METHODS

Experts in medicine, engineering, crash investigation, injury surveillance, and epidemiology generated a comprehensive listing of 31 data elements relevant to bicyclist injury research and prevention that could likely be reported by history. This expert group, which included the authors and experts from the United States CPSC, used the Haddon matrix⁹ to categorize the various elements identified. The Haddon matrix is a public health tool designed for the development of injury countermeasures and includes nine cells that divide the time sequence of an event into three stages (pre-event, event, and post-event) and the principal factors of the injury into three components (the agent of injury, the person, and the environment). The 31 elements identified were categorized into the nine Haddon matrix cells as demonstrated in table 1.

A retrospective chart review was conducted on emergency department records of children ages 5–18 years who sought care for bicycle related injuries in the emergency department of an urban level one pediatric trauma center between March 1995 and July 1998. This center serves as the primary source of emergency medical care for children of west and southwest Philadelphia, a predominantly poor, minority population. In addition, it is a major referral center for children from southeastern Pennsylvania and southern New Jersey. Cases were identified through an existing emergency department based injury surveillance system that utilized surveys of pre-hospital care providers and children to obtain information regarding the circumstances of injury.¹⁰ Participants were enrolled in accordance with a study protocol approved by the Institutional Review Board of the hospital.

Components of the emergency department based medical record that were reviewed included all physician and nursing notes, triage sheets, and, if applicable to the case, semistructured trauma team evaluations and inter-hospital transport team records. The semistructured evaluation forms include prompts for information such as previous medical history, pre-hospital care information, mechanism of injury, and exact location of injury. For each emergency department medical record, the documentation was reviewed for information on the 31 elements and the presence or absence of each element was determined. In order to ensure consistency in determining the documentation of elements, one clinician/researcher (AD) abstracted all of the records. Another researcher with experience in medical record coding (EM) independently reviewed 20% of the records in order to ensure the reliability of data abstraction.

The proportion (with 95% confidence intervals) of medical records that had documented information was determined for each of the 31 data elements. Analyses were conducted on an individual element level and as grouped in each of the cells of the Haddon matrix. A consecutive sample of at least 260 cases was necessary to estimate 95% confidence intervals of ± 3% to 6% on point estimates of data completion ranging from 10%–50%.

In addition, medical records of seriously injured children for whom the trauma team was alerted (“trauma-stat”) were compared to records of children who were not trauma-stat patients. Differences in documentation patterns were assessed using the χ² or exact test as indicated.

RESULTS

Altogether 278 cases of injured child bicyclists were identified from March 1995 to July 1998. The study sample was identified from a cohort of injured pediatric bicyclists who were 74% male and 74% African-American. The study sample included a range of injury severity with over 25% of children having sustained clinically significant injuries (maximum abbreviated injury scale score ≥2) and 29% were admitted to the hospital from the emergency department. A total of 268 of 278 records assessed had all components of the emergency department medical record [triage sheet, nurse notes, physician notes, trauma evaluation sheets (for trauma-stat patients only), and interhospital transfer records (if child was transferred from another hospital)] available for review. The assessment of inter-rater reliability of the medical record abstraction revealed excellent overall agreement (88% agreement, κ= 0.71, p<0.0001). The majority of 31 elements (96%) had >80% agreement as to whether the element was documented or not. Only 24 of the 31 categories had sufficient variability to calculate the κ statistic. Of those elements with sufficient variability, 63% had a κ indicating moderate or better agreement (κ≥0.4). All of those elements with κ<0.4, however did have >80% agreement.

Element level analysis

The proportion of medical records containing each of the 31 targeted data elements is presented in table 2. Elements most commonly documented (present in over 80% of records) included: previous medical history, diagnosis, indication as to whether pre-hospital care was given, and child traumatic contact points (documentation of minor injuries—for example, an abdominal wall contusion, that may provide insight into the mechanism of more serious injuries—for example, a handlebar related liver laceration). Of note, several elements that would be considered central to effective injury surveillance were not routinely documented. These included identification of motor vehicle/object involved in crash (58%), the precipitating event (24%), the location of the crash (23%), and documentation of helmet use (23%). Several of the elements were not present in any records at all (risk taking behavior of child, conspicuity of child, conspicuity of bike, fit of bike for child, condition of the road, and surface the child landed on after the crash).

Haddon cell groupings

Table 3 further presents the per cent completion rates of data elements grouped by Haddon matrix cell. Overall, the Haddon matrix cells with the highest completion rates were the post-event/child information (56%) and the post-event/environmental information (54%). Pre-event/child information was present 22% of the time. Cells pertaining to bicycle characteristics were not well documented (≤3%) in any time phase of the matrix.

Trauma-stat patients

Records of trauma-stat patients (n= 41) demonstrated significantly better (p<0.05) completion rates for documentation of pre-hospital care, description of pre-hospital care and child traumatic contact points. Records of trauma-stat patients demonstrated significantly worse (p<0.05) completion rates for documentation of previous medical history, child kinematics, main body parts impacted, and exact address of the incident (table 4). All other data elements had similar documentation rates between the trauma-stat and non-trauma-stat groups.

DISCUSSION

While the emergency department is uniquely positioned as a source of injury surveillance data, this study demonstrated that medical records routinely completed during the care of children after bicyclist trauma frequently lacked information that is important to injury surveillance and prevention efforts. Because the primary purpose of the medical record is to document care of the patient, documentation of more medically relevant information was generally better than documentation of information regarding the location and circumstances of the event.

Previous emergency department based injury surveillance systems have relied on questionnaires or surveys in addition to the medical record to gather this type of data. However, these systems have been plagued by low inclusion rates, staff objections to increased workload, and bias in inclusion rates based on injury severity.^11,12 These limitations commonly lead to a lack of sustainability in these surveillance systems.¹³ Use of the medical record as the primary source of surveillance data collection would alleviate many of these limitations and more effectively integrate data collection into the routine workings of the ED.

The documentation of selected trauma history elements improved with the use of the semistructured trauma flow sheet. One hypothesized reason for the increased frequency of some elements is that there are prompts posed on these semistructured forms that probe physicians and nurses to include this information on the record. For example, items such as motor vehicle involvement, bicycle helmet use, pre-hospital care, and child traumatic contact points are prompted for in the “trauma-stat” flow sheet. A previous study has shown formatted histories for emergency department visits were associated with higher physician satisfaction with the documentation process.¹⁴ A similar study of ambulatory care providers found that the structured medical record form improved provider performance and recording of the process of care.¹⁵ The decreased documentation of other elements may be inherent to the trauma-stat patient population. Often, parents of these more seriously injured children are not immediately available to provide a previous medical history or exact address of the incident. This possibility points to the need to link emergency department records of seriously injured children to inpatient medical data or trauma registries to complete the picture of the incident.

The data for this study were gathered by retrospective chart review, which limits our conclusions to the availability of written documentation. We do not know if a physician or nurse asked for the information but did not document it on the medical chart. However, injury surveillance and research is often done from chart abstraction and therefore any information asked but not documented is realistically inaccessible.

Previous research has demonstrated that injury circumstance information can be collected in an emergency department if its importance is highlighted. The Canadian Hospitals Injury Reporting and Prevention Program (CHIRPP) was developed as a national injury surveillance system in order to assist in the identification of emerging hazards as well as provide information that is useful in developing injury prevention programs. Emergency department staff use specialized data collection forms to obtain information about the injury event. The emergency department medical record is also used to provide information for the surveillance system. Studies have used these data to identify injury risks and the data have also been used to assist in the development of prevention programs.¹⁶

Injury information collected in the emergency department is important not only to identify injury trends and hazards but also for designing and evaluating injury prevention programs. Information about the location of the crash, for example, could help community program developers to appropriately target injury prevention programs. In order to develop programs and evaluate their effectiveness in the community, a sustainable and reproducible system of routinely collected and standardized data on injury circumstances and modifiable factors leading to injury must be created. The use of a semistructured emergency department record could provide this important injury surveillance information for patients with a range of injuries from the most severe trauma patients to the patients with minor injuries. Information on the entire spectrum of injury would be valuable to learn more regarding how the circumstances of an injury event impact the resulting severity of the injury.

Future work should compare documentation on a semistructured emergency department medical record with a free form emergency department medical record on the entire spectrum of emergency department treated injuries in order to determine the ability of the medical record to serve the dual purpose of documentation of clinical care and collection of injury surveillance data. In addition, future studies should assess the collection of data relevant to other types of injury and in other hospitals, both trauma and non-trauma centers.