Impact of child abuse clinical pathways on skeletal survey performance in high risk infants

Abstract

Objectives:

1) To examine the association between the presence of a child abuse pathway and the odds of skeletal survey performance in infants with injuries associated with high risk of abuse, 2) to determine if pathway presence decreased disparities in skeletal survey performance.

Methods:

In this retrospective study of children <1 year diagnosed with injuries associated with high risk of abuse at hospitals in the Pediatric Hospital Information System, information regarding the presence of a child abuse pathway was collected via survey. We examined whether the presence of a child abuse pathway was associated with the odds of obtaining a skeletal survey adjusting for patient level factors.

Results:

Among 2085 included cases 55% were male, 69% had public insurance and 64% were White. Fifty-eight percent presented to a hospital when a pathway was present. Skeletal surveys were performed in 86% of children between 0–5 months and 73% of children 6–11 months. In our regression model, adjusted for covariates (age, race, insurance, injury) the presence of a child abuse pathway in a hospital was associated with greater odds of skeletal survey performance (OR 1.46, 95% CI 1.02–2.08). Children with public insurance had greater odds of receiving a skeletal survey (OR 2.75, 95% CI 2.11–3.52 despite presence of pathway.

Conclusions:

When a child abuse clinical pathway was present, children with injuries associated with a high risk of abuse had a greater odds of receiving a skeletal survey. Differences in skeletal survey performance exist between infants with public vs. private insurance regardless of a pathway.

Introduction

In 2016 676,000 children were found to be victims of abuse in the United States. Children less than 1 year of age have the highest rates of serious injury and death with a victimization rate of nearly 25 per 1000.¹ Given the high number of child abuse cases frontline clinicians must accurately recognize and evaluate for abuse in a timely fashion. The American Academy of Pediatrics (AAP) and panels of experts have developed recommendations for evaluating children who may be victims of abuse.^2–4 These guidelines emphasize performing a skeletal survey (SS) in all cases of suspected physical abuse in infants.^2–4 The main purpose of the SS is to identify occult fractures. Between 11–30% of children who present with injuries associated with high risk of abuse have occult fractures identified on SS.^5–8 Despite the AAP guidelines, there remains variation in SS performance across the country in infants with concerning injuries including traumatic brain injury (TBI) and fractures.^9–12 Prior studies have shown that racial and socioeconomic disparities exist in SS performance leading to both under and over evaluation for occult fractures.^13–15 Since young victims of physical abuse frequently have occult fractures, adherence to occult fracture evaluation guidelines is crucial for identifying victims of abuse and protecting them from future harm.

Implementation of child abuse clinical pathways has been identified as a promising method to increase rates of SS performance.^16,17 Clinical pathways are defined as structured multidisciplinary care plans which detail essential steps in the care of patients with a specific clinical problem.^18,19 Clinical pathways for a variety of pediatric illnesses have been implemented in hospitals and the benefits are well described.^20,21 Several studies have found that clinical pathways improve patient care and reduce variability in outcomes.^18,21–24 Clinical pathways are known to exist in a variety of forms including publicly on the internet, embedded in the EMR, or paper versions. In a few specific settings the use of clinical pathways for children with injuries concerning for abuse has increased rates of SS performance and decreased bias based on race and socioeconomic status.^16,17

Although implementation of child abuse clinical pathways has demonstrated positive results in a few single center studies, the broader impact of clinical pathways on SS performance remains uncertain. The extent to which child abuse pathways are utilized across the country is unknown. Therefore, we aimed to 1) examine the association between the presence of a child abuse clinical pathway and the odds of SS performance in infants with injuries associated with high risk of abuse, 2) determine if pathway presence improves disparities in SS performance and 3) describe the nature of clinical pathways across U.S. children’s hospitals. We hypothesized that the presence of a child abuse clinical pathway would increase the odds of SS performance and reduce racial and socioeconomic biases in SS performance that have been previously described in the literature.

Methods

We performed a retrospective study of all infants < 1 year with a diagnosis of femur fracture, humerus fracture or TBI in 41 children’s hospitals in the Pediatric Hospital Information System (PHIS) database from 2004–2015. We included humerus and femur fractures because of the high rate of abuse in infants with such fractures^25–31 and recommendations that infants < 1 year of age with these fractures receive a SS.^2,26 We included infants with TBI because occult fractures are common in this group^32,33 and multiple groups have recommended that non-ambulatory children with head trauma undergo a SS.^4,7,16,32 We focused on children < 1 year because they are considered the highest risk population for abusive injuries and there are clear guidelines for performing a SS.^3,4 Patient-level data from the PHIS database were combined with hospital-level data regarding the presence of a child abuse clinical pathway. Information regarding the presence and characteristics of child abuse clinical pathways were obtained from a key informant identified at each hospital.

Data Collection

Data were obtained from the PHIS database, which includes demographic data, Internal Classification of Diseases, Revision 9, Clinical Modification (ICD-9-CM) diagnosis codes and radiologic tests performed at 49 non-competing children’s hospitals in the U.S. that are affiliated with the Children’s Hospital Association. The majority are teaching hospitals in large metropolitan areas, are level I or II trauma centers, and are located in diverse geographic areas. Included data are de-identified and subjected to rigorous reliability and validity checks.³⁴ Children under 1 year of age who were admitted to the emergency department(ED), inpatient unit, observational unit or operating room with ICD-9-CM codes for femur fracture, humerus fracture or TBI (skull fractures without intracranial hemorrhage not included) between January 2004 and September 2015 were eligible for inclusion (Appendix 1). We excluded patients with ICD-9-CM or external cause of injury codes for following conditions: birth-related injuries, motor vehicle crash-related injuries, bone disorders, bleeding disorders, metabolic/genetic disorders, and brain neoplasms (Appendix 2). We also excluded repeat admissions, duplicate hospital encounters for the same injury and transfers to or from another hospital because we were unable to determine if a SS was completed (Figure 1). We did not include codes for child abuse as our research question was not related to a diagnosis of abuse, but rather if a SS was obtained. We assumed the skeletal survey would be a systematically performed series of radiographic images that encompasses the entire skeleton as outlined by the American College of Radiology. All patients met inclusion and exclusion criteria at every site.

Figure 1.

Cohort flow diagram

MVC = Motor Vehicle Crash

We surveyed ED providers and designated child abuse providers at 49 PHIS hospitals to determine the presence and characteristics of a child abuse clinical pathway. We first reached out to the chief of the hospital’s ED. If the department chief was unable to answer the survey or recommended another respondent, we then reached out to the recommended respondent, followed by child abuse providers until we received a survey response. Survey respondents were asked to submit a copy of the relevant pathway(s). Survey data were collected and managed using REDCap database hosted at Children’s Hospital of Philadelphia. REDCap is a secure, web-based application designed to support data capture for research studies.³⁵ We had a completed survey response rate of 84%. The surveys included questions regarding year of pathway implementation, how providers access the pathway and the presence of a child abuse team. The survey evaluated whether the pathways had injury-specific recommendations for fractures, head injury, burns, bruising, abdominal injury and other injuries, such as poisoning. A study team member performed quality checks by reviewing submitted pathways and confirming they fell under the definition of pathway.^18,19 If the pathway was not available for review the study team reviewed the submitted survey to evaluate if enough detail was included to determine if the hospital pathway met our criteria for a clinical pathway. If clarification about a survey recipient’s response was needed, a study team member contacted the respondent to clarify. Our analysis included the 41 hospitals that responded to the survey and that had patient-level administrative data available during the study period.

Primary Predictor Variable

The primary predictor variable was the presence of a child abuse pathway (as determined by our survey data) during the year of the infant’s admission to the hospital.

Primary Outcome Variable

The primary outcome was the performance of SS as determined by the presence of a procedure and/or billing code for a SS or radionuclide bone scan (Appendix 3). In most cases SS is the preferred test for occult fractures but in select cases, radionuclide bone scans can serve as an alternative test.⁴

Covariates

Patient-level covariates included age (categorical 0–5 months and 6–11 months), race (White, Black, Other), insurance (private, public/none, unknown), gender, and type of injury (femur fracture, humerus fracture, or TBI). These covariates were chosen because they have been identified as predictors of SS performance in a child presenting with an injury.^4,13,14,32

Statistical Analysis

Descriptive statistics were used to summarize patient level and hospital level characteristics. Unadjusted association between predictor variables and occult fracture imaging were evaluated using the Chi-square test. Next, we performed a mixed effects logistic regression to examine the association between a pathway and SS performance. The following patient-level covariates were included in our model as fixed effects: age, sex, race, insurance, and injury type. We did not include the presence of a child abuse team in our model as 95% of included hospitals had at least a basic child abuse team as defined by at least one child abuse pediatrician and social work.³⁶ To account for intra-hospital correlation of outcomes, hospital was included as a random effect. An indicator variable for presence of a pathway was also included as a fixed effect; the odds-ratio for this variable represents the within hospital odds-ratio for SS performance after implementation of a pathway (versus before implementation). We did not include year in our model, as discharge year and presence of pathway were found to be co-linear and the exact time within a year in which a pathway was implemented is unknown. Age was included as a categorical variable. Next, the regression model was repeated, testing for an interaction between pathway presence and insurance type. Marginal standardization was utilized to obtain the predicted probability of SS performance from this model.³⁷ The marginal standardization method uses the entire sample as the standard population and estimates the probability of a SS by assuming that children were alternatively assigned to a hospital with and without a pathway.³⁷ Sensitivity analysis was performed dropping hospitals that had less than 10 children to look for statistically significant changes.

Results

Between the years 2004 and 2015, 2736 children under the age of 12 months were treated across the 49 hospitals with a diagnosis of a femur fracture, humerus fracture or TBI. Only the first visit for each of these diagnoses was included. We excluded 135 subjects due to the following ICD-9-CM or external cause of injury codes: birth injury, motor vehicle accident, bleeding disorder, brain neoplasm or metabolic bone disease. We excluded all subjects whose hospital did not complete our survey (n=342 patients and 8 hospitals), and duplicate admissions for the same injury (n=172). We also excluded patients for whom gender was unknown (n=2) for a final cohort of 2085. [See Figure 1 - Cohort flow diagram]

Subject Characteristics

Among the 2085 included cases, 55% were male, 69% had public insurance or no insurance and 64% were White (Table 1). Fifty-eight percent presented to an institution in a year when a pathway was present.

Table 1:

Characteristics of patient population (N=2085)

Characteristic	Total (%) (N=2085)	Femur Fracture(N=416)	Humerus Fracture(N=234)	TBI (N=1594)
Age
0–5 months	1305(62.6)	267(64.2)	148(63.2)	1014(63.6)
6–11 months	780(37.4)	149(35.8)	86(36.8)	580(36.4)
Gender
Male	1,145(54.9)	218(52.4)	131(56)	887(55.6)
Female	940(45.1)	198(47.6)	103(44)	707(44.4)
Race
White	1,331(63.8)	269(64.7)	141(60.2)	1,025(64.3)
Black	348(16.7)	61(14.7)	45(19.2)	270(16.9)
Other	308(14.8)	66(15.8)	37(15.8)	224(14.1)
Missing	98(4.7)	20(4.8)	11(4.7)	75(4.7)
Insurance
Private	487(23.4)	85(20.4)	31(13.2)	394(24.7)
Public/None	1446(69.4)	304(73.1)	192(82.1)	1078(67.6)
Unknown	152(7.3)	27(6.5)	11(4.7)	122(7.7)
Academic Center
Yes	2004(96.1)	395(94.9)	226(96.6)	1538(96.5)
No	81(3.9)	21(5.1)	8(3.4)	56(3.5)
Pathway
Yes	1,215(58.3)	242(58.2)	124(53)	944(59.2)
No	870(41.7)	174(41.8)	110(47)	650(40.8)

Pathway Characteristics

At the time of our survey, 24 of 41 (59%) hospitals reported having a pathway. The majority of pathways were implemented after 2007. Pathways varied in their complexity and level of guidance. All pathways offered guidance for fractures, whereas 17 (71%) provided specific information related to bruising, 13 (54%) had specific guidance for burns, 17 (70%) had guidance for head trauma and 15 (62%) had specific guidance for abdominal trauma. All pathways recommended a SS when abuse was suspected. Only 9 hospitals with pathways (37%) reported having alerts embedded in the electronic medical record that prompt clinicians to refer to a child abuse pathway for specific patients. Survey respondents reported accessing their hospitals clinical pathway a variety of different ways ranging from using the internet, to the EMR to paper versions. See Appendix 4 for pathway characteristics.

Skeletal Survey Evaluation: Unadjusted Results

In the unadjusted analysis, a SS was performed in 86% of children 0–5 months and 73% of children 6–11 months with a femur fracture, in 85% of children 0–5 months and 67% of children 6–11 months with a humerus fracture and in 74% of children 0–5 months and 67% of children 6–11 months with TBI (Table 2). In all three injury categories SS’s were more likely to be performed in the 0–5 month olds than 6–11 months (p < 0.005). SS’s were performed less often for those with private insurance than for those with public insurance in patients with femur fracture and TBI (Table 2).

Table 2.

Unadjusted percentages of infants undergoing skeletal survey

Characteristic	Femur Fracture (N=416)	P-value	Humerus Fracture (N=234)	P-value	TBI (N=1594)	P- value
Age
0–5 months	230 (86.1)*		126(85.1)		746(73.6)
6–11 months	108 (72.5)	0.001	58(67.4)	0.001	388(66.9)	0.005
Gender
Male	179(82.1)		106(80.9)		654(73.7)
Female	159(80.3)	0.637	78(75.7)	0.337	480(67.9)	0.803
Race
White	221(82.2)		115(81.6)		709(69.2)
Black	49(80.3)		33(73.3)		210(77.8)
Other	53(80.3)		27(73)		158(70.5)
Unknown	15(75)	0.866	9(81.8)	0.523	57(76)	0.035
Insurance
Private	59(69.4)		22(71)		216(54.8)
Public/None	259(85)		153(79.7)		835(77.5)
Unknown	20(74)	0.003	9(81.8)	0.528	83(68)	0.000
Pathway
Yes	198(81.8)		95(76.6)		694(73.5)
No	140(80.4)	0.726	89(80.9)	0.424	440(67.7)	0.012

^*numbers in parenthesis are percentages derived from totals within each injury category. See injury columns in table 1 for denominator.

Skeletal Survey: Adjusted Results

In analysis adjusting for covariates (age, race, insurance and injury type), the presence of a child abuse pathway in a hospital was associated with increased odds of getting a SS (OR 1.46, 95% CI 1.02–2.08) (Table 3). Black race was not associated with increased odds of SS performance; however, children with public insurance had a greater odds of receiving a SS (OR 2.75, 95% CI 2.11–3.52).

Table 3.

Adjusted Association of Pathway and Patient-Level Factors with Performance of Skeletal Survey

Factor	aOR	95% CI	P-value
Pathway
No	Ref
Yes	1.46	1.02–2.08	0.038
Age
0–5month	Ref
6–11 months	0.61	0.48–0.77	<0.001
Race
White	Ref
Black	1.21	0.86–1.66	0.265
Other	1.03	0.73–1.44	0.874
Insurance
Private	Ref
Public	2.75	2.11–3.52	0.000
Injury Category*
TBI	Ref
Femur Fracture	1.76	1.24–2.47	0.001
Humerus Fracture	0.85	0.55–1.29	0.448

^*The OR for SS performance base on injury type and independent of pathway presence. TBI serves as reference category

We further explored the impact of a pathway on the association of insurance type with odds of SS performance by interacting insurance status with pathway. We found that the impact of a pathway on SS performance was greater for children with public insurance than for children with private insurance (p-value <0.001). The presence of a pathway increased the predicted probability of SS performance for those with private insurance from 54% to 58% (p-value <0.001). For those with public insurance the presence of a pathway increased SS performance from 73% to 81% (p-value <0.001).

Discussion

In this multicenter study of children under the age of 12 months who presented with a femur fracture, humerus fracture or traumatic brain injury, we found that hospitals with a child abuse pathway had higher odds of SS performance. We found no difference in SS performance based on race; however, we found that the presence of a pathway increased the likelihood of SS performance for infants with public insurance.

This finding related to public insurance and SS performance is consistent with prior litturature^14,38 suggesting ongoing differences in evaluation potentially based on socioeconomic status(SES). We hypothesized that the presence of a pathway would equalize evaluation rates between infants with public and private insurance but instead found that the presence of a pathway intensified the difference. While we are unable to determine the reason for the increased difference, there are a few possible explanations. One could be continued conscious or unconscious bias amongst frontline clinicians. Although insurance is not a perfect indicator of SES multiple studies have shown that public insurance can be used as predictor of living at or below the federal poverty level.^39–41 Given that poverty has been shown to be a risk factor for abuse, clinicians may preferentially evaluate poor children for occult injury when compared to those with perceived higher SES, even when a pathway is present.^3,42,43 Another cause for our finding could be related to what has been described as social intuition. In a recent article by Keenen et al “social intuition” and “social information” was found to play a role in abuse evaluation. They found that Child Abuse Pediatricians were twice as likely to perform gold standard abuse evaluation for neurotrauma and long bone fractures if blinded to social information.⁴⁴ This social intuition has been theorized to cause over evaluation in certain groups and under evaluation in other groups leading to a wide variety of practice.

We recognize that decision making is a complex process guided by experience, available resources, analytic thinking and potentially bias. Our results demonstrate that even with the implementation of a decision making tool such as a pathway, the desired impact on practice may not be achieved. This finding underscores the need for ongoing implementation and quality improvement research regarding child abuse clinical pathways.

Prior studies have identified racial differences in evaluations for occult injury, with minority children receiving higher rates of SS’s.^13,16,45 Contrary to what some prior literature has shown, we did not identify an association between race and odds of occult injury evaluation in either the unadjusted or adjusted analyses. There are multiple reasons that could explain this finding. Prior literature has suggested that racial biases in child abuse evaluations are most prominent in children with presentations associated with a lower risk of abuse (older children with lower risk fractures).¹³ Thus, our decision to focus on a higher risk population may explain the lack of racial disparity in the approach to care. The AAP has recently addressed this issue by publishing guidelines aimed at increasing both awareness of child abuse and the known bias.³ It is possible that increased awareness regarding racial biases in child abuse evaluation decision making has led to decreased disparities in practice.

There are multiple limitations to our study. The study design was retrospective, thus we cannot infer causality from our results. Our patient-level data were limited to administrative data, which is subject to misclassification of diagnosis codes. Additionally, there are unmeasured confounders, including patient and provider level characteristics that may influence SS ordering. We did not have access to the history obtained by the frontline clinician, which may have influenced the decision to obtain a SS. We attempted to address this limitation by only including patients that, based on their age and injury type, should receive a SS in most cases. Second, temporal trends may account for some of the identified increase in evaluations for occult injuries. We were unable to account for any changes in a pathway over time after year one of implementation. Due to the high correlation between the presence of pathway and time, we were unable to adjust for this in our model. We attempted to address this limitation by using mixed effects regression analysis, looking at within-hospital practice pre and post pathway implementation based on the presence or absence of a pathway. Another limitation is that we are unable to evaluate drivers of compliance and institutional infrastructure related to pathway implementation and usability. The presence of a pathway may simply be a proxy for greater child abuse education at a hospital. Perhaps the implementation process was more of a driver than the actual pathway. Based on our data we are unable to determine if the pathway itself is the key driver towards improved SS performance or if it serves as a surrogate for other influential drivers. Finally, while survey respondents provided the best information available, recall bias could contribute to inaccurate responses regarding timing of pathway implementation. Based on the heterogeneity of the clinical pathways, we were unable to account for whether characteristics of the pathway influenced SS performance. We were unable to determine if clinical pathways with more detailed recommendations regarding SS’s led to higher rates of evaluations, as our regression model was based solely on the presence or absence of a pathway. Since there is no current gold standard for a child abuse clinical pathway, it remains to be determined if certain features of child abuse pathways impact the effectiveness of the pathway.

Despite these limitations, our study provides a contribution to the literature about the potential influence of pathways on adherence to guidelines and evaluation for occult injuries in high-risk infants. Based on this study, we recommend that hospitals consider implementing clinical pathways to standardize and improve evaluation practices in infants with high-risk injuries. Further research is needed to 1) identify pathway characteristics that result in the most accurate evaluation and 2) evaluate other drivers of compliance that lend to successful use of the pathway.

In addition we recommend careful monitoring post-pathway implementation to ensure that it has the desired impact on patient care and outcomes

Conclusion

When a dedicated child abuse pathway was present, children with injuries associated with a high risk of abuse were more likely to have a skeletal survey performed. Differences in SS performance continue to exist between patients with private versus public insurance. Further research is needed to better understand the full role pathways play as a decision-guiding tool in cases of possible abuse and their impact on outcomes.

What’s New.

Child abuse clinical pathways may serve an important role in increasing skeletal survey performance amongst infants with injuries associated with a high risk of abuse. Infants with public insurance receive skeletal surveys at higher rates that those with private insurance.

Abbreviations:

AAP

American Academy of Pediatrics

ICD-9-CM

Internal Classification of Diseases, Revision 9, Clinical Modification

PHIS

Pediatric Hospital Information System

OR

Odds Ratio

CI

Confidence Interval

SS

Skeletal Survey

TBI

Traumatic Brain Injury

ED

Emergency Department