2004 CAEP/ACMU Scientific Abstracts - Oral Abstracts (#17 to #32)

CAEP Abstracts

CJEM 2004;6(3):173-212

abstracts:  001-016 | 017-032 | 033-052 | 053-075 | 076-092 | 093-111 | 112-132

017 Evaluation of a program to address frequent emergency visits.

Grafstein E, Burton K, Innes G, Pope D, Kalla D, Stenstrom R. Providence Health Care, Vancouver, BC

OBJECTIVE: The management of frequent emergency department (ED) users can be vexing and difficult for clinicians. We developed a designated medical profile (DMP) program to optimize and standardize the care of these patients. When ED staff refer a frequent visitor to the program, a multidisciplinary team (emergency physician, nurse, social worker and the patient's primary care provider) meet and develop a care plan to guide treatment on subsequent ED visits. Our objective was to assess the impact of the DMP program on the number of visits and mean ED length of stay (LOS) for enrolled patients. METHODS: A retrospective study, at St. Paul's Hospital, a Canadian inner city hospital. All patients enrolled in the DMP program between May 1999 and December 2002 were included, and study data were obtained from the ED administrative database after auditing database fields for reliability and validity. Each patient served as their own control, comparing the six-month period before to the six-month period after DMP enrolment. RESULTS: During the study period, 85 patients, accounting for 2.6% of all annual ED visits, were enrolled in the DMP program. Mean age was 43 years and 54 (64%) were male. The most common reasons for referral were substance misuse (43.5%) and chronic medical problems (43.5%). Most patients (76%) had more than one reason for referral, and concurrent disorder (psychiatric and substance misuse) patients comprised 25% of the study group. The mean number of ED visits per patient was 14.1 (SD = 11.6) before DMP and 5.6 (SD 6.6) after DMP (P < 0.00001). Mean ED LOS was 150 minutes (SD = 131.3) before DMP and 151 (SD 153.4) after DMP (P = 0.88). The number of hospital admissions fell from 70 (0.82/pt) before DMP to 32 (0.38/pt) after DMP (P < 0.001). The rate of left without being seen visits changed from 1.1/pt to 0.6/pt (P < 0.01). CONCLUSION: Patient specific care plans can reduce the number of ED visits and hospitalizations by frequent ED users.

Key words: difficult patients, care path, substance misuse

018 Emergency department length of stay and the risk of a missed diagnosis of acute myocardial infarction.

Schull MJ, Vermeulen MJ. Institute for Clinical Evaluative Sciences, University of Toronto, Toronto, ON

OBJECTIVE: Accurate diagnosis of acute myocardial infarction (AMI) is contingent on the timing of serum marker and ECG tests with respect to symptom onset. The need for serial testing is controversial, and results in longer ED length of stay (LOS). Shorter ED LOS is more efficient, but may increase the risk of misdiagnosis. We sought to examine the relationship between varying ED LOS and the risk of a missed AMI. METHODS: We linked hospitalization records for patients admitted to hospital for AMI in Ontario in fiscal 2001-02 with ED visit records in the 7 days preceding admission. Prior ED visits were considered related to the AMI admission if the primary ED discharge diagnosis matched a pre-specified list. RESULTS: A total of 16,533 patients were admitted to hospital with AMI. ED LOS was <4 hours in 57%, 4 to <8 in 27%, 8 to <12 in 9.3%, and 12 or more hours in 6.9% of patients. In the 7 days preceding admission, 434 (2.6%) had an ED visit in which the AMI diagnosis was missed. The rate of missed AMI was 2.8% among patients who had an ED LOS of <4 hours; 3.1% for LOS from 4 to <8 hours; 1.8% from 8 to <12 hours, and 0.9% for LOS >12 hours. The risk of a missed AMI diagnosis for an ED LOS of <8 hours was 2.9% compared with 1.4% for an LOS of > = 8 hours (odds ratio 0.48; 95% CI 0.34-0.68). In bivariate analyses, an ED LOS of <8 hours was associated with younger age (p < 001), male sex (p < 0.001), rural residence (p < 0.001), absence of AMI history (p < 0.01), registration on the day shift (p < 0.001), weekend (vs. weekday) arrival (p = 0.01), arrival by ambulance (p < 0.05), triage score (p < 0.001), small hospitals (p < 0.001), and fewer visits to other EDs in the previous year (p < 0.001). Neighborhood income and median number of visits to the admitting ED in the past year were not associated with a shorter ED LOS. CONCLUSION: AMI diagnoses are more likely to be missed among patients with an ED LOS of less than 8 hours. Comparison of median LOS across EDs may represent a useful measure of ED quality of care for cardiac patients.

Key words: myocardial infarction, error, chest pain

019 Does emergency department activity level affect triage categorization and admission to the hospital?

Spence JM, Murray MJ, Morrison LJ. St. Michael's Hospital, Division of Emergency Medicine, University of Toronto, Toronto, ON

OBJECTIVES: To determine whether there was a correlation between low and high emergency department (ED) volume and triage categorization. To determine whether the proportion of patients admitted in each triage category varied during low and high volume periods. METHODS: A retrospective analysis was conducted for all 1999 ED visits from a community hospital in Ontario. Analysis of level of ED activity was defined by: daily ED volume, and 2-hour volume. Low volume (LV) periods were considered less than 25th percentile and high volume periods (HV) were greater than 75th percentile. Subgroup analyses for 2-hour periods were conducted for shifts (day, evening, night). The 5-level Canadian Triage and Acuity Scale (CTAS) was used for triage scoring. Chi square analyses were conducted. RESULTS: There were 68,757 visits available for analysis. Mean number of patients seen daily was 191 (range 132-278). Fewer patients were seen at night. When daily census was reviewed, small downward shifts in triage level were seen on HV days (p < 0.0001, OR 0.84 (95% CI 0.78-0.91)). The proportion of patients admitted on HV days, after controlling for triage level, was significantly different (p < 0.0001), decreasing for all levels (see Table, next column). When ED shifts were reviewed, acuity levels significantly decreased for HV periods (p < 0.0001), however, the proportion of patients admitted in each triage level did not change for days (p = 0.56) or evenings (p = 0.14). Lower adjusted admission rates were seen on night shifts (p < 0.0001). CONCLUSIONS: High volume days are associated with lower acuity ratings, as defined by CTAS level. However, even after controlling for acuity, admission rates decrease during periods of higher departmental activity.

Key words: triage, emergency health services

020 A process evaluation of patient flow in the emergency department (ED) prior to initial physician assessment.

Uy C, Guttman A, Verter V, Colacone A, Rosenthal S, Afilalo M. SMBD-Jewish General Hospital and Faculty of Management Science, McGill University, Montreal, PQ

INTRODUCTION: Emergency department crowding (EDC) is a serious problem facing hospitals nationwide. Applying operations management methodology to analyze complex ED processes can be helpful in identifying EDC causes. OBJECTIVE: To evaluate patient flow (PF) patterns and system bottlenecks in the ED from patient arrival to initial physician (MD) assessment. METHODOLOGY: An observational time-and-motion study was performed at a busy urban teaching hospital. Critical factors that affect PF were initially mapped by interviewing medical, nursing, and clerical staff. The second phase consisted of a thorough observation of PF through the entire ED from the time of patient arrival to discharge. Finally, a 2-month data collection period ensued with a single observer tracking patients from arrival to MD assessment between 10:00 am and 5:00 pm. The hospital's information system provided data on socio-demographics, triage code, mode of arrival, chief complaint, final disposition and timing. RESULTS: 360 patients in the acute care unit (ACU) and 341 in the ambulatory care unit (AMCU) were observed. The mean and standard deviation of waiting time for triage in the ACU was 19 min +19. Waiting for physician assessment (28 min + 27) was the most time-consuming activity for patients in the ACU. In contrast, AMCU patients spent the longest time waiting to enter an examination cubicle (86 min + 72). Frequency distributions of the seven sub-processes studied revealed high variability in flow times. Results also indicated that triage code was a good gauge for waiting time to be triaged (each unit increase in triage code resulted in an additional 5-min. wait to be triaged) and for waiting time for an examination cubicle in the ACU. CONCLUSION: Operations management methodology is effective in identifying system bottlenecks in the ED. The development of long-term solutions to EDC requires detailed analyses through more sophisticated methods, such as simulation-based optimization.

Key words: overcrowding, quality improvement

021 Overcrowding of Ontario emergency departments and door-to-needle time delays in acute myocardial infarction.

Schull MJ, Vermeulen MJ, Slaughter G, Morrison LJ, Daly P. Institute for Clinical Evaluative Sciences, University of Toronto, Toronto, ON

OBJECTIVE: To estimate the impact of emergency department (ED) overcrowding on door-to-needle time (DNT) for patients with suspected acute myocardial infarction (AMI) in EDs linked by ambulance diversion systems. METHODS: Data were obtained on patients thrombolysed in the ED for suspected AMI from 1998 to 2000 in 25 Ontario hospitals. Neighboring EDs linked by ambulance diversion systems were assigned to networks of between two to five hospitals. For each patient, the ambulance diversion status of all network EDs at the time of patient registration was determined. The proportion of network EDs that were diverting ambulances at the time of patient registration was used to categorize the network as having none (0%), moderate (<60%), or severe (> = 60%) network overcrowding. DNT was defined as time from ED registration to drug administration. Covariates included age, sex, diagnostic first ECG, past medical history, type of AMI, vital signs, day of week, time of day and hospital type. RESULTS: A total of 3452 thrombolysis patients were included: mean age was 62.9 years and 72.7% were male. Overall median DNT was 43 minutes (IQR 27-82); for 29.2% of patients it was <30 minutes, and for 34.9% it was >60 minutes. Median DNT increased from 40, to 45, to 47 minutes in conditions of no, moderate and severe network overcrowding respectively (p < 0.001). When compared with no overcrowding in multivariate analyses, moderate and severe overcrowding conditions were associated with increased median DNT (3.0min [95% CI 0.09-6.00] and 5.8 min [95% CI 2.68-8.95] respectively). The odds ratio for DNT exceeding target (>30 minutes) and major delay (>60 minutes) were 1.32 (95% CI 1.06-1.65) and 1.40 (95% CI 1.13-1.73) respectively for severe network overcrowding as compared to none. CONCLUSION: ED overcrowding is independently associated with increased median door-to-needle times for patients with suspected AMI. The severity of overcrowding may be underestimated unless conditions at all EDs linked by ambulance diversion systems are considered.

Key words: overcrowding, emergency health services

022 Does triage level (Canadian Triage and Acuity Scale ) correlate with resource utilization for emergency department visits?

Murray MJ, Levis G. Royal Victoria Hospital, Barrie, ON

INTRODUCTION: The CTAS defines the level of acuity of emergency patients. CTAS may be a valid indicator and predictor of ED resource utilization. OBJECTIVE: The purpose of this study is to determine if a relationship exists between triage level and resource use in the emergency department. The areas of study are: any imaging use, computerized tomography(CT),ultrasound(US),plain film radiography (extremity and non extremity), electrocardiograms (ECG's), microbiology procedures (culture and sensitivities), lab tests (one or more), and blood bank tests (blood typing with or without crossmatch). METHODS: This was a retrospective study of a database of emergency department visits (n = 164,812) in a large community hospital for a 26 month period from April 2001 to June 2003. The Z-test for difference in two proportions was used to see if there was a statistically significant difference between the percentage test utilization rate for a given triage level and all lower levels. RESULTS: The information was available for 164,589 patients. The distribution of patients over the triage levels from highest acuity to lowest was: CTAS 1 - 0.5%, CTAS 2 - 11.1%, CTAS 3 - 32.7%, CTAS 4 - 42.2 %, CTAS 5 - 13.5%. The rates of use of any imaging (CT, US, plain films) was CTAS 1 - 82.3%, CTAS 2 - 73.2%, CTAS 3 - 46.2%, CTAS 4 - 28.8 %, CTAS 5 - 24.2%. There was a statistically significant difference (p < 0.001) in imaging rates for a given triage level and the next lower level for all CTAS levels for any imaging, CT, and plain film non-extremity exams. The relationship did not exist across all levels for ultrasound or plain film extremity radiography. There was a statistically significant difference (p < .001) across all CTAS levels for rates of use of ECG, lab tests, and blood bank tests. CONCLUSIONS: CTAS level which defines level of acuity correlates with resource utilization for any imaging procedure, CT, non extremity plain film exams, ECG's, lab tests and blood bank tests. Triage level may be a valid indicator and predictor of ED resource utilization.

Key words: triage, Canadian triage and acuity scale

023 Does the Canadian emergency department triage and acuity scale correlate with admission to the hospital from the emergency department?

Spence JM, Beaton DE, Murray MJ, Morrision, LJ. St. Michael's Hospital, Division of Emergency Medicine, University of Toronto, Toronto, ON

BACKGROUND: Triage level has been linked to emergency department outcome. However, the relationship between the 5-level Canadian Triage and Acuity Scale (CTAS) and admission to hospital has not been fully explored. OBJECTIVES: To determine the relationship between CTAS level and admission to hospital at a single site. To identify factors associated with admission using multivariable modeling. METHODS: This was a retrospective study of ED visits to a community hospital in 1999. Univariate analysis was used to define the relationship of CTAS level and admission. Logistic regression (LR) was used to define predictors of admission. For statistical reasons (i.e. quasi-separation of data) the two highest triage levels, CTAS 1 and CTAS 2, were combined. RESULTS: Information for all visits (N = 68,757) was available. The hospital was representative of a community hospital in Ontario. The proportion of patients in each category was: CTAS 1 - 0.4%, CTAS 2 - 12.0%, CTAS 3 - 36.8%, CTAS 4 - 32.0%, and CTAS 5 - 18.9%. Admission rates were: CTAS 1 - 94.0%, CTAS 2 - 40.3%, CTAS 3 - 16.3%, CTAS 4 - 2.1%, and CTAS 5 - 0.3%. Univariate analysis calculated the odds ratios (OR) for admission for triage levels as: CTAS 1 & 2 - 212 (95% CI 157-268), CTAS 3 - 57 (95% CI 43-77), CTAS 4 - 6 (95% CI 5-9), CTAS 5 - 1.00. Other important associations in the univariate analysis were age over 65 with 6-fold and ambulance use with 8-fold increased odds of admission. Sex, income, and time of visit correlated poorly. In the LR model including age, sex, ambulance use, income, diagnosis, weekend, and night visit, CTAS remained a strong predictor of admission. Adjusted OR were: CTAS 1 & 2 - 64 (95% CI 48-88), CTAS 3 - 24.7 (95% CI 18-33), and CTAS 4 - 4.8 (95% CI 4-6). CONCLUSIONS: Increased acuity, as defined by CTAS level correlates with admission to hospital.

Key words: triage, Canadian triage and acuity scale

024 Reassessing the methods of medical record review (MRR) studies in emergency medicine research ten years after.

Worster A, Bledsoe D, Cleve P, Eva K, Fernandes C. Division of Emergency Medicine, Hamilton Health Sciences Corporation, Hamilton, ON

INTRODUCTION: A unique study assessing the methods of medical record review (MRR) studies in emergency medicine (EM) journals during a 5-year period ending in 1993 demonstrated significant deficiencies. Strategies for improving the quality of MRR studies using methodologic standards of health research were provided. Since then, no further assessments of MRR studies have been reported. The objective of this study was to reassess quality of MRR studies published 10 years after the initial assessment to determine if and where further improvements are still needed. METHODS: Researchers conducted computerized and hand searches of all articles published between 2000 and November 2003 in three EM journals. 20 articles from each of these journals meeting a published definition of MRR were analyzed using eight criteria from the original MRR study and eight additional criteria. Reviews were conducted independently by two researchers blinded to the study hypothesis and disagreements resolved by arbitration. RESULTS: MRR studies continue to made up a large proportion of all original research articles (28.2% in 2003 vs. 25% in 1993). Statistically significant improvement (<0.001) is seen in only two of the original eight categories: use of data abstraction forms, having increased from 10.7% (7 - 15) to 40.3% (38.7-41.9) and; testing of inter-rater reliability up from 0.4% (0 - 2) to 4.8% (CI: 4.1%-5.5%). A number of criteria are found in less than half of articles: abstractor training 11.3% (10.3-12.3), performance monitoring 6.5% (5.7-7.3), and blinding to hypothesis 1.6% (1.2-2.0); reporting 9.7% (8.7-10.7) and testing of inter-rater agreement; sample size calculation 3.2% (2.6-3.8) and; management of missing and conflicting data 25.8% (24.4-27.2). Inter-rater reliability is reported for each variable measured. CONCLUSIONS: MRR remains a frequently used method of EM research. However, methodologic standards of these studies have improved only minimally in the past 10 years leaving in question the validity of their results.

Key words: research, methodology

025 Emergency department overcrowding as ecologic problem: A comparison of measurement methods and the effect on time to thrombolysis.

Vermeulen MJ, Schull MJ. Institute for Clinical Evaluative Sciences, University of Toronto, Toronto, ON

OBJECTIVE: The effect of emergency department (ED) overcrowding has typically been studied in patients in the overcrowded ED alone. We examined the effect of overcrowding on patient outcome when measured at the treating ED alone vs. across a network of EDs linked by an ambulance diversion system. METHODS: We studied door-to-needle time (DNT) in a registry of AMI patients thrombolysed in the ED from 1998-2000 in 25 Ontario hospitals, linked to AD data in each ED at the time of patient arrival. Each ED was assigned to a network of between 2 to 5 surrounding EDs, all linked by a common ambulance diversion system. We compared the effect of overcrowding (defined as ambulance diversion) on DNT when it was measured only at the treating ED vs. measured as the percentage of overcrowded EDs (including the treating ED) across the network [which was grouped as none (0% of EDs), moderate (<60%), or high (> = 60%)]. Logistic regression was used to estimate odds ratios (ORs) for missing the target DNT (<30 minutes) and for major delay (>60 minutes). Quantile regression was used to estimate median DNT. Estimates were adjusted for age, sex, clinical presentation, ECG findings, past history, thrombolytic used, and setting. RESULTS: 2,927 AMI patients had complete data for analysis. Adjusted ORs for overcrowding at the treating ED only, compared with high network overcrowding, were 1.23 (95% CI 1.01, 1.49) vs. 1.32 (95% CI 1.06, 1.65) for missing the 30-minute target DNT and 1.23 (95% CI 1.02, 1.49) vs. 1.40 (95% CI 1.13, 1.73) for major DNT delay. The increase in median DNT was 3.6 minutes (95% CI 0.4-6.8) when the treating ED only was considered vs. 5.8 minutes (95% CI 2.7-8.9) when a high proportion of network hospitals were overcrowded. Moderate network overcrowding was associated with a 3.0-minute (95% CI 0.1-6.0) increase in median DNT but not with missed target DNT. CONCLUSION: In hospital networks linked by an ambulance diversion system, measurement of overcrowding at the treating ED alone may underestimate the adverse effect on DNT delays.

Key words: overcrowding, emergency health services

026 Clinically meaningful values of the verbal numeric rating scale of pain severity.

Lee JS, Hobden EJ, Stiell IG, Wells G. Sunnybrook and Women's College Health Sciences Centre, University of Toronto, Toronto, ON

INTRODUCTION: Clinicians use the verbal Numeric Rating Scale or NRS ("Rate your pain severity from 0 to 10") more frequently than any other measure of pain severity. Previous studies have compared changes in the NRS with descriptive categories and patient satisfaction. The objectives of this study was to define the minimum clinically important difference for the Numeric Rating Scale (NRS) of pain severity by measuring the change in NRS associated with adequate pain control. METHODS: We conducted a prospective, observational study. Adult emergency department (ED) patients with undifferentiated acute pain (<1 week) were eligible. Chest-pain patients were excluded. Patients rated their pain severity on a 10-point verbal NRS on presentation and at discharge. Patients were asked if they would accept any analgesic, then if they would accept a parenteral analgesic prior to treatment. At discharge, they were asked whether they had received adequate pain control. RESULTS: We enrolled 143 patients (mean age 36 years, 54% female). The mean decrease in NRS was -2.41 (95% CI, -3.5 to -1.3) for the 116/143 (81%) patients with adequate pain control at discharge vs. -0.74 (95% CI, -1.3 to -0.16 ) for the 27 with inadequate pain control (p = 0.009). At discharge, the mean NRS was 4.53 for those with adequate pain control vs. 6.07 for those without. The mean NRS for the 114/143 patients who would accept any analgesics initially was 7.05 vs 6.41 for the 29 reporting no analgesic need. Initially 77 patients would accept parenteral analgesics (mean NRS 7.73), and the mean NRS for the 65 patients not requiring IV analgesics was 5.97. CONCLUSION: A mean reduction in NRS of 2.41 represents a clinically important difference in pain severity that corresponds to patients' perception of adequate pain control. Defining MCID based on adequate analgesic control rather than minimal change may be more appropriate for the design of future analgesic trials, given the availability of effective treatments for acute pain.

Key words: pain management, pain measurement

027 Optimal two-hour cardiac marker strategies to identify low risk patients not requiring chest pain unit evaluation.

Innes GD, Christenson J, Wong H, Yu E, Rosenberg F, Grafstein E. St. Paul's Hospital; The University of British Columbia; Vancouver, BC

INTRODUCTION: Chest pain unit (CPU) protocols involve 6-12 hr stays, multimarker analyses, imaging and stress tests. Applied unselectively, these may increase diagnostic cost. Our objective was to identify two-hour, marker-based strategies that are 98% sensitive for ACS (AMI or unstable angina) and that identify at least 25% of patients safe to discharge without CPU evaluation. METHODS: This prospective cohort study enrolled consenting patients with chest pain and no traumatic or x-ray cause. Myoglobin, TnI and CKMB assays (Biosite Diagnostics) were drawn at 0 and 2 hrs. At 30 days, a final diagnosis of ACS (AMI or UA) or no ACS was assigned based on explicit criteria. We correlated ED test results with 30-d outcomes, used ROC analysis to find optimal test cut-offs, and evaluated marker strategies using these. RESULTS: Of 686 patients, 58 (8.5%) had AMI and 81 (11.8%) had UA, for a total of 139 (20.3%) with ACS. Optimal cutoffs were 3.0 for CKMB, 0.02 for TnI, and 100 (or doubling) for myoglobin. The Table (below) summarizes for each strategy diagnostic performance and proportion of patients who would be discharged without CPU evaluation (DC%). A triple marker strategy was 95% sensitive and 19% specific, but would increase CPU evaluations. Other strategies would reduce CPU evaluation rates but had poor sensitivity. No strategy was 98% sensitive for ACS. CONCLUSIONS: Two-hour marker strategies are promising but do not identify patients safe for discharge without CPU evaluation. Combining key clinical predictors with marker data may identify patients safe for early discharge.

Key words: chest pain, acute coronary syndrome, myoglobin, troponin, cardiac marker

028 ALARMED: Adverse events in low risk ACS patients receiving continuous cardiac monitoring in the emergency department: A pilot study.

Atzema C, Schull MJ, Borgundvaag B, Slaughter G, Lee C. Division of Emergency Medicine, SWCHSC, University of Toronto, Toronto, ON

INTRODUCTION: Most patients who present to an emergency department (ED) with chest pain are placed on continuous electrocardiographic monitoring (CEM). We intend to determine the rate of adverse events (AEs) identified by CEM in ED patients with chest pain. METHODS: We prospectively enrolled stable patients who presented to a single ED over a two-month period with a chief complaint of chest pain, who were placed on CEM. Patients were classified according to risk of poor outcome using published risk stratification tools. Trained observers recorded number of monitored hours, alarms, and any pre-defined changes in management or AEs. AEs were either "vital sign" or "arrhythmia" events. The primary outcome measure was the rate of AEs detected by CEM. The secondary outcome measure was the rate that alarms resulted in a change in management. RESULTS: We enrolled 72 patients, the majority (59%) of whom were categorized as the lowest risk scores by TIMI-IIB criteria, although six patients (8%) had ST segment deviation on initial ECG. During 371 monitored-hours, we recorded 1762 alarms (1019 arrhythmia alarms and 743 minor, bed alarms), a rate of 4.7 alarms per monitored-hour. There were no vital sign events during the monitoring period. There were 12 arrhythmia events identified by alarms (0.68%; 95% CI: 0.35% - 1.2%); none were hemodynamically significant, and three were associated with a change in management (0.2%; 95% CI: 0.04% - 0.5%). Eight of these patients had a heart rate of >=45, and received no treatment. The ninth was asymptomatic and given atropine for a heart rate of 32. The other three AEs were an SVT that required no therapy, a sinus pause that triggered a change in rate of IV Nitroglycerin, and a run of PVCs after which heparin was given. Of the three patients with a change in management, none was in the lowest risk categories. CONCLUSIONS: Routine continuous ECG monitoring in low and intermediate risk ED chest pain patients results in a large number of alarms, the vast majority of which require no change in management.

Key words: cardiac monitoring

029 Missed diagnosis of acute myocardial infarction in the emergency department varies by region and hospital size.

Schull MJ, Vermeulen MJ. Institute for Clinical Evaluative Sciences, University of Toronto, Toronto, ON

OBJECTIVE: Accurate diagnosis of acute critical medical conditions is essential to emergency department (ED) care. We sought to estimate the population-based rate of missed diagnoses among patients admitted to hospital with acute myocardial infarction (AMI). METHODS: We linked hospitalization records for patients admitted to hospital in Ontario in fiscal 2001-02 for AMI with ED visit records in the 7 days preceding admission. Prior ED visits were considered related to the AMI admission if the primary ED discharge diagnosis matched a pre-specified list. RESULTS: A total of 16,533 patients were admitted to hospital with AMI; 84% were in mid-size community hospitals. Mean age was 68 years (SD 14), 63% were male, 18% had a rural residence, and 29% arrived by ambulance; 51% were triaged as higher acuity and 6.3% as lower acuity. Survival to discharge and to 1 year was 90% and 79% respectively. Prior to AMI admission, 434 (2.6%; 95% CI 2.4%, 2.9%) had an ED visit in which the AMI diagnosis was missed; of these, 278 (64%) had attended the ED within 2 days of their admission. In bivariate analyses, prior related ED visits were associated (p < 0.01) with younger age, rural residence, lower acuity triage, walk-in arrival, shorter ED length of stay, and a higher number of visits to both the treating and other EDs in the previous year. The rate varied significantly by region, from 2.0% in the Toronto metropolitan area to 3.3% in the east and remote north of the province (p < 0.01). Missed AMI was also less likely to occur in teaching (1.3%) than in small hospitals (5.1%). No associations with sex, neighborhood income, history of AMI, time of day, or day of week were found. CONCLUSION: AMIs are missed in a prior ED visit in about 2.6% of Ontario AMI patients. This rate compares well with published rates of missed AMI, but our analysis suggests substantial regional variations and a more than 3-fold increase in risk in small hospitals compared with teaching institutions.

Key words: myocardial infarction, error, acute coronary syndrome

030 A prospective analysis of decision making by emergency physicians following successful resuscitation from out-of-hospital cardiac arrest.

Forrest DM, Abu-Laban RB, van Beek CA, Gin K. Division of Critical Care Medicine, Department of Medicine, University of British Columbia, Vancouver, BC

INTRODUCTION: The process of decision-making about continuation or withdrawal of therapy in the Emergency Department (ED) after resuscitation from cardiac arrest (CA) is poorly understood. The objective of this study was to determine the frequency of and factors related to withdrawal of therapy by emergency physicians (EPs) in patients resuscitated from CA. METHODS: From 07/2002 to 08/2003, patients presenting to the ED with at least 20 minutes of pulse return following out-of-hospital CA were eligible. Data was collected prospectively and descriptive statistics with 95% confidence intervals were generated. Logistic regression analysis was performed using 7 covariates defined a priori as potentially influential in EP decision-making (age, witnessed arrest, bystander CPR, initial rhythm, pulselessness duration, comorbidities, and family presence in the ED). RESULTS: Fifty-nine patients were enrolled with the following characteristics: witnessed arrest 55%; bystander CPR 28%; VF/VT rhythm 34%; median pulselessness duration 26 minutes; 2 or more comorbidities 48%; family presence 48%. Therapy was withdrawn in 8 cases (13.6%: 95% CI 6.0% to 25.0%); continuation of therapy usually resulted in referral to ICU (14/51) or CCU (34/51). The only factor significantly associated with therapy withdrawal on logistic regression was patient age (p = 0.05). The median/mean/SD age of patients in whom therapy was withdrawn was 84.3/81.2/13.2 years vs 64.5/63.1/16.1 years for those in whom therapy was continued (univariate p = 0.004). CONCLUSIONS: EPs withdraw care in a substantial number of patients after successful CA resuscitation. The only factor significantly associated with therapy withdrawal is patient age, although age has been shown not to be associated independently with prognosis in CA. It is unclear whether age is a surrogate for other unmeasured factors (such as an advanced directive), or whether our findings arise from institutional or physician bias. Further research in this area is warranted.

Key words: cardiac arrest, resuscitation

031 Missed diagnosis of subarachnoid hemorrhage in the emergency department: A population-based analysis.

Schull MJ, Vermeulen MJ. Institute for Clinical Evaluative Sciences, University of Toronto, Toronto, ON

OBJECTIVE: A missed diagnosis of subarachnoid hemorrhage (SAH) is associated with increased mortality, but there are no population-based estimates of the frequency of this event. We sought to estimate the frequency with which a SAH was missed in patients presenting to an Ontario emergency department (ED). METHODS: We linked hospitalization records for all patients admitted to hospital in Ontario in fiscal year 2001-02 with a final diagnosis of non-traumatic SAH to all ED visit records in the 7 days preceding the admission date. Prior ED visits were considered related to the SAH admission if the primary ED discharge diagnosis matched a pre-specified list of diagnoses and symptoms. RESULTS: 382 patients were admitted to hospital with SAH; 64% were in mid-size community hospitals. Mean age was 58 years (SD 16), 41% were male, 16% had a rural residence, and 52% arrived by ambulance. 47% of patients were triaged as higher acuity, vs. 11% triaged as lower acuity. Survival to discharge and to 1 year was 62% and 54% respectively. In the 7 days before admission, 25 (6.5%; 95% CI 4.3%, 9.5%) had a related ED visit in which the SAH diagnosis was missed. In bivariate analyses, missed SAH was associated with walk-in arrival (p < 0.001), lower acuity triage (p < 0.01), and visits to other EDs in the previous year (p < 0.001). In small hospitals 6 of 13 (46%; 95% CI 19%, 75%) SAHs were missed vs. 17 of 246 (6.9%; 95% CI 4.1%, 10.8%) and 2 of 123 (1.6%; 95% CI 0.2%, 5.8%) in community and teaching hospitals, respectively. Missed diagnosis was not associated with age, sex, income, hospital region, rural residence, time of day, day of week, ED length of stay, or visits to the admitting ED in the past year. CONCLUSION: About 6.5% of SAH patients are missed on their initial presentation to an Ontario ED. The rate is likely higher among patients with a normal or near-normal clinical exam since we included all SAHs regardless of severity. The rate of missed SAH was in small hospitals was markedly higher in small hospitals as compared with larger institutions.

Key words: subarachnoid hemorrhage, error

032 Detection of alpha-II spectrin and breakdown products in humans after severe traumatic brain injury.

Papa Ricci L, D'Avella D, Aguennouz M, Angileri FF, de Divitiis O, Germano A, Toscano A, Tomasello F, Vita G, Pineda J, Hayes RL, Pike B, Wang KK. Department of Emergency Medicine, University of Florida, Gainesville, FL

INTRODUCTION: Alpha-II-spectrin as a biomarker of injury severity in traumatic brain injury (TBI) has been shown in animal models. This study in humans: 1) examined levels of spectrin and spectrin breakdown products (SBPs) measured in the CSF of adults with severe TBI and 2) examined the relationship between levels and clinical outcome. METHODS: This prospective case control study enrolled a convenience sample of adult patients presenting to a tertiary care hospital following a severe TBI defined by a GCS score of <8 and requiring invasive intracerebral monitoring. Controls were patients without TBI requiring CSF drainage for other medical reasons. Ventricular CSF was drained from each patient at 6, 12, 24, 48, 72 and 96 hours following TBI and measured for spectrin and SBPs. Outcome was assessed using the Glascow Outcome Score at 6 months after injury. Outcome was classified into two groups: patients in Group 1 (worse outcome) died or were in a vegetative state and patients in Group 2 (better outcome) were disabled. Data are expressed as mean%-control (+/-SD). Data were analyzed using Fisher's exact test, Mann Whitney U, and ANOVA. Association with outcome was tested using the unpaired 2-tailed t-test (p < 0.05). RESULTS: Over 3 months 8 patients with severe TBI were enrolled: mean age 40 years (range 18-85), 82% male, and a median GCS = 5 on ED arrival. CSF spectrin and SBP levels were increased a mean of 26-fold (range 10 to 96 fold) over control at all time points (p < 0.001). CSF spectrin decreased by 2900(+/-1840)(p = 0.019) and SBPs decreased 2500(+/-2120) (p = 0.043) from 6 to 96 hours in patients with better outcome. Patients whose spectrin and SBP levels remained elevated or failed to decline by 2500 had a worse outcome. CONCLUSIONS: There is a significant increase in CSF spectrin and SBPs following a severe TBI in humans compared to uninjured controls. If these findings can be duplicated using larger sample sizes, this novel biomarker may be used as a predictor of injury severity and outcome.