CAEP/ACMU 2000 Scientific Abstracts: 51-53

CAEP Abstracts

CJEM 2001;2(3):169-187

051 Comparison of Recursive Partitioning and Logistic Regression Modelling in the Derivation of the Canadian C-Spine Rule.
Stiell I, Wells G, De Maio V, MacPhail I, Cass D, Clement C, Worthington J, Schull M, Reardon M, Morrison L, McKnight D, MacPhail I, Greenberg G, Dreyer J, Brison R. Clinical Epidemiology Unit, University of Ottawa, Ottawa.

OBJECTIVES: The Canadian C-Spine Rule (CCR) for cervical spine radiography was derived by use of recursive partitioning (RP). This study compared the accuracy of the CCR algorithm to a model developed by another statistical technique, logistic regression (LR).
METHODS: This secondary data analysis was based on a prospective cohort study conducted in 10 Canadian EDs and involved alert, stable adult trauma patients at risk for neck injury. Physicians completed a 20-item data form for all patients who then underwent radiography to determine the outcome, C-spine injury. Variables correlated with this outcome on univariate analysis and having kappa values >0.6 were then assessed by two multivariate statistical techniques. Chi-square RP analysis (KnowledgeSEEKER) was used for the original CCR derivation and backwards stepwise LR (SPSS) was used for this analytic study. The two models were compared by receiver operating characteristic (ROC) curve analysis. RESULTS: The Canadian C-Spine/CT Head (CCC) Study dataset contained 8,933 potential neck injury cases, including 148 (1.7%) with cervical spine injury. 31 variables demonstrated a univariate p value <0.05 and 12 had kappa values >0.6. The RP model lost no cases to missing values and contained 7 variables. The LR model lost 71 cases to missing values and contained 8 variables, including 3 not in the original CCR (high-risk mechanism, immediate neck pain, and ability to flex). Comparing the RP to the LR models, respectively: Hosmer­Lemeshow goodness-of-fit statistic 0.93 vs. 0.94, area under the ROC curve 0.87 vs. 0.91 (p < 0.01), sensitivity 100% vs. 99.3% (p = 0.31), and specificity 46.6% vs. 55.8% (p < 0.01). CONCLUSIONS: LR provides a model with better overall discrimination and only slightly lower sensitivity. Further research should determine the clinical acceptability of a decision rule based upon an LR-derived scoring system rather than an RP-based algorithm.
Key words: clinical prediction rule, methodology

052 Impact of Computerized Maps on Ambulance Dispatch Response Times in Eastern Ontario.
McLelland K, Jones G, Pickett W. Queen's University, Kingston, Ont.

OBJECTIVES: To study the effect of computerized mapping on ambulance dispatch response times.
METHODS: Response times for all priority 4 calls taken by the Quinte Thousand Island Central Ambulance Communication Centre were recorded for July through September 1998 (baseline year; no computerized mapping system in place) and then compared to the same measures taken during July through September 1999 (intervention year; mapping system in place). Various time intervals in the dispatch process were evaluated to determine the impact of the computerized mapping. RESULTS: Between July and September 1998, 970 priority 4 (urgent/life-threatening) calls were dispatched in the absence of a computerized mapping program, whereas 1136 priority 4 calls were dispatched between July and September 1999 (computerized mapping in place). There was no appreciable difference in the median T1­T2 time (call received until ambulance notified) between the two years. There was, however, a significant decrease in the percentage of T1­T2 times greater than 75 seconds (13.8% in 1998 and 9.8% in 1999; chi-square (1df): 6.76; p = 0.009). There was no difference in the median T0­T4 time (call connect until ambulance arrival on scene). CONCLUSIONS: While there was some suggestion that the proportion of excessively long dispatch response times was reduced due to the new mapping system, we were unable to determine a difference in the overall response times between study years. We are currently investigating whether or not factors such as adequacy of initial training, physical location of map terminal, compliance with map use and increased call volume may have masked potential benefits of computerized maps.
Key words: emergency medical services, dispatch, response time

053 Retropharyngeal Abscess: A Ten Year Experience.
Murray H, Keogh K. Queen's University, Kingston, Ont.< P>

OBJECTIVES: Retropharyngeal abscess (RPA) is an uncommon but potentially life-threatening pediatric illness. The goal of this study was to review the experience of one academic centre over a 10 year period.
METHODS: This retrospective review took place at a university affiliated pediatric tertiary care centre. Data was abstracted from all charts with an ICD-9 diagnosis of RPA from 1989 to 1999. Descriptive statistics with 95% CIs were calculated.
RESULTS: 51 patients were identified, with a mean age of 5.6 years (95% CI 4.3­6.8). 6 patients (11.8%) had a history of pharyngeal trauma; 4 had FB penetration and 2 were postoperative. Torticollis was documented in 29 patients (56.9%) while 10 (19.6%) had evidence of stridor. Mean WBC was 21.3 (95% CI, 18.5­24.1). 39 patients (76.5%) underwent soft tissue neck x-ray examinations. 5 of these were reported as normal, 4 noted retropharyngeal gas and 30 reported a widened pre-vertebral space. 38 patients (74.5%) had CT examinations, 33 (78.6%) of which indicated a RPA. The remaining 5 showed swelling of the prevertebral soft tissues. 10 patients developed airway obstruction and required definitive airway management (9 ETT, 1 surgical airway). RPA was managed operatively in 33 patients (64.7%) and conservatively with IV antibiotics in 18 (35.3%). 15 (29.4%) patients had persistence or recurrence of RPA after treatment and 4(7.8%) required repeat operative I&D.
CONCLUSIONS: This case series reviews the epidemiology of RPA at one centre. CT scanning has become the gold standard for confirming the diagnosis of RPA. Future research will focus on the utility of soft tissue neck x-rays as a screening tool in patients with potential RPA.

CONCLUSIONS: This sub-study confirms the validity of the "Proxy Outcome Assessment Tool" to rule out important negative outcomes in neurologically intact CCC Study patients. This was an essential step in developing the Canadian CT Head Rule.
Key words: retropharyngeal abscess