924
Newgard et al. • CRASH SCENE PHOTOGRAPHY
Crash Scene Photography in Motor Vehicle Crashes without Air Bag Deployment Craig D. Newgard, MD, MPH, Katherine A. Martens, MD, Evelyn M. Lyons, RN, MPH Abstract Objective: To determine whether vehicle characteristics, measured using crash scene photography, are associated with anatomic patterns of injury and severity of injury sustained in motor vehicle crashes (MVCs) without air bag deployment. Methods: A prospective observational study was conducted over 22 months, using 12 fire departments serving two hospitals. Two vehicle photographs (exterior and interior) were taken at each MVC. Vehicular variables were assigned by grading the photographs with a standardized scoring system, and outcome information on each patient was collected by chart review. Results: Five hundred fifty-nine patients were entered into the study. Frontal crashes and increasing passenger space intrusion (PSI) were associated with head, facial, and lower-extremity injuries, while rear
crashes were associated with spinal injuries. Restraint use had a protective effect in head, facial, and upper and lower extremity injuries, yet was associated with higher odds of spinal injury. Lack of restraint use, increasing PSI, and steering wheel deformity were associated with an increased hospital length of stay and hospital charges, yet only steering wheel deformity was associated with increasing injury severity when adjusting for other crash variables. Conclusions: Out-of-hospital variables, as obtained from crash vehicle photography, are associated with injury site, injury severity, hospital length of stay, and hospital charges in patients involved in MVCs without air bag deployment. Key words: crash photography; motor vehicle; trauma. ACADEMIC EMERGENCY MEDICINE 2002; 9:924–929.
Mechanism of injury is recognized as a potential predictor of injury in motor vehicle trauma (MVT).1 While associations between vehicular damage and injury severity2–4 and between impact site and injury site1,5 have been demonstrated, a method of accurately recording predictive variables in a timely manner by on-scene personnel has not been developed.2,6,7 On-scene photographic documentation of damage from motor vehicle crashes (MVCs) has been suggested as a potentially effective means
of obtaining accurate and timely out-of-hospital information.6,8 To date, no study has linked injury pattern and injury severity to data from crash scene photographs. In this study, we assess crash vehicle photographs taken by out-of-hospital personnel for site and severity of impact, passenger space intrusion (PSI), steering wheel deformity, and windshield condition in vehicles without air bag deployment. These vehicle characteristics were examined to determine their relationship to anatomic injury patterns, severity of injury, length of hospital stay, and hospital charges. We hypothesized that such photographs could provide an objective tool that could be used to help predict injury patterns and injury severity in MVT patients.
From the Department of Emergency Medicine, Harbor–UCLA Medical Center, Torrance, CA (CDN); the Division of Emergency Medicine, Loyola University Medical Center, Maywood, IL (CDN, KAM, EML); and EMS & Highway Safety, Illinois Department of Public Health, Maywood, IL (EML). Dr. Newgard is currently in the Department of Emergency Medicine, Oregon Health & Science University, Portland, OR. Received January 7, 2002; revision received April 23, 2002; accepted April 29, 2002. Presented at the annual meeting of the Society for Academic Emergency Medicine, Denver, CO, May 1996, and the annual meeting of the Illinois College of Emergency Physicians, Chicago, IL, June 1996. Supported in part through funding from the Emergency Medicine Education & Research Group (EME&RG), grant number F32 HS00148 from the Agency for Healthcare Research and Quality, and the SAEM Research Training Grant. Polaroid provided the cameras used in the study. Address for correspondence and reprints: Craig D. Newgard, MD, MPH, Department of Emergency Medicine, Oregon Health & Science University, 3181 S.W. Sam Jackson Park Road, Mailcode CR-114, Portland, OR 97201-3098. Fax: 503-494-4997; email:
[email protected].
METHODS Study Design. This was a prospective, observational study. The study was approved by our hospital institution review board (IRB), who waived the requirement for informed consent. Study Setting and Population. The population consisted of mixed urban/suburban communities served by 12 fire departments in the western suburbs of Chicago transporting patients to one of two participating hospitals, a Level I trauma center in a
ACAD EMERG MED • September 2002, Vol. 9, No. 9 • www.aemj.org
university hospital setting and a Level II trauma center in a community hospital setting. Study Protocol. Specialized, durable instant cameras produced by Polaroid were distributed to 12 fire departments. Members of each department received instruction on camera use and appropriate documentation of on-scene variables prior to implementation of the study. Emergency medical services (EMS) personnel were instructed to take one photograph of the primary point of external vehicular impact and a second photograph of the occupant compartment at the scene of passenger vehicle crashes when patients were transported to participating hospitals. If there was any question of compromising patient care or transport time, the photographs were not taken unless backup firefighters or police were available. The photographs were collected over a 22-month period (April 1993–January 1995) for patients taken to one of the two participating hospital emergency departments (EDs). Measurements. The photographs were scored as to primary site of impact (18 options) and severity of impact (mild, moderate, or severe) based on a standardized grading tool produced by the National Safety Council.9 The pictures were also graded for PSI, steering wheel deformity, and windshield integrity. PSI was defined as any encroachment of vehicle deformity into the passenger compartment, regardless of proximity to the occupant. Minor compartment intrusion was defined as less than 18 inches and major intrusion was greater than or equal to 18 inches of encroachment, as assessed from the photographs. Vehicle structures were used as points of reference to estimate the extent of encroachment in the photographs (e.g., intrusion beyond the midway point of a given seat location, measured from the door in a lateral collision, from the dash in a frontal collision, or from the rear seatback in a rear collision, was defined as major intrusion). Steering wheel deformity was noted if any angulation or deformity was seen in the photograph. A nonintact windshield was defined as any windshield encompassing a crack, spidering, or blowout. Each photo was independently graded by two of four researchers [two emergency medicine (EM) attendings, an EM registered nurse, and a medical student/emergency medical technician]. A third researcher arbitrated any significant discrepancies in the observations. Kappa scores were calculated to assess interrater agreement. Because there was a very low prevalence of air bag deployment and because injury patterns may differ with air bag deployment, vehicles with air bag deployment were excluded from the analysis.
925 We conducted a chart review to determine patient demographic information, Abbreviated Injury Scale (AIS), Injury Severity Score (ISS), length of hospital stay, and total hospital charges. A single researcher, blinded to the accompanying photograph, assigned AIS codes and the resulting ISS. Data Analysis. To assess the relationship between out-of-hospital variables and anatomic injury, we used logistic regression models. Seven different models (one for each type of anatomic injury by AIS code: head, face, spine, thorax, abdomen, upper extremity, and lower extremity) were analyzed using five variables obtained from the photographs (type and severity of external vehicle damage, PSI, steering wheel deformity, and windshield damage) and three variables from out-of-hospital providers (restraint use, seat location, and patient age). Variable selection was performed using a backward stepwise selection process and a p < 0.15 for significance to allow for improved model fit and for an appropriate number of predictor variables used to model a given outcome. Five predictor variables were included in the final models (crash type, PSI, restraint use, seat location, and patient age) for all injury types. Due to the high prevalence of certain anatomic injury types in certain types of crashes, crash type was modeled as a specific type of collision versus all other types. Each type of collision (i.e., frontal, lateral, or rear) was tested in each model for association with the anatomic injury and for model fit. In vehicles with more than one impact site, the site with the greatest amount of external damage was considered the main impact site for purposes of the analysis. The 18 regions of external damage originally used to grade the photographs were categorized into three regions for the analysis: frontal (front-center, front-distributed, front-left, and frontright), rear (rear-center, rear-distributed, rear-left, and rear-right), and lateral (left-back-quarter, leftdistributed, left-front-quarter, left-middle, left-top, right-back-quarter, right-distributed, right-frontquarter, right-middle, and right-top). Models using two different definitions of seat location (front versus rear seat position or drivers versus non-drivers) were compared. Injuries within a given anatomic region were analyzed without respect to injury severity due to the variable number of injuries with AIS ⱖ 2 severity by body region, and no adjustment was made for clustering of patients within vehicles. Proc LOGISTIC (SAS 8.1, SAS Institute, Cary, NC) was used to compute the models and the Hosmer–Lemeshow goodness-of-fit test was used to assess model fit.
926
Newgard et al. • CRASH SCENE PHOTOGRAPHY
We analyzed ISS, total hospital stay, and hospital charges for differences with respect to restraint use, PSI, steering wheel deformity, and windshield damage using the nonparametric methods of SAS Proc NPARLWAY (univariate analysis). The reported unadjusted p-values are products of the Wilcoxon rank sum test or the Kruskal-Wallis test, based on the number of groups being compared. Adjusted p-values were calculated using three multivariate linear regression models, each modeling a separate continuous variable as an outcome (ISS, hospital stay, and hospital charges) using seven predictors (restraint use, PSI, steering wheel deformity, windshield damage, age, seat position, and external vehicle damage). Models using two different definitions of seat location (front versus rear seat position or drivers versus non-drivers) were compared. All reported p-values are two-sided. Proc REG was used to compute the models.
RESULTS There were 559 patients included in the study. Patient and vehicle characteristics of the sample are included in Table 1. We analyzed 502 automobile photographs, and 28 photographs were used to assess more than one patient. Impact site was unable to be categorized in 18 photographs (3%), while impact severity could not be categorized in six photographs (1%), due to poor quality. These photographs were excluded from the respective analyses. Fifteen percent of patients (83/559) had at least one injury of moderate severity or greater (AIS ⱖ 2), while 2% of the patients studied (11/559) were severely injured (ISS ⱖ 16), and four patients died (0.7%). All MVT patients transported by participating fire departments to one of the two study hospitals were tracked for a six-month period during the study to monitor out-of-hospital personnel compliance. Of the 377 total MVT patients transported during this six-month period, 224 (59%) had photographs taken and were included in the study. Information on MVT patients who did not have photographs taken was unavailable for comparison. Interrater reliability scores (kappa scores) were calculated for primary vehicular impact site (18 categories) and damage severity (3 categories), as determined in the photograph. The overall kappa scores for impact site and impact severity were 0.59 and 0.52, respectively, indicating fair reliability. Kappa scores were not calculated for assessment of PSI, steering wheel deformity, or windshield damage. The association between anatomic injury and out-of-hospital variables is seen in Table 2, and the
TABLE 1. Patient and Vehicle Characteristics Variable Sample size
559
Age—median (interquartile range)
29 (21–41) yr
Gender—male
48%
Restrained
56%
Windshield damage
59%
Steering wheel deformity
23%
Drivers
71%
Front seat position
91%
Impact type Frontal Rear Lateral
49% 16% 35%
External vehicle damage None Mild Moderate Severe
3% 29% 36% 33%
Passenger space intrusion (PSI) None
