Randomized Prospective Study to Evaluate Child Abuse Documentation in the Emergency Department
Presented at the Pediatric Academic Societies/American Academy of Pediatrics Meeting, Washington, DC, May 2005.
This study was supported by a National Institutes of Child Health and Human Development (NICHD) grant to Dr. Guenther (K23HD043145).
Address for correspondence and reprints: Elisabeth Guenther, MD, MPH; e-mail: firstname.lastname@example.org.
Objectives: The objective was to determine whether an educational intervention for health care providers would result in improved documentation of cases of possible physical child abuse in children <36 months old treated in the emergency department (ED) setting.
Methods: This study had a statewide group-randomized prospective trial design. Participating EDs were randomized to one of three intervention groups: no intervention, partial intervention, or full intervention. Medical records for children <36 months of age were abstracted before, during, and after the intervention periods for specific documentation elements. The main outcome measure was the change in documentation from baseline. Generalized estimating equations (GEEs) were used to test for intervention effect.
Results: A total of 1,575 charts from 14 hospitals EDs were abstracted. Hospital and demographic characteristics were similar across intervention groups. There were 922 (59%) injury visits and 653 (41%) noninjury visits. For each specific documentation element, a GEE model gave p-values of >0.2 in independent tests, indicating no evidence of significant change in documentation after the intervention. Even among the 26 charts in which the possibility of physical abuse was noted, documentation remained variable.
Conclusions: The educational interventions studied did not improve ED documentation of cases of possible physical child abuse. The need for improved health care provider education in child abuse identification and documentation remains. Future innovative educational studies to improve recognition of abuse are warranted.
Child abuse and neglect is a serious public health problem in the United States. The U.S. Department of Health and Human Services reported that an estimated 905,000 children experienced child abuse or neglect in the United States in 2006.1 The consequences of child abuse and neglect for children and their families include increased risk for suicide, depression, substance abuse, and chronic health problems.2–7 As a result, primary and secondary prevention of child abuse are important to the long-term health of children.
Several studies have shown that children who have experienced abuse are often seen in an emergency department (ED) before the diagnosis of the abuse is made.8–12 Chang and colleagues8 estimate that in the United States, 10% of all children presenting to an ED have experienced abuse. A recent study found that children with substantiated child abuse have higher ED use prior to a diagnosis of abuse, when compared with the general pediatric population.13 These statistics suggest that the ED is one of the health care settings in which the diagnosis of child abuse should be considered.
Appropriate documentation in the medical record is important to show that child abuse was considered in the differential diagnosis of an injury. In addition, adequate documentation helps differentiate abusive from nonabusive injury in the legal realm. Unfortunately, a study by Ziegler and colleagues9 found that the staff in a general hospital often did not document consideration of child abuse in pertinent cases, nor did the staff routinely document or assess all of the indicators of child abuse in children at risk for abuse. Similarly, several studies conducted in the ED setting showed that documentation in the medical chart is inadequate to differentiate unintentional from inflicted injury.14–17
Child abuse or neglect may not be consistently diagnosed and/or reported in the medical record in the outpatient health care setting for multiple reasons, including misleading histories by the caretaker, differences in health care training of practitioners,18,19 practitioner bias,20–22 concerns for the legal implications of this diagnosis,23 and physician error in recognition and/or interpretation of presenting signs and symptoms.16,24,25 As a result, documentation of the diagnosis of child abuse in the outpatient medical record remains suboptimal. To overcome these barriers, knowledge, skills, and attitudes of physicians need to be changed.25
One way to change knowledge, skills, and attitudes is through targeted education.18,26,27 We conducted a randomized prospective study to evaluate two educational interventions designed for ED health care providers (i.e., physicians, nurses, and technicians), compared to no intervention. The two interventions were designed to increase general knowledge and understanding of physical child abuse, to improve recognition and detection of physical child abuse, and to meet current management and documentation recommendations when physical child abuse is considered in children younger than 36 months.16 The two interventions contained similar material; however, one intervention was a one-time-only intervention, while the other intervention was more comprehensive, required more time of the participants, and included additional information. The goal of the educational program was to educate health care providers regarding signs and symptoms suggestive of physical child abuse in children under 36 months, provide information regarding referrals to the Division of Child and Family Services (DCFS) if child abuse is considered, and review the requirements of documentation for suspected physical child abuse in the ED chart. We hypothesized that an ED-based educational intervention for health care providers would result in improved documentation in institutions receiving education of the consideration of the diagnosis of physical child abuse in injured children presenting in the ED setting. Furthermore, we hypothesized that an ongoing intervention would create a greater change than a single intervention.
A group-randomized prospective trial design was used to evaluate the effectiveness of two different structured educational programs designed for health care providers in the ED setting. This study was considered exempt by our university institutional review board. Separate institutional review board approval was obtained from each of the hospitals involved in this study.
Study Setting and Population
The sampling frame was each of the 39 hospitals in Utah with an ED that evaluates children. We excluded the only pediatric trauma center and children’s hospital due to a lack of comparable institutions in the state. Fourteen hospitals agreed to participate in this study. None of the participating hospitals had a hospital-based specialized team to evaluate possible child abuse and neglect; however, all 14 hospitals had access to the child abuse specialist at the tertiary pediatric hospital in the state.
Hospitals were divided into five groups, dependent on the number of annual preadolescent visits to the ED and blocked by stratum. Blocking was used to ensure equal numbers of small to large hospitals in the sample so that results could generalize to state hospitals of any size; the experimental design was not developed to statistically evaluate the effect of blocks itself. Of these five initial groups, three were randomly assigned to one of the three intervention groups. Initially, 15 hospitals were randomly selected from the sampling frame to participate in the study, with the goal of 5 hospitals in each of three intervention groups described below. If the representatives of a hospital chose not to participate in the study, the next hospital in the equivalent stratum was approached for participation. We were limited to 15 institutions by the available resources to conduct this study.
The educational program was designed for general practitioners, emergency physicians (EPs), family practice physicians, pediatricians, nurse practitioners, physician assistants, registered and licensed nurses, and technicians employed to work in the ED setting. To minimize intrinsic selection bias and assure that the attendees were not more likely to be those already interested in improving the recognition of child abuse, the educational interventions were most often presented as part of an ED staff meeting or an ED physicians’ meeting. In addition, we felt that this would maximize ED health care provider attendance. Health care providers at each of the participating hospitals were not aware of the educational arms of the study or the chart abstraction for specific documentation elements.
The three intervention groups were as follows. Intervention Group 1 consisted of no intervention (five hospitals). Participants did not receive an intervention. Intervention Group 2 consisted of single-contact intervention (five hospitals). During the intervention period, the primary investigator (PI) or a trained health care educator presented a standardized educational program to the emergency health care providers. The educational program was a 1-hour didactic session consisting of a 45-minute PowerPoint presentation on the recognition of physical child abuse in the ED and appropriate documentation and 15 minutes for case-centered discussion. There were no additional site visits. Intervention Group 3 consisted of ongoing education and chart reminder intervention (four hospitals). An extended educational program consisting of three 1-hour didactic sessions held over a 6-month period was presented to this group by the PI or the trained health care educator during the intervention period. The 1-hour didactic sessions were as follows: 1) Session 1—the 45-minute PowerPoint presentation used for Intervention Group 2 was presented. 2) Session 2—two journal articles, one on child abuse recognition22 and the second on appropriate documentation of child physical abuse,16 were presented and discussed approximately 1 to 3 months after Session 1. Case-centered discussions were also held at the second educational session. In addition, a screening instrument, designed to remind the health care provider to consider the diagnosis of physical child abuse, was presented and discussed (see Data Supplement S1 for instrument, available as supporting information in the online version of this paper). Following this presentation, the checklist was then inserted by the ED triage nurse into the chart of every patient less than 36 months of age seen in that ED with a chief complaint of injury, from the time Session 2 was conducted until the end of the intervention period. 3) Session 3—in addition to all of the interventions from Session 2, a child abuse specialist was available to answer questions posed by the group on documentation elements, child abuse recognition, and the most appropriate disposition of patients in which there is a concern for child abuse. The total intervention time was 3 to 4 months.
Instruments Two focus groups were used to develop, test, and modify the 1-hour didactic training session used for Intervention Groups 2 and 3, and the checklist used in Intervention Group 3. The educational interventions and guidelines for the evaluation tool used in the chart abstraction were constructed a priori from elements of published child abuse evaluations.16
Data Collection A review of randomly chosen medical records before, during, and after the intervention period was conducted at each participating hospital to assess the effect of the standardized educational program on the medical chart documentation of the consideration and diagnosis of physical child abuse. Time 1 chart abstraction was conducted to measure baseline documentation for each institution. Time 2 chart abstraction (midintervention) was performed to ascertain the immediate effect of education on documentation. In those programs in which no intervention occurred, charts were abstracted from dates that coincided with the time of the educational presentations at the intervention sites. Time 3 chart abstraction (postintervention) was conducted on charts from all groups to discern presence of long-term effects of the education intervention on documentation.
Baseline Documentation (Time 1). For Intervention Groups 1, 2 and 3, all medical charts of children less than 36 months of age evaluated in the ED setting from November 1, 2004, to February 28, 2005, were obtained. From these medical charts, those with discharge diagnoses described in the medical literature as most often associated with the presentation of possible physical child abuse were obtained for data abstraction (see online Data Supplement S2 for a description of the clinical indicators, available as supporting information in the online version of this paper). Due to the high volume of pediatric charts from this time period, a subset of charts from 30 randomly generated dates was then obtained for data abstraction.
Midintervention Documentation (Time 2). We abstracted this from medical charts as follows. Intervention Group 1—Medical charts for children <36 months of age from the Group 1 hospitals were derived from 30 random dates generated from October 18, 2005, to February 2, 2006. Intervention Group 2—All medical charts of ED visits for children <36 months of age were abstracted from visits on Day 1 through Day 30 following the first didactic session. Intervention Group 3—All medical charts of ED visits for children <36 months of age were abstracted from Visit Days 1 through 30 following the second didactic session (presentation and discussion of the journal articles).
Postintervention Chart Abstraction (Time 3). Data were abstracted from all medical charts of visits for children less than 36 months of age evaluated in the ED 30 days through 60 days for all three groups after the date of the final intervention for Intervention Group 3.
Two of the study authors abstracted the following data elements from each medical chart: child age, sex, ethnicity and race, date and time of arrival to the ED, chief complaint, date and time of presenting injury, type of presenting injury, history of past injuries, description of presenting injuries (including location, size, and color), genital exam results (if performed), radiograph results (if obtained), lab test results (if obtained), involvement of child protective services, disposition of the patient, and International Classification of Diseases, Ninth Revision (ICD-9), codes for discharge diagnosis. If the patient presented with a chief complaint or evidence of injury, the following specific items, noted in the literature to be vital to appropriate documentation for possible child abuse,16 were abstracted or marked as present or absent in the medical chart: 1) documentation of delay between injury and seeking medical advice of greater than 24 hours without plausible explanation, 2) documentation of consistency of history with the injury (any notation of history’s consistency with the presenting injury was recorded as present), 3) a complete description of the injury, 4) documentation of an anterior fontanelle exam in patients less than 12 months of age, 5) adequate skin exam documentation, 6) documentation of a skeletal survey for possible abuse, 7) documentation of a computed tomography (CT) of the head, and 8) documentation of a referral to either DCFS or a pediatric subspecialist to investigate the possibility of abuse.
For analysis of documentation patterns, only Questions 1–5 were considered and were used as five separate outcome variables. Questions 6–8 were included for data abstraction to help determine the scope of the diagnostic work-up. In addition, any chart that included documentation of possible physical child abuse was extracted for further analysis to determine the documentation patterns of the ED health care provider once the diagnosis of child abuse was considered.
Our goal was to determine if ED practice changed, not to find out whether individual physician’s behavior changed. As a result, the unit of analysis for this group-randomized trial was the institution, not the individual; therefore, we did not measure change by practitioner. Descriptive analysis included describing the number of visits and the demographic composition (age, sex, race, and ethnicity) by intervention group. Management, consultations, and disposition were described by intervention group, as were hospital characteristics, including the mean number of beds and the number of charts abstracted. Descriptive analyses were repeated for the subset of patients documented as possible child abuse in the differential diagnosis.
Emergency department visits were categorized as injury or noninjury according to ICD-9 codes and chief complaint. The top chief complaints and diagnoses, as well as the disposition and the number of head CT scans, were described for each of the injury and noninjury groups.
A statistical model was used to test for a change in documentation due to the intervention. The outcome of interest was a change in documentation from baseline for injury visits. Generalized estimating equations (GEEs) were used to test for an intervention effect while accounting for the clustering of patients in a hospital.28 An exchangeable working correlation structure was used. Individual GEE models were fit to each of the five responses relating to the documentation of child abuse. Adequate documentation (yes or no) for each outcome was used as the dichotomous response in the GEE models. Independent variables included time (pre-, mid-, and postintervention), intervention group (none—Group 1, mid—Group 2, high—Group 3), and a time–intervention interaction. A significant interaction term would suggest an effect of the intervention compared to baseline.
Frequency of documentation for each intervention group and time period was described for each of the five questions used in GEE models. Exact 95% confidence intervals (CI) for the percentage of documentation were also calculated for each intervention group separately for the five questions. No a priori power analysis was performed as we were limited to 15 institutions by the available resources to conduct this study.
Data were abstracted from 1,575 charts for visits of children less than 36 months of age evaluated in 1 of the 14 hospital EDs participating in this study. Table 1 shows that total visit numbers were similar across intervention groups. Of the 1,575 charts, 34.5% (n = 544) were preintervention visits (Time 1), 31% (n = 488) were midintervention visits (Time 2), and 34.5% (n = 543) were postintervention visits (Time 3). The median patient age was 18 months (interquartile range = 11, 26) and the majority (78.4%) were white, reflecting the general population of this state (89.2% white).29
Table 1. Hospital Characteristics and Demographics Abstracted from the Medical Charts of Children <36 Months Evaluated in the ED Setting over the Study Period (N = 1,575)*
|Total number of charts abstracted during the study period, n (%)||559 (35.55)||473 (30)||543(36)|
|Number of teaching hospitals in each group||1/5||1/5||1/4|
|Number of hospitals with 100% of total physician attendance at educational intervention||N/A||2/5||0/4|
|Mean number of total hospital beds||163||144||201|
|Patient Demographics|| No Intervention (n = 559)||Single-contact Intervention (n = 473)||Ongoing Intervention (n = 543)|
|Mean age, months (SD)||18.8 (9.4)||18.4 (9.2)||18.2 (9.4)|
|Male, n (%)||320 (57.2)||284 (60)||288 (53)|
|Race, n (%)|
| White||328 (79)||326 (78)||405 (78)|
| Other||87 (21)||91 (22)||113 (22)|
|Ethnicity, n (%)|
| Hispanic||63 (15)||77 (18)||92 (17)|
| Non-Hispanic||365 (85)||352 (82)||444 (83)|
|Trauma-related visit, n (%)||324 (58)||290 (61)||308 (57)|
|Management, n (%)|
| Head CT performed ||31 (6)||35 (7)||51 (9)|
| Skeletal survey performed ||1 (<1)||1 (<1)||1 (<1)|
|Subspecialty consult, n (%)|
| DCFS||2 (<1)||4 (1)||5 (1)|
| Police||2 (<1)||3 (1)||1 (<1)|
| Social work||1 (<1)||0 (0)||0 (0)|
| Pediatrics||32 (6)||21 (4)||28 (5)|
| Orthopedics||11 (2)||9 (2)||3 (1)|
| Other||0 (0)||7 (2)||4 (1)|
| None||492 (88)||413 (87)||478 (88)|
|Disposition, n (%)|
| Discharge||539 (96)||461 (98)||534 (98)|
| DCFS||1 (<1)||0 (0)||1 (<1)|
| Transfer to another institution||13 (2)||11 (2)||3 (1)|
Type of ED Visits, by Visit Type
When grouped by visit category, there were 922 (58.5%) visits for injury-related complaints and 653 (41.4%) for noninjury complaints among abstracted charts. The three primary chief complaints for injury-related visits were laceration, facial injury, and extremity fracture. The top three noninjury chief complaints were vomiting, fever, and diarrhea. The leading three discharge diagnostic codes for injury-related visits were ICD-9 codes 920 (contusion of face, scalp, and neck), 959.01 (injury, other, and unspecified: head and neck), and 873.42 (other open wound of head). The three discharge diagnoses coded most often for non–injury-related visits were ICD-9 codes 787.03 (symptoms involving digestive system: nausea and vomiting), 382.9 (suppurative and unspecified otitis media), and 780.31 (general symptoms: convulsion, febrile convulsions).
Results of GEE Models
For each of the five documentation questions, a GEE model was performed based on the presence of predetermined outcome documentation measures. Data were too sparse for a GEE model to converge using the first outcome; explanation of the delay between injury, and seeking medical advice. Delays were documented in only 55 visits, and explanations were documented in 22 of these visits. All of the remaining four questions had p-values of >0.2 in independent tests, indicating no evidence of significance or trend.
Table 2 shows the frequency of documentation for each intervention group and time period. The number of visits with a delay between injury and seeking medical advice was small overall (n = 55), and documentation of an explanation for this delay was present in only 22 charts. Documentation of a history consistent with the injury and description of the injury, Questions 2 (documentation of consistency of history with the injury) and 3 (a complete description of the injury), respectively, were prevalent across all intervention groups and time points. In contrast, documentation of the anterior fontanelle and skin exams, Questions 4 (documentation of an anterior fontanelle exam in patients less than 12 months of age) and 5 (adequate skin exam documentation), respectively, were much less prevalent and showed no significant improvement over time.
Table 2. Frequency of Documentation Relating to Child Abuse
|None||Preintervention|| 1/5|| 92/93|| 87/94|| 2/18||31/94|
|Midintervention|| 6/9||102/104||104/105|| 1/12||43/104|
|Postintervention|| 5/13||104/112||108/112|| 2/21||49/112|
|% (95% CI)||44 (25, 65)|| 96 (94, 98)|| 96 (93, 98)||10 (3, 21)||40 (34, 45)|
|Single-contact||Preintervention|| 2/5|| 96/99|| 96/100|| 4/22|| 6/99|
|Midintervention|| 0/4|| 74/78|| 76/81|| 4/15||25/79|
|Postintervention|| 4/8|| 94/97|| 97/99|| 3/22||20/99|
|% (95% CI)||35 (14, 62)|| 96 (93, 98)|| 96 (93, 98)||19 (10, 31)||18 (14, 23)|
|Ongoing||Preintervention|| 1/3|| 99/101|| 98/103|| 2/19||26/103|
|Midintervention|| 1/3|| 78/83|| 78/83|| 2/16||23/83|
|% (95% CI)||36 (11, 69)|| 96 (93, 98)|| 96 (93, 98)||26 (16, 39)||26 (21, 32)|
Documentation and Management Once Child Abuse Was Considered
Documentation of the possibility of abuse was present in 26 (2%) of the medical charts during the study period (Table 3). Of the 26 charts, 24 showed that the patient presented for injury-related complaints. In this subset of 24 charts, we found marked variability in documentation as well as management of the injury. No documentation of a child abuse subspecialist consultation was found in the charts we reviewed, although documentation of the DCFS consultation occurred in 10 (37%) and the police were consulted in 4 (15%) cases. Two (7%) injured patients were discharged to DCFS custody.
Table 3. Demographics and Documentation in Records in which Child Abuse Was Considered and Documented (n = 26)*
|Mean age, months (SD)||16 (7)||18 (13)||14 (11)|
|Male, n (%)||4 (40)||5 (56)||5 (71)|
|Race, n (%)|
| White||7 (78)||6 (86)||6 (86)|
| Other||2 (22)||1 (14)||1 (14)|
|Ethnicity, n (%)|
| Hispanic||2 (22)||1 (14)||1 (14)|
| Non-Hispanic||7 (78)||6 (86)||6 (86)|
|Trauma-related visit, n (%)||9 (90)||9 (100)||6 (86)|
|Documentation, n (%)|
|Was there delay between injury and seeking medical advice without satisfactory explanation?†||n = 1|
|n = 0|
|n = 2|
|Is there commentary on whether the history given is consistent with the injury?||9 (90)||7 (78)||6 (86)|
|Is there a complete description of the injury?||9 (90)||7 (78)||6 (86)|
|Is an anterior fontanelle exam documented if the patient is less than 12 months of age? (n = number of children <12 months of age)||n = 5|
|n = 4|
|n = 3|
|Is an adequate skin exam documented?||5 (50)||5 (56)||4 (57)|
|Management, n (%)|
| Head CT performed ||0 (0)||1 (11)||4 (57)|
| Skeletal survey performed ||1 (10)||1 (11)||0 (0)|
|Subspecialty consult, n (%)|
| DCFS||2 (20)||3 (33)||5 (71)|
| Police||2 (20)||1 (11)||1 (14)|
| Social work||1 (10)||0 (0)||0 (0)|
| Pediatrics||3 (30)||1 (11)||0 (0)|
| Orthopedics||0 (0)||1 (11)||0 (0)|
| Other||0 (0)||2 (22)||0 (0)|
| None||4 (40)||4 (44)||2 (29)|
|Disposition n (%)|
| Discharge||8 (80)||7 (78)||6 (86)|
| DCFS||1 (10)||0 (0)||1 (14)|
| Transfer to another institution||1 (10)||2 (22)||0 (0)|
This multicenter randomized study of an educational intervention to improve documentation of possible child abuse in the ED setting has two main findings. First, neither of our educational interventions appeared to significantly affect ED health care provider documentation behavior. Second, even when consideration of possible child abuse was documented in the chart, compliance with specific documentation recommendations remained sporadic.
Our first main finding is that a comprehensive educational intervention for health care providers in the ED setting did not appear to significantly affect EP documentation behavior. The diagnosis of child abuse is challenging for health care providers. Children less than 36 months of age present to the ED with a wide range of signs and symptoms for conditions, a variety of which are ultimately thought to be due to child abuse and neglect.30 Many health care providers in the outpatient setting, however, lack comfort in the diagnosis and management of possible child abuse.20,31–33 One consequence of this lack of comfort in the diagnosis and management of possible child abuse has been variable related documentation and reporting patterns.15–17,34 We specifically designed our intervention to address these concerns and concentrated our efforts on the diagnosis and documentation of possible child abuse. Similar efforts at physician education have been successful in areas of outpatient care including asthma treatment35 and management of non–ST-segment elevation myocardial infarction.36 In contrast to our study, however, the asthma education program curriculum is delivered in two 2½-hour interactive sessions with mandatory physician attendance,35 while the study on non–ST-segment elevation myocardial infarction management was performed as a quality improvement initiative.36 Our study did not allow for mandatory physician attendance, which may have diluted the efforts we were trying to achieve. In addition, the recognition of child abuse remains a more challenging area for the success of educational interventions.27,37 It is possible that mandated physician compliance with an educational program in conjunction with procedural changes and quality assurance review of pediatric trauma charts may be more successful in addressing the problem of child abuse recognition than relying solely on optional health care provider education.
Our second main finding was that even when consideration of possible child abuse was documented in the medical record, compliance with specific documentation recommendations16,38 remained sporadic. Although our educational interventions stressed best documentation practices, including the importance of historical factors and unique pertinent physical findings, consistent documentation of these factors was variable. In some areas, such as the likelihood of the child’s injury resulting from the mechanism stated by the caregiver, baseline documentation was found in the medical chart. Documentation of the presence or absence of bruising on a skin exam, however, remained poor. When the diagnosis of possible child abuse is considered but not adequately documented, crucial medical information is lost, including pertinent historical items, physical examination findings, supplemental diagnostic procedures, specific wording of the diagnostic impression including use of terminology such as “rule-out abuse,” and even final disposition the child.15–17,39 This may ultimately result in increased childhood morbidity or mortality due to child abuse.22,40–43
Physician attendance at the educational programs was variable between hospitals. ED health care providers work unusual hours, including many evenings, holidays, and weekends. Thus, one of the great challenges in providing education to health care providers in the ED setting is the difficulty in finding a forum in which such an educational program can be presented to a majority of the ED staff. Although we were able to present to the physicians at a mandatory physician meeting at two hospitals participating in this study, in the other hospitals there was a mix of health care providers, but primarily nurses and technicians. In addition, there was no guarantee that the same individuals for Group 3 were present at each of the sessions. This underscores the difficulty of traditional educational interventions, such as lectures and presentations, targeting EPs who have variable work schedules and hours of availability. Finally, there was no obligation or pressure (i.e., continuing medical education [CME] credit or quality assurance/credentialing requirements) for the health care providers to attend these sessions. Thus, innovative educational modalities, such as Web-based educational experiences for CME credit or credentialing, may prove to be a more effective method to reach ED health care providers.
Although we were able to abstract over 1,500 charts, the design of the study required the analysis to be performed at the institution level, limiting the true sample size to the 14 institutions and ultimately limiting the ability of GEE models to detect significant improvements in documentation, even though GEE models have been shown to give liberal results for small sample sizes.44 GEE models did not converge for Question 1 due to sparse data. For outcomes with very low (Question 4) or very high (Questions 2 and 3) rates of documentation, a much larger sample size would be required to detect significant improvement. The modest number of visits and small number of institutions likely limited the power to detect improvement in documentation for Question 5.
Child abuse is a relatively rare diagnosis made in the ED setting, which again may have affected our ability to detect a small but significant difference. However, when the analysis was additionally performed at the chart level, it still did not show an effect.
Finally, our outcome was a change in documentation. This may not be an accurate reflection of the thought process of the health care provider involved or the health care providers reached by this educational intervention.
Our multifaceted educational intervention did not appear to be successful in improving documentation in the medical record regarding possible physical child abuse in the ED setting. The need for improved and ongoing health care provider education in the area of child abuse identification and documentation remains. This study suggests that a traditional educational model alone may not be the best mode for transmitting the pertinent information. Future studies based in innovative educational interventions may be indicated.
The authors thank Dr. Bruce Herman for his assistance with development of the didactic portion of the educational intervention for this study and Drs. Lori Frasier and Bruce Herman for their expertise in child abuse. The authors are also indebted to Kathleen Merkely, PNP, for her tireless hours of assisting with the educational aspects of this study. Finally, the authors thank Dr. Brian Johnston of the Harborview Medical Center for sharing with us the checklist used to help raise awareness of possible child abuse.