Pilot intervention to improve the documentation of pediatric injuries in the emergency department

EM Advances

CJEM 2000;2(4):252-257

Abstract

Résumé

Introduction

"Surveillance" refers to the systematic collection and evaluation of disease-related data in an effort to track trends, evaluate countermeasures and devise strategies for control.¹ In Canada, largely because of poor emergency department (ED) information systems, systematically collected injury data has historically been limited to fatalities and inpatient hospitalizations, although some EDs have attempted to improve injury surveillance. For example, selected EDs participate in the Canadian Hospitals Injury Reporting and Prevention Program (CHIRPP), which gathers detailed injury data. Unfortunately, existing data is generally not population based,² and many clinically important injuries remain unreported.^3,4

The following example illustrates the value of ED injury data. In 1999, the American Academy of Pediatrics (AAP) recommended a ban on home trampolines.⁵ This was based primarily on ED-derived data,^6–11 which showed that at the time of injury trampoline safety regulations were generally not in effect (e.g., no spotters and multiple users) and that many trampoline injuries caused long-term activity limitation despite having required only outpatient treatment (e.g., surgery, casting). Had the AAP considered only hospitalization and mortality data, this information would have been unavailable.

In the Edmonton region, there are over 300,000 ED patient encounters annually, and 30% of these are associated with injury. Since April 1995, all ED encounters have been recorded in an electronic database. Medical record nosologists assign ICD-9 external cause of injury (E) codes, so that injury rates can be approximated. But E codes alone are not sensitive enough to track injury trends; nor do they provide specific information about injury circumstances, particularly in the area of sports and recreation.^12,13 To obtain important data regarding injury circumstances, specific cause, setting and geographic location, it is often necessary to interview patients or review charts.

Our hypothesis was that by teaching ED staff about the importance of injury surveillance and instructing them on what information should be recorded during the patient visit, we could improve the documentation of critical data elements on ED charts. Ultimately, enhanced ED data could be used to determine trends, plan injury reduction strategies and evaluate interventions. Our objective was to determine the effectiveness of an intervention designed to improve ED documentation of pediatric injuries.

Methods

Setting
The Capital Health Region includes the greater Edmonton area, a population of over 800,000. Within this region, 6 hospital-based EDs provide the bulk of emergency care. The study intervention was conducted at an urban community teaching hospital and trauma centre that handles approximately 80,000 patients per year.

Intervention
The intervention consisted of injury surveillance training, pocket reminder cards and visible "flagging" of injury charts. In May 1997, 2 investigators (G.E.C. and C.P.) provided all ED physicians and nursing staff with specific injury surveillance training. Staff learned that injury is the leading cause of death and disability for people under 45 years and that comprehensive injury surveillance data can be used to design and evaluate measures to reduce injury frequency and severity. The concept of the injury pyramid was introduced, and the contribution of ED data to the pyramid was discussed.^14,15 To improve charting, staff were then introduced to an injury documentation manual (available from the authors upon request). To minimize the burden on staff, no additional data forms were introduced; however, existing ED charts were modified by adding labels and space for injury-related data. (Table 1 summarizes the specific data elements under study.) Each physician was given a pocket card specifying the injury data to be charted and, finally, triage clerks were asked to flag the chart of each pediatric injury patient with a "Children's Health Program" symbol to prompt clinical staff. Data elements 1 through 10 (Table 1) were considered key items that should appear on all charts. Items 11 through 14 were only required if relevant to the specific injury. To reduce the Hawthorne effect,¹⁶ staff were not informed that their charting practices would be audited.

Patients
Patients under 17 years of age were eligible for study inclusion if they were treated during the study periods and had an ICD-9 E code in their electronic chart.

Data collection
Six hundred and forty-five "pre-intervention" charts were selected from a 3-month period (June through August) in 1996. Random sampling was based on the terminal digit of the chart number. The same process was used to select 321 post-intervention charts (flagged with the "Children's Health Program" symbol) from the corresponding months in 1997. During the selection process, it became apparent that not all pediatric injury charts had been flagged by a triage clerk prior to treatment; therefore, to assess the presence of selection bias, a further random sample of 323 unflagged, post-intervention charts was reviewed. The sample size of 321 per cell provided 90% statistical power to detect a 10% improvement in documentation,¹⁷ which we felt would be clinically important.

An experienced medical record nosologist not connected to the intervention ED reviewed all charts. Data extraction reliability of the nosologist was established by one of the authors (K.B.) in pilot work prior to the main study (Kim Borden, unpublished data, 2000).

Analysis
Patient characteristics were described using means for continuous variables and percentages for categorical data. Chi-squared analysis was used to determine the statistical significance of observed differences in categorical outcomes, while one-way ANOVA was used to compare age variations among groups and pre- versus post-intervention charting differences. Scheffe post-hoc analysis was used to determine the significance of specific group differences. Logistic regression analysis was used to calculate the odds that each specific data element would appear on the post-intervention chart. In the regression model, the intervention (i.e., pre vs. post) was the independent variable and the appearance of each specified data element on the chart (yes/no) was the dependent variable. Where elements appeared on the chart with an aggregate frequency >95%, no regression analysis was conducted. Group imbalances on age and diagnosis were controlled for in the regression analysis. All analyses were conducted using SPSS statistical software, and statistical significance was established at p < 0.05.

Results

A total of 2,517 pediatric injury charts were available for review during the 2 study periods. Of these approximately 75% were flagged as per protocol. Table 2 summarizes patient characteristics for the study group. It shows that selected and non-selected charts were similar, based on patient age, sex and injury pattern, suggesting that our study group is representative of all pediatric injury patients seen at our hospital. Table 2 also shows that post-intervention patients with flagged charts were significantly younger than pre-intervention patients and patients with unflagged charts (p < 0.001). In addition, the "flagged" group had a significantly higher percentage of lacerations than other study groups (p < 0.05). There were no significant group differences by time or day of presentation.

Table 3 shows the proportions of the pre-defined injury data elements that were recorded on pre- and post-intervention charts. For the key items (1 through 10), the mean number of documented data elements was 6.9 for control charts, 7.3 for unflagged post-intervention charts and 8.1 for flagged post-intervention charts (p < 0.001). All groups were significantly different from each other (p < 0.01; Scheffe post-hoc test). Eight percent of pre-intervention charts, 11% of unflagged post-intervention charts, and 28% of flagged post-intervention charts had all 10 key data elements recorded.

Table 3 also summarizes logistic regression data, estimating the odds (adjusted for age and diagnosis) that specified data elements would be documented on the post- versus pre-intervention charts. This table shows that there was significant improvement for 6 items (time of injury, activity at time of injury, location of injury [e.g., home, school], address where injury occurred, presence of an adult observer, and relevant environmental conditions), but significant deterioration for 2 items (possible prevention measures, and misuse or absence of sports equipment).

Discussion

Our data indicate that the study intervention, consisting of education, chart modification and visual flagging of injury-related charts, led to a significant improvement in injury documentation without the need to introduce new forms or data sheets.

However, although we improved documentation, we found that triage clerks failed to consistently flag injury charts, and we believe this reduced the intervention's effectiveness. Our data show that triage clerks were more likely to flag the charts of younger patients and patients with visible injuries (e.g., lacerations). This inconsistency set up a natural experiment, which allowed us to compare flagged versus unflagged charts in the post-intervention phase. Table 3 shows that documentation was substantially better on flagged charts, with higher post-intervention odds ratios and greater percentage improvements. Table 3 also shows that the items for which documentation failed to improve in the post-intervention phase were already well documented in the pre-intervention phase, suggesting little potential for improvement.

Of note, an average of 6.9 (of 10) key items appeared on pre-intervention patient charts. This increased to 7.3 items on unflagged and 8.2 items on flagged post-intervention charts. Similarly, only 8% of pre-intervention charts had all 10 key items documented, but this increased to 11% for unflagged and 28% for flagged post-intervention charts. These data demonstrate that our simple intervention can enhance charting quality and that visual cues (flagging) are probably the most important component of the intervention.

It is interesting that the documentation of "prevention measures" (item 11) was substantially worse after the intervention. This may be a chance effect or it could have occurred because the enhanced injury data collection process created a time constraint that led clinicians to omit a "touchy" subject (i.e., what the caregiver could have done to prevent or mitigate the injury). Certainly, in a busy ED, there is a ceiling on the amount of time clinicians can spend with a patient and the amount of information they can be expected to gather.

ED injury documentation is problematic, and there is no established minimal data set.^2,18–20 Consequently, injury data are collected with variable success. In one child injury surveillance system, Beattie²¹ reported that 65% of charts had complete data but did not itemize the data points studied. Christopher and colleagues²² evaluated the documentation of 15 injury variables on 669 ED pediatric injury charts. In this study, the mean number of variables documented was 6.2 (8 of 15 was defined as acceptable), and only 1 chart documented all 15. Schwartz and coworkers²⁰ reviewed 109 charts and found that only 46% of these had cause-of-injury information recorded.

Others have studied various interventions to improve injury charting. Pre-formatted charts for ED data collection have shown promise,^23–25 particularly to improve patient history documentation. In 1 study, pre-formatted charts increased the proportion of charts judged as "complete" from 60% to 78%.²³ In another study, complaint-specific charts for laceration and closed head injury failed to improve the documentation of injury mechanism but did improve other clinical documentation.²⁴ In 1994, Wallace and colleagues²⁵ reported that the use of a specialized form improved ED head injury documentation substantially.

Limitations and future questions

This study was conducted at a single institution, and it is possible that physicians in other settings might respond differently to a similar intervention. Because our controls were historical rather than concurrent (before­after design), it is conceivable that other factors (e.g., public injury awareness programs) could also have affected our outcome of interest; however, the authors know of no such confounding factors. A significant limitation is that we measured the outcome of the intervention at one point in time, therefore we cannot say whether improved charting will be sustained. Finally, it is of concern that, despite the intervention, many injury charts were not flagged by triage clerks. Investigators attempting similar interventions will need to develop mechanisms to address this problem and should emphasize the key role that triage clerks have in improving chart quality.

In the future, we intend to perform data audits over longer periods to determine the optimum timing for refresher training. In addition, we will extend this intervention into the other 5 regional EDs and monitor ongoing program effectiveness.

Conclusions

A simple intervention, consisting of staff training, chart modification and visual flagging of charts, can increase the amount of relevant injury information documented by ED clinicians while imposing minimal additional burden. Efforts to improve ED charting are most likely to succeed if they include visual prompts for clinicians.

References

1. Last JM. A dictionary of epidemiology, 3rd ed. Toronto: Oxford University Press; 1995.
2. Laboratory Centre for Disease Control. Children's Hospitals Injury Research and Prevention Program (CHIRPP). Technical report #2. Ottawa: The Centre; 1991.
3. Baker SP, O'Neill B, Ginsburg MJ, Guohua LI. The injury fact book, 2nd ed. New York: Oxford University Press; 1992.
4. Deane M. Child accident data: Accessible and available? J Public Health Med 1993;15:226-8.
5. Trampolines at home, school, and recreational centers. American Academy of Pediatrics. Committee on Injury and Poison Prevention and Committee on Sports Medicine and Fitness. Pediatrics 1999;103:1053-6.
6. Woodward GA, Furnival R, Schunk JE. Trampolines revisited a review of 114 pediatric recreational trampoline injuries. Pediatrics 1992;89:849-54.
7. Larson BJ, Davis JW. Trampoline-related injuries. J Bone Joint Surg [Am] 1995;77:1174-8.
8. Hume PA, Chalmers DJ, Wilson BD. Trampoline injury in New Zealand: emergency care. Br J Sports Med 1996;30:327-30.
9. Furnival RA, Street KA, Schunk JE. Too many pediatric trampoline injuries. Pediatrics 1999;103:e57.
10. Smith GA. Injuries to children in the United States related to trampolines, 1990­1995: a national epidemic. Pediatrics 1998;101:406-12.
11. Smith GA, Shields BJ. Trampoline-related injuries to children. Arch Pediatr Adolesc Med 1998;152:694-9.
12. Williams JM, Furbee PM, Prescott JE, Paulson DJ. The emergency department log as a simple injury-surveillance tool. Ann Emerg Med 1995;25:686-91.
13. Smith GS, Langley JD. Drowning surveillance: How well do E codes identify submersion fatalities? Injury Prev 1998;4:135-9.
14. Guyer B. Injury prevention: meeting the challenge. New York: Oxford University Press; 1989.
15. Monash University Accident Research Centre. Injury research and prevention: a text. Melbourne: The Centre; 1995.
16. Campbell JP, Maxey VA, Watson WA. Hawthorne effect: implications for prehospital research. Ann Emerg Med 1995;26:590-4.
17. Cohen J. Statistical power analysis for the behavioral sciences. New Jersey: Lawrence Erlbaum Associates; 1988.
18. Langlois JA, Buechner JS, O'Connor EA, Nacar EQ, Smith GS. Improving the E coding of hospitalizations for injury: Do hospital records contain adequate documentation? Am J Public Health 1995;85:1261-5.
19. Schwartz DF. It is time to get serious about childhood injury surveillance in the United States [editorial]. Am J Prev Med 1995;11:134-5.
20. Schwartz RJ, Boisoneau D, Jacobs LM. The quality of cause-of-injury information documented on the medical record: an appeal for injury prevention. Acad Emerg Med 1995;2:98-103.
21. Beattie TF. An accident and emergency based child accident surveillance system: Is it possible? J Accid Emerg Med 1996;13:116-8.
22. Christopher NC, Anderson D, Gaertner L, Roberts D, Wasser TE. Childhood injuries and the importance of documentation in the emergency department. Pediatr Emerg Care 1995;11:52-7.
23. Humphries T, Shofer FS, Jacobson S, Coutifaris C, Stemhagen A. Preformatted charts improve the documentation in the emergency department. Ann Emerg Med 1992;21:534-40.
24. Wrenn K, Rodewald L, Lumb E, Slovis C. The use of structured, complaint-specific patient encounter forms in the emergency department. Ann Emerg Med 1993;22:805-12.
25. Wallace SA, Gullan RW, Byrne PO, Bennett J, Perez-Avila CA. Use of a pro forma for head injuries in the accident and emergency department -- the way forward. J Accid Emerg Med 1994;11:33-42.