Prevalence of abuse among young children with femur fractures: a systematic review

A systematic review of the literature on abuse in children with fractures was performed using a pre-specified protocol with inclusion criteria (available upon request). This paper covers the subset of articles specific to femur fractures (Figure 1). We performed searches for studies published in English between January 1990 and July 2013 in the PubMed/MEDLINE and CINAHL databases using the search terms listed in Additional file 1. We included terms related to both abuse and accidental trauma to avoid bias toward studies focused exclusively on abuse. Studies were also identified by iteratively reviewing reference lists of articles identified during the search.

Randomized controlled trials (RCT), prospective non-RCT studies, and retrospective data analyses were included; surveys, reviews, editorials, case series and textbooks were excluded. Studies were included if subjects were ≤5 years old or if the data for the subset of children ≤5 years old could be extracted. Studies including fewer than 10 children ≤ 5 years old with femur fractures were excluded, as were animal and post-mortem studies. Methodologically weak studies due to significant bias in selection of subjects, such as studies including only cases seen by the investigator for medical-legal review or studies including only the subset of patients who were eligible for a specific treatment modality, were excluded. Titles and abstracts of studies were screened by one of four reviewers (JW, OF, Maria Fatima de Reyes, VM), and non-relevant studies were eliminated (Figure 1). Full manuscripts for relevant studies were assessed for eligibility by two reviewers (JW and OF or VM) in an unblinded standardized manner, with disagreements resolved by consensus.

Two reviewers (JW, OF) independently extracted the following information from the studies using a standardized form: 1) study population characteristics (ages, type(s) of fractures, study location and dates), 2) inclusion and exclusion criteria for subjects, 3) potential biases, and 4) number of fractures attributed to abuse and accidental mechanisms in overall study population as well as within study subpopulations. The reviewers assessed the level of the evidence presented in each study using the 2011 Oxford Centre for Evidence-Based Medicine (CEBM) Levels of Evidence table (Table 1). Per the CEBM Evidence Levels, articles providing the strongest evidence are likely to be assessed as a level 1, and those providing the weakest evidence are likely to be assessed as a level 5 [14, 15]. The CEBM Evidence Levels for “How common is the problem” were applied to studies of the prevalence of abuse in children with femur fractures. The CEBM Evidence Levels for diagnosis studies were adapted for assessing studies examining the association of specific clinical features with likelihood of abuse determination. Studies providing information on both questions were assigned a single score based on the question for which they had the weakest level of evidence. The level of evidence assigned to a study is specific to the level of evidence for answering the questions posed in this review and may not be reflective of the overall quality of the study. For example, the level of evidence was downgraded for studies with small numbersof children with femur fractures although the studies may have had appropriate methods and large study populations that included children with other types of fractures. Similarly, studies were classified as non-current and received a lower level if any of the data was from prior to 2000. Finally, the CEBM Evidence Levels for prevalence studies include assessment of whether the study population is local or not, but we chose to eliminate this factor as we were not attempting to estimate the prevalence of abuse in a particular location. Instead we provide the location of each study and leave it to the reader to determine the applicability of the study data to their population of interest. For each study, the reviewers also rated the methodology used to determine that an injury was due to abuse using a scale adapted from Maguire et al. [16] which assigned the highest rank (1) to studies requiring that abuse be either confirmed at a child abuse case conference or civil or criminal court proceedings, or admitted by a perpetrator, or witnessed. The lowest rank (5) was accorded to studies providing no stated criteria for categorizing cases as suspected abuse (Table 2). Reviewers resolved disagreements through discussion and consensus.

The abuse prevalence for each study was calculated with 95% confidence intervals (CIs) from the data reported in cohort and cross-sectional studies. The study-specific sensitivity, specificity, positive likelihood and negative likelihood ratios with 95% CIs of different clinical characteristics for abuse were also computed. Formulae described by Simel et al. [17] were applied to calculate 95% CIs for likelihood ratios. All other analyses were performed using Stata 12 (StataCorp, College Station, TX). The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist was utilized in reporting the results of the systematic review (Additional file 2).

The comprehensive literature search identified 7,009 non-duplicate citations on all fractures, of which 271 were deemed relevant. Twenty-four studies on femur fractures met inclusion criteria (Figure 1). [2, 4‐6, 10, 18‐36] A total of 10,717 children of ages 0–5 with femur fractures were examined in the 24 studies. Two studies using a national database [10, 19] received a study methodology quality rating of L1, 1 study [31] received a rating of L2, 3 studies [2, 20, 24] received a rating of L5, and the remainder of studies received ratings of L3 or L4 (Table 3) [4‐6, 18, 21‐23, 25‐30, 32‐36]. Three studies requiring that abuse be confirmed by case conference or court proceedings following a child protective services (CPS) investigation received a rating of 1 for the abuse determination methods [18, 22, 26]. Ten studies reported specific clinical criteria applied in diagnosing abuse (rating 2 or 3) [4, 20, 21, 24, 25, 29‐31, 33, 36]. Eleven studies, including 8 studies relying on administrative data, received abuse determination ratings of 4 or 5 [2, 5, 6, 10, 19, 23, 27, 28, 32, 34, 35].

The inclusion and exclusion criteria applied in subject selection varied among the 24 studies, with differences in the types of femur fractures and possible etiologies considered. Only 5 studies included children with any femur fracture type from any etiology [10, 19, 21, 27, 36]. Thirteen studies excluded children with pathologic fractures and/or children with a clear accidental etiology such as motor vehicle crash (MVC) [2, 5, 6, 18, 20, 22, 26, 30, 32‐35]. Eight included children with specific types of fractures or specific reported trauma histories [22, 23, 25, 26, 28, 31‐34]. Three studies excluded children with additional injuries, [4, 22, 26] potentially biasing the abuse prevalence lower.

Among studies including children 0–36 months with all types of femur fractures from any reported etiology, estimated prevalence of abusive fractures ranged from 11.0%-31.2% (Figure 2) [4, 10, 21, 29, 30] Exclusion of MVC-related cases increased the range of reported abuse prevalence in the studies to 11.6%-50.0% [4, 21, 30]. Restricting the population in these studies to children <12 months old resulted in a range of 16.7%-30.5% abuse prevalence if MVCs were included [10, 36] and 16.7%-35.2% if MVCs were excluded [5, 6, 30]. One study which included cases categorized as “child abuse suspected” in the trauma registry and which received the lowest child abuse methodology rating of 5, reported an abuse prevalence of 68.3% in children <18 months old [27]. In a study of children ≤36 months old with a reported stair fall history, 13.8% of children with femur fractures were categorized as abused [31]. Among children >12 months old with all types of femur fractures, abuse was diagnosed or confirmed in 1.5-6.0% of cases [10, 18, 30].

In a single Swedish study relying on administrative data, 4.2% of children <12 months old with diaphyseal femur fractures were categorized as abused (Figure 3) [34]. A higher abuse prevalence, 13.7%, was reported in an American study of children <24 months old, also relying on administrative data [23]. Two studies, one American and one Thai, reported even higher rates of suspected abuse among children <60 months old with non-MVC-related diaphyseal femur fractures: 31.6% and 31.0% respectively [32, 33]. Two studies focused on children with diaphyseal femur fractures and no additional injuries reported lower prevalences of cases substantiated as abusive by Child Protective Services (CPS) in their samples (2.4% and 7.9%) [22, 26].

In the single study of children <12 months old with a complete distal metaphyseal fracture, abuse was suspected in 75% and diagnosed in 50% of cases [28].

Age <12 months was associated with increased abuse likelihood in 3 of 5 studies, with positive likelihood ratios ranging from 3.3-19.7 (Table 4) [10, 20, 24, 30, 32]. Another study reported increased probability of abuse among children aged <18 months, as compared to counterparts ≥18 months [2]. Two studies examined the relationship between age in years as a continuous variable and risk of abuse: one found significant association between younger age and increased risk [26], but the other did not [22]. Non-ambulatory status was associated with increased abuse likelihood in both studies examining this association [26, 30].

Male gender and Medicaid/uninsured status were each associated with increased likelihood of abuse likelihood in only one [2, 32] of several studies examining these associations [22, 26, 28, 30, 35]. No association was found between abuse likelihood and race [26, 35]. In a single study, lower scores on the Hollingshead Occupational Scale (HOS) for fathers, but not mothers, were associated with reports to CPS [22].

A suspicious history was associated with abusive femur fracture in 3 of 3 studies (Table 5) [2, 22, 30]. The definition of suspicious history varied but generally included history of no trauma, an unwitnessed trauma history, or a history considered inconsistent with the injury. Delay in care of >24 hours and unknown trauma mechanism were associated with abuse, while a trauma history witnessed by a non-parent was associated with non-abusive femur fracture [22, 30, 32]. In one study, injuries occurring at home or an unknown location were associated with abuse compared to injuries occurring in public places [35].

Additional injuries, including bruises, other fractures identified on skeletal survey, and subdural hemorrhages, were significantly associated with determination of abusive femur fracture in 3 of 4 studies (Table 5 )[2, 30‐32]. Universal screening for additional injuries was, however, not performed in any of these studies.

In 2 of 3 studies, diaphyseal fractures were associated with decreased abuse likelihood, compared to other fracture types (Table 6) [2, 30, 35]. Distal diaphyseal fractures, in comparison with proximal and midshaft diaphyseal fractures, were associated with abuse in another study [26]. In one report, distal metaphyseal fractures showed increased abuse probability [2], but this finding was not replicated in a second study [30]. Fracture side, spiral fracture type, or bilateral fractures did not show significant association with abuse [22, 26, 28, 30].