid-
1355
ered over a period from September 1, 1997, to December 1, 1997. Among these patients, those presenting to the ED were enrolled after informed consent within one hour of arrival if they had: 1) two of four systemic inflammatory response syndrome (SIRS)16 criteria (temperature >38⬚C or 90 beats/min, respiratory rate >20 breaths/min, or white blood cell count >12,000 cells/mm3, 10% band cells) and a systolic blood pressure 4 mmol/L). No patient who met the inclusion criteria was excluded due to lack of informed consent. All patients received standard ED intervention and enrollment did not affect clinical decision making by the ED staff. Patients were evaluated in the ED by a critical care fellow prior to the ICU admission. Patients with an age less than 18 years, with a primary diagnosis of myocardial infarction, unstable angina, pulmonary edema, seizure, pregnancy, hemorrhage, or trauma, or requiring acute surgical intervention were excluded. Patients with do-notattempt-resuscitation (DNAR) orders were also excluded from the study. The study investigators were not informed of excluded patients. Measurements. Demographic data and admission diagnoses were recorded. All patients were followed throughout the hospital stay. The ED length of stay was the time from ED arrival to discharge from the ED. The hospital length of stay was the time from ED discharge to hospital discharge. The number of critical-patient-days per year delivered in the ED was computed by multiplying the number of ICU admissions per year by the average ED length of stay (hours), then divided by 24 hours/ day. The number of critical-patient-days delivered per month is the number of critical-patient-days per year divided by 12 months/year. The variables required for the APACHE II,13 SAPS II,14 and MODS15 scoring systems were obtained at ED admission, at ED discharge, and at 24-hour intervals up to 72 hours in the ICU. These variables are routinely obtained for critically ill patients and the study did not alter clinical decision making. The chronic health history was obtained from the patient, the patient’s family, a computerized patient information system, and the patient’s medical record. The predicted mortality was obtained from the sum of individual predicted mortalities (calculated from APACHE II or SAPS II scores) divided by the number of survivors present at the corresponding time point. A computer database was developed (Paradox-Corel, Jericho, NY) to record patient demographics, therapy given, and physiologic variables, and to calculate APACHE II, SAPS II, and MODS scores and predicted mortalities. Actual in-hospital mortality was determined by dividing the number of deaths occurring in the hos-
1356
CRITICAL CARE
Nguyen et al. • CRITICAL CARE IN ED
TABLE 1. Demographic Information* Number of ED visits/year (1997) Number of ICU admissions/year Number of ward admissions/year Number enrolled in study Number admitted to the ICU Number admitted to the ward Number who died in the ED Age (yr) Gender Male Female
84,279 1,889 9,590 81 71 9 1 64 ⫾ 18 40 41
ED length of stay (hr) Hospital length of stay (days) Critical-patient-days/year Critical-patient-days/month Critical-patient-hours/day
5.9 ⫾ 2.7 12.2 ⫾ 16.6 464.4 38.7 30.5
Survivors Age (yr) ED length of stay (hr) Hospital length of stay (hr)
56 64 ⫾ 18 5.6 ⫾ 2.5 382.1 ⫾ 440.6
Nonsurvivors Age (yr) ED length of stay (hr) Hospital length of stay (hr)
25 64 ⫾ 18 6.5 ⫾ 3.0 90.8 ⫾ 139.6
Nonsurvivors during the study period ED admission ED discharge 24 hours in the ICU 48 hours in the ICU 72 hours in the ICU In-hospital mortality
0 1 4 8 13 30.9%
*Data are presented as absolute count or as mean ⫾ standard deviation. ED = emergency department; ICU = intensive care unit.
pital by the total number of patients in the study group. Data Analysis. The Statistical Analysis System (SAS Institute, Cary, NC) was used for data analysis. Unless otherwise indicated, data are presented as mean ⫾ standard deviation. Statistical significance is determined at p < 0.05. Mortality predictions were compared with actual in-hospital mortality. Binary variables such as gender were tested for survivor and nonsurvivor differences with a chi-square test. Continuous variables such as age, physiologic score, and predicted mortality were tested for differences between survivors and nonsurvivors or between time intervals with a twotailed Student’s t-test. If variances were unequal, Welch’s test17,18 was used as an alternative to Student’s t-test to account for the effect of unequal variances on type I error at the desired p < 0.05. The hourly rate of change in physiologic scores was computed by dividing the individual difference in
scores between two time points by the total number of hours elapsed. In order to examine how temporal changes in scores are related to survival, general linear mixed modeling19,20 was used. This model is a generalization of analysis of variance (ANOVA) with repeated measures, which accounts for correlations between time points and allows each patient to serve as his or her own control. The predictive abilities of the physiologic scoring systems, APACHE II and SAPS II, were evaluated using receiver operating characteristic (ROC) curve analysis.21 ROC curves were constructed to compare the sensitivities and specificities over the range of predicted mortalities at ED admission, at ED discharge, and at 24-hour intervals up to 72 hours in the ICU. Comparison of the predictive ability of APACHE II with that of SAPS II at these time points is not within the purpose of this study. The sensitivity (true-positive rate) indicates the percentage of patients predicted to die who actually die. The specificity indicates the percentage of patients predicted to survive who actually survive, or equivalently, 1⫺ specificity (falsepositive rate) indicates the percentage of patients predicted to die who actually live. The area under the curve (AUC) indicates the discriminatory power of the scoring system for the indicated time point. An AUC of 1.0 is perfect discrimination, whereas less than 0.5 would be less than chance. Previous studies using ROC curves for validation of mortality prediction models have reported acceptable AUCs of 0.78 to 0.90.22
RESULTS Eighty-one patients, 40 men and 41 women, who met the inclusion criteria out of 472 actual ICU admissions, or a 17.2% representation of critically patients, were enrolled during the study period. The remainder of patients who did not meet the enrollment criteria after initial assessment were not entered in the study. Among the enrolled patients, there were 13 deaths occurring during the study period (72 hours) and 25 (or 30.9%; 95 CI = 20.8% to 40.9%) in-hospital deaths. Thirty-five patients remained in the ED for more than six hours, ten patients more than nine hours, and five patients more than 12 hours. By Student’s t-test there was no significant difference in age (p = 0.86) or ED length of stay (p = 0.06) in survivors vs nonsurvivors. The total hospital length of stay for survivors was significantly greater than that for nonsurvivors (p < 0.001). The ED length of stay represented 1.5% and 7.2% of the total length of hospitalization for the survivors and nonsurvivors, respectively (Table 1). There were 113 available ICU beds with an average daily occupancy of 92% during the enrollment period. Nine (six survivors and three non-
ACADEMIC EMERGENCY MEDICINE • December 2000, Volume 7, Number 12
1357
survivors) or 11.1% of the patients initially accepted for ICU admission by the critical care fellow were later admitted to the general ward after ED intervention. Seventy-nine percent of the patients had septic shock as the admitting diagnosis (Table 2). Resuscitation and intravenous antibiotTABLE 2. Admitting Diagnoses Diagnostic Category
n
Decompensated heart failure Diabetic ketoacidosis Mesenteric ischemia Hypovolemic shock Pulmonary embolism Septic shock
4 3 1 8 1 64
TABLE 3. Monitoring and Therapeutic Interventions n Procedures performed Pulse oximetry Cardiac monitoring Central venous pressure monitoring Intra-arterial pressure monitoring Orotracheal intubation Paracentesis Esophageal Doppler hemodynamic monitoring
81 81 72 72 22 1 4
Vasoactive and inotropic agents used Dopamine Norepinephrine Dobutamine Neosynephrine Nitroglycerin
23 18 6 8 1
TABLE 4. APACHE II, SAPS II, and MODS Scores* Survivors
Nonsurvivors
p-value†
APACHE II 0 (ED admission) 6 (ED discharge) 24 (ICU) 48 72
19.8 16.5 12.6 12.0 11.8
⫾ ⫾ ⫾ ⫾ ⫾
6.5 5.9 5.1 4.9 4.4
23.0 21.2 21.0 18.9 18.7
⫾ ⫾ ⫾ ⫾ ⫾
6.0 5.8 6.2 5.5 5.5
0.04 0.001
