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

  • Joanne N Wood1, 2, 3Email author,

    Affiliated with

    • Oludolapo Fakeye1,

      Affiliated with

      • Valerie Mondestin1,

        Affiliated with

        • David M Rubin1, 2, 3,

          Affiliated with

          • Russell Localio4 and

            Affiliated with

            • Chris Feudtner1, 2, 3

              Affiliated with

              BMC Pediatrics201414:169

              DOI: 10.1186/1471-2431-14-169

              Received: 15 January 2014

              Accepted: 19 June 2014

              Published: 2 July 2014

              Abstract

              Background

              Clinical factors that affect the likelihood of abuse in children with femur fractures have not been well elucidated. Consequently, specifying which children with femur fractures warrant an abuse evaluation is difficult. Therefore the purpose of this study is to estimate the proportion of femur fractures in young children attributable to abuse and to identify demographic, injury and presentation characteristics that affect the probability that femur fractures are secondary to abuse.

              Methods

              We conducted a systematic review of published articles written in English between January 1990 and July 2013 on femur fracture etiology in children less than or equal to 5 years old based on searches in PubMed/MEDLINE and CINAHL databases. Data extraction was based on pre-defined data elements and included study quality indicators. A meta-analysis was not performed due to study population heterogeneity.

              Results

              Across the 24 studies reviewed, there were a total of 10,717 children less than or equal to 60 months old with femur fractures. Among children less than 12 months old with all types of femur fractures, investigators found abuse rates ranging from 16.7% to 35.2%. Among children 12 months old or greater with femur fractures, abuse rates were lower: from 1.5% - 6.0%. In multiple studies, age less than 12 months, non-ambulatory status, a suspicious history, and the presence of additional injuries were associated with findings of abuse. Diaphyseal fractures were associated with a lower abuse incidence in multiple studies. Fracture side and spiral fracture type, however, were not associated with abuse.

              Conclusions

              Studies commonly find a high proportion of abuse among children less than 12 months old with femur fractures. The reported trauma history, physical examination findings and radiologic results must be examined for characteristics that increase or decrease the likelihood of abuse determination.

              Keywords

              Child abuse Child maltreatment Femur fracture Accident Trauma

              Background

              Femur fractures are the most common orthopedic injury for which children are hospitalized in the United States [1, 2]. Although the majority of childhood femur fractures result from accidental trauma, abuse is also a common cause of these fractures, especially in children less than 1 year old. Thus, medical providers caring for children with femur fractures should recognize and evaluate children who might be abuse victims. Abuse evaluation and diagnosis rates among children with femur fractures have, however, been noted to vary among hospitals and providers [36]. Furthermore, studies have shown that failing to recognize and evaluate for abuse in young children with fractures can result in children suffering complications from additional undiagnosed injuries as well as ongoing abuse [7].

              Although femur fractures have been associated, in general, with a high abuse risk in young children, [810] the prevalence of abuse in children with different types of femur fractures has not been established. Moreover, other clinical features that increase or decrease the likelihood of abuse determination in children with femur fractures have not been well elucidated. This uncertainty regarding which children with femur fractures might have been abused may contribute to variation in care and missed opportunities to diagnose abuse in the pediatric population.

              Given these uncertainties, we systematically reviewed published studies in order to: 1) provide estimates of abuse prevalence among children ≤5 years old with femur fractures and 2) describe the association of specific clinical features with likelihood of abuse determination. Recognizing that the vast majority of abusive fractures occur in the infants and young toddlers, we included children up to age 5 years in our review as abusive fractures have been occasionally reported in preschool age children [1013]. Due to the heterogeneity of study populations, we did not perform a meta-analysis. Instead, we present the proportions of children diagnosed with abuse in each study as well as details of the study population, in order to provide a richer understanding of the prevalence of abuse in different sub-populations of children with femur fractures.

              Methods

              Search strategy

              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.
              http://static-content.springer.com/image/art%3A10.1186%2F1471-2431-14-169/MediaObjects/12887_2014_1093_Fig1_HTML.jpg
              Figure 1

              Processes of study identification, screening, eligibility assessment and inclusion.

              Study selection

              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.

              Data extraction, assessment of methodological quality, and analysis

              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.
              Table 1

              Study methodology rating scales; levels of evidence scale*

              Question

              How common is physical abuse?

              Is this factor associated with risk of physical abuse?

              Level 1

              Current random sample surveys (or censuses)

              Systematic review of cross sectional studies with consistently applied reference standard and blinding

              Level 2

              Systematic review of surveys that allow matching to local circumstances

              Individual cross sectional studies with consistently applied reference standard and blinding

              Level 3

              Non-random or non-current sample

              Non-consecutive studies, or studies without consistently applied reference standards

              Level 4

              Case-series

              Case–control studies, or “poor or non-independent reference standard

              Level 5

              n/a

              Mechanism-based reasoning

              *Adapted from the 2011 Oxford CEBM Evidence Levels of Evidence table [14]. Level may be graded down on the basis of study quality, imprecision, indirectness, or because the absolute effect size is very small. Current was defined as data from 2000 or later.

              Table 2

              Study methodology rating scales; abuse determination methodology rating scale*

              1

              Abuse confirmed at case conference or family, civil, or criminal court proceedings; admitted by perpetrator; or witnessed abuse

              2

              Abuse confirmed by stated criteria including multidisciplinary assessment

              3a

              Abuse defined using specific stated case based criteria

              3b

              Abuse including cases of likely or probable abuse defined by specific stated case based criteria

              4

              Abuse stated but no supporting detail given as to how a determination of abuse was made#

              5

              Suspected abuse

              *Adapted from a scale with 5 levels developed by Maguire et al. (2005). [16] For studies using specific stated case based criteria to make a determination of abuse (rating 3), we distinguished studies that included only definite abuse cases (3a) from those that also included likely or probable abuse cases (3b). Assessment by multidisciplinary hospital-based child protection team as part of routine clinical care did not qualify as multidisciplinary assessment. #Level 4 includes studies relying on ICD-9 and E-codes for identifying abuse cases in administrative data sets and studies relying on diagnoses of abuse made by clinical teams without providing specific criteria by which these diagnoses were made.

              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).

              Results

              Description of studies

              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, 46, 10, 1836] 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) [46, 18, 2123, 2530, 3236]. 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, 2931, 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].
              Table 3

              Summary of study characteristics

              Study (1stauthor, year)

              Study Location/Source

              Dates

              Age (mo)

              Inclusion criteria

              Exclusions

              Study Methods Rankinga

              n

              Dalton, 1990 [4]

              3 Michigan Hospitals, USA

              1979-1983

              <36

              Femur fracture all types

              Additional injuriesb

              L4 / 3a

              138

              Thomas, 1991 [20]

              Yale-New Haven Hospital, USA

              1997-1984

              <36

              Femur fracture all types

              Pathologic fracture

              L5 / 3b

              25c

              Kowal-Vern, 1992 [21]

              Loyola University Medical Center, USA

              1984-1989

              <36

              Femur fracture all types

              None

              L4 / 3a

              14c

              Blakemore, 1996 [22]

              C.S. Mott Children’s Hospital, USA

              1979-1993

              12-71

              Isolated femur diaphyseal fracture

              Pathologic fractures, MVC related fractures, additional injuriesb

              L4 / 1

              42

              Hinton, 1999 [23]

              Hospital Discharge Database of the Maryland Health Services Cost Review Commission, USA

              1995-1996

              <24

              Femur diaphyseal fractures

              Non-acute fracture, multiple admissions

              L3 / 4

              73c

              Rex, 2000 [24]

              Manchester Children’s Hospitals, UK

              1992-1996

              <60

              Femur fracture all types, definite abuse or accident

              Unclear etiology, non-acute fractures

              L5 / 3a

              33c

              Scherl, 2000 [25]

              The University of Chicago Children’s Hospital & King’s County Hospital, USA

              1986-1996

              <72

              Closed diaphyseal femur fracture

              Non-diaphyseal fractures, open fracture, pathologic fracture

              L3 / 2

              207

              Schwend, 2000 [26]

              Children’s Hospital of Buffalo, New York, USA

              1993-1997

              <48

              Diaphyseal femur fracture

              Pathologic fracture, additional injuriesb

              L4 / 1,3b

              139

              Banaszkiewicz, 2002 [36]

              Royal Aberdeen Children’s Hospital, UK

              1995-1999

              <12

              Femur fracture all types

              None

              L4 / 3b

              12c

              Jeerathanyasakun, 2003 [33]

              Queen Sirikit National Institute of Child Health, Thailand

              1996-2001

              <60

              Diaphyseal femur fracture

              Non-diaphyseal fracture, distal greenstick fracture, pathologic fracture

              L4 / 3b

              39

              Coffey, 2005 [27]

              Children’s Hospital, Columbus, OH, USA

              1998-2002

              < 18

              Femur fracture all types

              None

              L4 / 5

              41c

              Pierce, 2005 [31]

              Children’s Hospital of Pittsburgh, PA, USA

              1999-2002

              ≤36

              Femur fracture all types, reported history of stair fall

              Reported history other than stair fall

              L2 / 3a

              29

              Rewers, 2005 [35]

              Colorado Trauma Registry, USA

              1998-2001

              <36

              Femur fracture all types

              Pathologic fracture, non-Colorado resident, repeat admission for complication

              L3 / 4

              332 c

              Loder, 2006 [19]

              National (Kids’ Inpatient Database, USA)

              2000

              < 24

              Femur fracture all types

              None

              L1 /4

              1,076c

              Arkader, 2007 [28]

              Two Level I pediatric centers, USA

              1995-2005

              ≤ 12

              Complete distal metaphyseal femur fracture

              Incomplete metaphyseal and epiphyseal fractures

              L3 / 4, 5

              20

              Trokel, 2006 [5]

              National (Kids’ Inpatient Database, USA)

              1997

              <12

              Femur fracture all types, admitted through ED

              MVC, gunshot, or stabbing related fracture; no external cause of injury code

              L3 / 4

              426c

              Leventhal, 2007 [29]

              Yale-New Haven Children’s Hospital, CT, USA

              1979-1983 1991–1994 1999-2002

              <36

              Femur fracture all types

              Pathologic fracture

              L4 / 3b

              81c

              Hui, 2008 [30]

              Alberta Children’s Hospital, Canada

              1994-2005

              <36

              Femur fracture all types

              Pathologic fracture

              L4 / 3b

              127

              Leventhal, 2008 [10]

              National (Kids’ Inpatient Database, USA)

              2003

              <36

              Femur fracture all types

              None

              L1 / 4

              4,026c

              Baldwin, 2011 [2]

              The Children’s Hospital of Philadelphia, PA, USA

              1998+

              <48

              Femur fracture all types

              Pathologic fracture, cause of fracture not clearly determined

              L5 / 4

              209

              Heideken, 2011 [34]

              Swedish National Hospital Discharge Registry, Sweden

              1987-2005

              <12

              Diaphyseal femur fracture

              Non-diaphyseal fracture, pathologic or birth fracture, multiple femur fractures

              L3 /4

              313c

              Shrader, 2011 [32]

              Phoenix Children’s Hospital, AZ, USA

              2003-2008

              <60

              Diaphyseal femur fracture

              Pathologic fractures, non-diaphyseal fractures

              L3 / 4

              137

              Wood, 2012 [6]

              Pediatric Health Information System Database (40 pediatric hospitals), USA

              1999-2009

              <12

              Femur fracture all types

              MVC, birth, or neoplasm related fractures

              L3 / 4

              2,975c

              Capra, 2013 [18]

              The Hospital for Sick Children, Toronto, Canada

              1995-2004

              12-59

              Femur fracture all types

              Non ambulatory children, pathological fractures

              L3 /1

              203

              aPresents overall study methodology ranking (L1-L5) and abuse determination methodology ranking (1–5). See Table 1 for description of the ranking scales. Some studies utilized multiple different methods to define cases of abuse or suspected abuse and therefore received more than one ranking.

              bPatients with other injuries (in addition to the femur fracture) were excluded from these studies but the exact definition of additional injuries varied.

              cThe data presented are for the relevant subset of a larger study population which may have included children with other types of injuries, other ages, and/or from other time periods.

              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, 3235]. Eight included children with specific types of fractures or specific reported trauma histories [22, 23, 25, 26, 28, 3134]. Three studies excluded children with additional injuries, [4, 22, 26] potentially biasing the abuse prevalence lower.

              Abuse prevalence in young children with femur fractures: all types

              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].
              http://static-content.springer.com/image/art%3A10.1186%2F1471-2431-14-169/MediaObjects/12887_2014_1093_Fig2_HTML.jpg
              Figure 2

              Probability of Abuse in Children with a Femur Fracture, All Types. Proportion of cases with abusive femur fractures in included studies, by subject age criteria of inclusion. aUpper age limit was 11 months in Leventhal [10] & Wood [6] and 12 months in Hui [30]. bData for only the subset of children admitted to children’s hospitals or to general hospitals without children’s hospitals could be extracted. cStudy was limited to children with isolated femur fracture and no additional injuries. dIncluded children with reported history of stair fall ≤36 months old only. ePresents overall study methodology ranking (L1-L5) and abuse determination methodology ranking (1–5).

              Probability of abuse diagnosis in young children with femur fractures: diaphyseal type

              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].
              http://static-content.springer.com/image/art%3A10.1186%2F1471-2431-14-169/MediaObjects/12887_2014_1093_Fig3_HTML.jpg
              Figure 3

              Probability of Abuse in Children with Diaphyseal Femur Fractures. aStudy was limited to children with isolated femur fractures without additional injuries. bPresents overall study methodology ranking (L1-L5) and abuse determination methodology ranking (1–5).

              Probability of abuse diagnosis in young children with femur fractures: distal metaphyseal type

              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].

              Patient and family demographics associated with abuse determination

              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].
              Table 4

              Association of demographic characteristics with likelihood of abuse

              Characteristics

              Study

              Sensitivity (95% CI) (95%CI)

              Specificity (95%CI)

              LR+ (95% CI)

              LR- (95% CI)

              P a

              Age

                    

                   < 12 m.o. vs. 12 m.o.-35 m.o.

              Thomas, 1991

              66.7 (35.9-97.5)

              75.0 (53.8-96.2)

              2.7 (1.0-7.0)

              0.4 (0.2-1.2)

              0.09

                   < 12 m.o. vs. 12 m.o.-35 m.o.

              Hui, 2008

              71.4 (47.8-95.1)

              55.8 (46.6-64.9)

              1.6 (1.1-2.4)

              0.5 (0.2-1.2)

              0.09b

                   < 12 m.o. vs. 12 m.o.-35 m.o.

              Leventhal, 2008

              81.6 (78.0-85.0)

              75.4 (74.0-76.8)

              3.3 (3.1-3.6)

              0.2 (0.2-0.3)

              <0.001

                   < 12 m.o. vs. 12 m.o.-59 m.o.

              Rex, 2000

              92.9 (79.4-100.0)

              73.7 (53.9.-93.5)

              3.5 (1.6-7.6)

              0.1 (0.0-0.7)

              <0.001

                   < 12 m.o. vs. 12 m.o.-71 m.o.

              Shrader, 2011

              41.9 (27.1-56.6)

              97.9 (95.0-100.0)

              19.7 (4.8-81)

              0.6 (0.5-0.8)

              <0.001

                   < 18 m.o. vs. 18 m.o.-47 m.o.

              Baldwin, 2011

              90.0 (83.0-97.0)

              68.3 (60.6-76.1)

              2.8 (2.2-3.7)

              0.1 (0.1-0.3)

              <0.001

              Ambulatory Status

                    

                   Non-ambulatory vs. ambulatory

              Schwend, 2000

              76.9 (54.0-99.8)

              88.1 (82.4-93.7)

              6.5 (3.7-11.3)

              0.3 (0.1-0.7)

              <0.001

                   Non-ambulatory vs. ambulatory

              Hui, 2008

              71.4 (47.8-95.1)

              69.0 (60.5-77.6)

              2.3 (1.5-3.5)

              0.4 (0.2-1.0)

              0.006

              Health Insurance Status

                    

                   Uninsured vs. insured

              Baldwin, 2011

              7.1 (1.2-13.2)

              90.6 (85.8-95.5)

              0.8 (0.3-2.0)

              1.0 (0.9-1.1)

              0.80

                   Uninsured/Medicaid vs. private

              Shrader, 2011

              76.7 (64.1-89.4)

              51.1 (41.0-61.2)

              1.6 (1.2-2.0)

              0.5 (0.3-0.8)

              0.003

                   Uninsured/Medicaid vs. private

              Blakemore, 1996

              37.5 (13.8-61.2)

              88.5 (76.2-100.0)

              3.3 (0.9-11.2)

              0.7 (0.5-1.1)

              0.06

              Gender

                    

                   Male vs. female

              Schwend, 2000

              61.5 (35.1-88.0)

              27.8 (20.0-35.6)

              0.9 (0.5-1.3)

              1.4 (0.7-2.9)

              0.52

                   Male vs. female

              Baldwin, 2011

              51.4 (39.7-63.1)

              32.4 (24.6-40.2)

              0.8 (0.6-1.0)

              1.5 (1.1-2.1)

              0.002

                   Male vs. female

              Hui, 2008

              50.0 (23.8-76.2)

              37.2 (28.3-46.1)

              1.2 (0.7-2.0)

              0.8 (0.5-1.4)

              0.39

                   Male vs. female

              Rewers, 2005

              62.5 (48.4-76.2)

              31.7 (26.2-37.1)

              0.9 (0.7-1.2)

              1.2 (0.8-1.8)

              0.51

                   Male vs. female

              Arkader, 2007

              80.0 (55.2-100)

              40.0 (9.6-70.4)

              1.3 (0.7-2.4)

              0.5 (0.1-2.1)

              0.63

              Race

                    

                   Black vs. White/other

              Schwend, 2000

              46.2 (19.1-73.3)

              77.0 (69.6-84.3)

              2.0 (1.0-3.9)

              0.7 (0.4-1.2)

              0.09

                   Black vs. White/Hispanic/other

              Rewers, 2005

              54.2 (40.1-68.3)

              44.2 (38.4-50.1)

              1.0 (0.7-1.3)

              1.0 (0.7-1.4)

              0.88

              aComputed using two-sided Fisher’s exact test. P-values of 0.05 or less were considered statistically significant and are presented in bold text.

              b P-value reported in Hui [30] paper was 0.037 but calculated to be 0.09 using two-sided Fisher’s exact test.

              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].

              History characteristics associated with abuse

              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].
              Table 5

              Association of History & Examination Characteristics with Likelihood of Abuse

              Characteristics

              Study

              Sensitivity (95%CI)

              Specificity (95%CI)

              LR+ (95% CI)

              LR- (95% CI)

              p a

              Reported History of Trauma c

                    

                   Suspicious vs. non-suspicious

              Blakemore, 1996

              68.7 (46.0-91.5)

              88.0 (75.3-100.0)

              5.7 (1.9-17.4)

              0.4 (0.2-0.7)

              <0.001

                   Suspicious vs. non-suspicious

              Baldwin, 2011

              32.9 (21.9-43.9)

              95.7 (92.3-99.1)

              7.6 (3.2-17.8)

              0.7 (0.6-0.8)

              <0.001

                   Suspicious vs. non-suspicious

              Hui, 2008

              71.4 (47.8-95.1)

              97.3 (94.4-100.0)

              26.9 (8.4-86.2)

              0.3 (0.1-0.7)

              <0.001

                   Unknown vs. known history

              Shrader, 2011

              39.5 (24.9-54.1)

              94.7 (90.1-99.2)

              7.4 (2.9-18.8)

              0.6 (0.5-0.8)

              <0.001

                   Unwitnessed vs. witnessed

              Blakemore, 1996

              43.8 (19.4-68.1)

              72.0 (54.4 -89.6)

              1.6 (0.7-3.6)

              0.8 (0.5-1.3)

              0.33

                   Witnessed by non-parent (yes vs. no)

              Blakemore, 1996

              18.8 (0.0-37.9)

              48.0 (28.4-67.6)

              0.4 (0.1-1.1)

              1.7 (1.1-2.7)

              0.05

                   History of fall vs. other history

              Blakemore, 1996

              93.8 (81.9-100.0)

              26.9 (9.9-44.0)

              1.3(1.0-1.7)

              0.2 (0.0-1.7)

              0.13

              Reported time from injury to care

                    

                   Delay >24 hours vs. no delay

              Hui, 2008

              42.9 (16.9-68.8)

              92.0 (87.0-97.0)

              5.4 (2.3-12.9)

              0.6 (0.4-1.0)

              0.002

              Location Injury Occurred

                    

              Home/unknown vs. public placed

              Rewers, 2005

              97.9 (93.9-100)

              22.5 (17.65-27.4)

              1.2-1.4

              0.1 (0.0-0.7)

              <0.001

              Additional Injuries

                    

                   Fractures, bruises, or SDHs (yes vs. no)

              Hui, 2008

              42.9 (16.9-68.8)

              92.0 (87.0-97.0)

              3.7 (1.7-8.2)

              0.6 (0.4-1.0)

              0.007

                   Bruises vs. no bruises

              Pierce, 2005

              100 (NAe)

              72.0 (NAe)

              3.6 (1.9-6.7)

              0.0 (NA)

              0.01 b

                   Current polytraumae (yes vs. no)

              Baldwin, 2011

              52.9 (41.2-64.6)

              92.1 (87.6-96.6)

              6.7 (3.6-12.3)

              0.5 (0.4-0.7)

              <0.001

                   Occult injury on imaging (yes vs. no)

              Pierce, 2005

              75.0 (19.4-99.4)

              100.0 (86.0-100.0)

              NA

              0.3(0.0-1.4)

              0.001

                   Any

              Shrader, 2011

              20.9 (8.8-33.1)

              90.4 (84.5-96.4)

              2.2 (0.9-5.1)

              0.9 (0.7-1.0)

              0.10

              aComputed using two-sided Fisher’s exact test. P-values of 0.05 or less were considered statistically significant and are presented in bold text.

              b P-value reported in Pierce [31] was 0.055 but computed as 0.010 using two-sided Fisher’s exact test.

              cDefinition of suspicious history generally included no known history of trauma, or a history of trauma that was unwitnessed or considered inconsistent with the injury.

              dPublic place included locations categorized as public, recreation, and street.

              eCurrent polytrauma was defined as another concurrent long bone, clavicle, axial fracture, or other body system injury that would require hospitalization.

              Additional injuries associated with abuse

              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, 3032]. Universal screening for additional injuries was, however, not performed in any of these studies.

              Fracture characteristics associated with abuse

              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].
              Table 6

              Association of fracture characteristics with likelihood of abuse

              Characteristics

              Study

              Sensitivity (95%CI)

              Specificity (95%CI)

              LR+ (95% CI)

              LR- (95% CI)

              p a

              General Fracture Position

                    

                   Diaphyseal vs. all other

              Hui, 2008

              78.6 (57.1-100.0)

              36.3 (27.4-45.1)

              1.2 (0.9-1.7)

              0.6 (0.2-1.7)

              0.38

                   Diaphyseal vs. all other

              Rewers, 2005

              58.3 (44.4-72.3)

              22.4 (17.5-27.3)

              0.8 (0.6-1.0)

              1.9 (1.2-2.8)

              0.007

                   Diaphyseal vs. all other

              Baldwin, 2011

              44.4 (33.0-55.9)

              33.8 (25.9-41.7)

              0.7 (0.5-0.9)

              1.6 (1.2-2.2)

              0.003

                   Subtrochanteric vs. all other

              Hui, 2008

              7.1 ( 0.0-20.6)

              95.6 (91.8-99.4)

              0.5 (0.1-1.8)

              1.2 (0.9-1.5)

              0.51

                   Subtrochanteric vs. all other

              Baldwin, 2011

              19.4 (10.3-28.6)

              86.3 (80.6-92.0)

              1.4 (0.8-2.7)

              0.9 (0.8-1.1)

              0.32

                   Distal metaphyseal vs. all other

              Hui, 2008

              14.3 (0.0-32.6)

              70.8 (62.4-79.2)

              0.5 (0.1-1.8)

              1.2 (0.9-1.5)

              0.35

                   Distal metaphyseal vs. all other

              Baldwin, 2011

              36.1 (25.0-47.2)

              79.9 (73.2-86.5)

              1.8 (1.1-2.8)

              0.8 (0.7-1.0)

              0.02

              Diaphyseal Fracture Position

                    

                   Distal vs. mid/proximal

              Schwend, 2000

              53.8 (26.7-80.9)

              88.4 (82.7-94.1)

              4.7 (2.3-9.4)

              0.5 (0.3-0.9)

              0.001

              Diaphyseal Fracture Type

                    

                   Spiral vs. non-spiral

              Blakemore, 1996 11119961996

              68.8 (46.0-91.5)

              11.5 (0.0-23.8)

              0.8 (0.5-1.1)

              2.7 (0.7-9.8)

              0.23

                   Transverse vs. all other

              Pierce, 2005

              75.0 (32.6-100.0)

              84.0 (69.6-98.4)

              4.7 (1.6-13.6)

              0.3 (0.1-1.6)

              0.03

              Fracture Side

                    

                   Left vs. right

              Schwend, 2000

              58.3 (30.4-86.2)

              39.5 (30.9-48.1)

              1.0 (0.6-1.6)

              1.1 (0.5-2.1)

              1.00

                   Left vs. right

              Blakemore, 1996

              37.5 (13.8-61.2)

              46.2 (27.0-65.3)

              0.7 (0.3-1.4)

              1.4 (0.8-2.4)

              0.35

                   Left vs. right

              Hui, 2008

              57.1 (31.2-83.1)

              52.2 (43.0-61.4)

              1.2 (0.7-2.0)

              0.8 (0.4-1.5)

              0.58

                   Left vs. right

              Arkader, 2007

              60.0 (29.6-90.4)

              30.0 (1.6-58.4)

              0.9 (0.4-1.6)

              1.3 (0.4-4.5)

              1.00

              Bilateral

                    

                   Bilateral vs. unilateral

              Schwend, 2000

              7.7 (0.0-22.2)

              98.4 (96.2-100.0)

              4.8 (0.5-49.9)

              0.9 (0.8-1.1)

              0.26

              aComputed using two-sided Fisher’s exact test. P-values of 0.05 or less were considered statistically significant and are presented in bold text.

              Discussion

              Our review identified 24 studies providing information about prevalence of abuse among young children with femur fractures. The confirmed or suspected abuse prevalence varied widely across the studies with point estimates ranging from 1.5% - 68.3%, reflecting the heterogeneity of the studies’ populations and methodologies. The structure of this review, focusing on smaller groups of studies with similar populations and stricter child abuse determination methods, produced narrower ranges of abuse proportions.

              Our results indicate that there is substantial risk of abuse (16.7%- 35.5%) in children <12 months old with non-MVC-related femur fractures. The abuse prevalence in this age group was lower, however, than the 50.1% (95% CI, 34.1-66.1) reported for children <18 months in a recent meta-analysis by Maguire et al., [8] likely due to differences in search strategy and inclusion criteria. We included accidental trauma as well as abusive trauma search terms to prevent bias towards studies focused on abusive fractures alone. In addition, we limited our search to articles published in 1990 or later, while Maguire et al. included earlier papers. Although the shorter time frame for our review may have excluded relevant papers, we included only papers published after 1990 for two important reasons. First, data suggest that the abuse probability among children with fractures has changed over time [29]. Second, the subject matter is explored mostly in retrospective studies which depend upon clinical evaluation practices that may have changed over time. Child abuse is a relatively new field; many pediatric centers did not establish child abuse programs until after 1990 [37]. Finally, our estimates for prevalence of abusive femur fractures included data from cohort and cross-sectional studies, while Maguire et al. also extrapolated from case–control studies. Although the two reviews provide different point estimates, both found a high prevalence of abuse in children <12 months old with femur fractures.

              The results of this review also emphasize the significance of reported history and the presence of additional injuries for abuse determination in young children with femur fractures. Femur fracture type and position were, in general, not significantly associated with abuse, with the exception that diaphyseal fractures and mid/proximal diaphyseal fractures particularly might be associated with decreased abuse likelihood. On the other hand, distal metaphyseal femur fractures may indicate increased likelihood of abuse.

              This review has three principal limitations. First, the studies were mostly retrospective in design and not all subjects were assessed consistently for clinical findings, potentially contributing to detection bias. Second, as there is no clear gold standard for diagnosing child abuse, there is concern for circular reasoning in the identification of factors associated with increased risk of abuse. Third, influence of the abuse determination method on abuse prevalence estimates was evident: studies employing more stringent criteria for making a determination of abuse reported lower abuse rates than studies with less stringent criteria. This observation is highlighted in a study where 38% of cases were reported to CPS for suspected abuse, but only 2% were confirmed secondary to abuse at court hearings [22]. Another key concern lies in the potential differences across physicians in calibration of their determinations of abuse [38, 39]. As a result, the variability of findings of abuse might reflect the spectrum of criteria for those determinations, rather than an underlying variation in the prevalence of abuse if all studies applied the same criteria for determining abusive fractures.

              The results of this review underscore the need for prospective studies of children with femur fractures to allow for more accurate estimation of abuse risk in this population. In such studies, standardized evaluation and data collection procedures would be utilized to minimize detection bias and circular reasoning. Lacking a gold standard method for determining abuse, these ideal studies would report whether abuse was diagnosed using specific stated case criteria, as well as whether abuse was confirmed by CPS or court proceedings.

              Conclusions

              This comprehensive literature review underscores the high prevalence of abuse among children <12 months old with femur fractures. In addition, non-ambulatory status, a suspicious history, and the presence of additional injuries were associated with increased likelihood of abusive femur fracture in multiple studies. No significant association was found between probability of determination of abuse and the following characteristics: fracture side; spiral fracture type; bilateral fractures. Additional prospective studies are needed to further elucidate the characteristics that affect abuse probability in children with femur fractures.

              Abbreviations

              cps: 

              Child protective services

              mvc: 

              Motor vehicle crash.

              Declarations

              Acknowledgements

              We would like to acknowledge Maria de Fatima Reyes for her assistance with identification and screening of abstracts of studies obtained from search databases.

              Financial disclosure

              This review was funded in part by grant 1K23HD071967-01from the National Institute of Child Health and Human Development (NICHD). Dr. Joanne Wood’s institution has received payment for expert witness court testimony that Dr. Wood has provided in cases of suspected child abuse.

              Authors’ Affiliations

              (1)
              Division of General Pediatrics and PolicyLab, The Children’s Hospital of Philadelphia
              (2)
              Leonard Davis Institute of Health Economics, Colonial Penn Center
              (3)
              Department of Pediatrics, Perelman School of Medicine at the University of Pennsylvania
              (4)
              Department of Biostatistics and Epidemiology, Perelman School of Medicine at the University of Pennsylvania

              References

              1. Galano GJ, Vitale MA, Kessler MW, Hyman JE, Vitale MG: The most frequent traumatic orthopaedic injuries from a national pediatric inpatient population. J Pediatr Orthop 2005,25(1):39–44.PubMed
              2. Baldwin K, Pandya NK, Wolfgruber H, Drummond DS, Hosalkar HS: Femur fractures in the pediatric population: abuse or accidental trauma? Clin Orthop Relat Res 2011,469(3):798–804. 10.1007/s11999-010-1339-zPubMed CentralPubMedView Article
              3. Oral R, Blum KL, Johnson C: Fractures in young children: are physicians in the emergency department and orthopedic clinics adequately screening for possible abuse? Pediatr Emerg Care 2003,19(3):148–153.PubMed
              4. Dalton HJ, Slovis T, Helfer RE, Comstock J, Scheurer S, Riolo S: Undiagnosed abuse in children younger than 3 years with femoral fracture. Am J Dis Child 1990,144(8):875–878.PubMed
              5. Trokel M, Waddimba A, Griffith J, Sege R: Variation in the diagnosis of child abuse in severely injured infants. Pediatrics 2006,117(3):722–728. 10.1542/peds.2004-2731PubMedView Article
              6. Wood JN, Feudtner C, Medina SP, Luan X, Localio R, Rubin DM: Variation in occult injury screening for children with suspected abuse in selected US children’s hospitals. Pediatrics 2012,130(5):853–860. 10.1542/peds.2012-0244PubMed CentralPubMedView Article
              7. Ravichandiran N, Schuh S, Bejuk M, Al-Harthy N, Shouldice M, Au H, Boutis K: Delayed identification of pediatric abuse-related fractures. Pediatrics 2010,125(1):60–66. 10.1542/peds.2008-3794PubMedView Article
              8. Maguire S: What does the recent literature add to the identification and investigation of fractures in child abuse: an overview of review updates 2005–2013. Evid-Based Child Health: A Cochrane Rev J 2013, 8:2044–2057.View Article
              9. Anderson WA: The significance of femoral fractures in children. Ann Emerg Med 1982,11(4):174–177. 10.1016/S0196-0644(82)80492-7PubMedView Article
              10. Leventhal JM, Martin KD, Asnes AG: Incidence of fractures attributable to abuse in young hospitalized children: results from analysis of a United States database. Pediatrics 2008,122(3):599–604. 10.1542/peds.2007-1959PubMedView Article
              11. Worlock P, Stower M, Barbor P: Patterns of fractures in accidental and non-accidental injury in children: a comparative study. Br Med J (Clin Res Ed) 1986,293(6539):100–102. 10.1136/bmj.293.6539.100View Article
              12. Taitz J, Moran K, O’Meara M: Long bone fractures in children under 3 years of age: Is abuse being missed in Emergency Department presentations? J Paediatr Child Health 2004,40(4):170–174. 10.1111/j.1440-1754.2004.00332.xPubMedView Article
              13. Leventhal JM, Martin KD, Asnes AG: Fractures and traumatic brain injuries: abuse versus accidents in a US database of hospitalized children. Pediatrics 2010,126(1):e104-e115. 10.1542/peds.2009-1076PubMedView Article
              14. OCEBM Levels of Evidence Working Group: The Oxford Levels of Evidence 2. 2011. http://​www.​cebm.​net/​index.​aspx?​o=​5653. Accessed 04/24/2014
              15. Howick J, Chalmers I, Glasziou P, Greenhalgh T, Heneghan C, Liberati A, Moschetti I, Phillips B, Thornton H: The 2011 Oxford CEBM Levels of Evidence (Introductory Document). 2011. http://​www.​cebm.​net/​index.​aspx?​o=​5653. Accessed 04/29/2014
              16. Maguire S, Mann MK, Sibert J, Kemp A: Are there patterns of bruising in childhood which are diagnostic or suggestive of abuse? A systematic review. Arch Dis Child 2005,90(2):182–186. 10.1136/adc.2003.044065PubMed CentralPubMedView Article
              17. Simel DL, Samsa GP, Matchar DB: Likelihood ratios with confidence: sample size estimation for diagnostic test studies. J Clin Epidemiol 1991,44(8):763–770. 10.1016/0895-4356(91)90128-VPubMedView Article
              18. Capra L, Levin AV, Howard A, Shouldice M: Characteristics of femur fractures in ambulatory young children. Emerg Med J 2013,30(9):749–753. doi:10.1136/emermed-2012–201547. Epub 2012 Oct 4 10.1136/emermed-2012-201547PubMedView Article
              19. Loder RT, O’Donnell PW, Feinberg JR: Epidemiology and mechanisms of femur fractures in children. J Pediatr Orthop 2006,26(5):561–566. 10.1097/01.bpo.0000230335.19029.abPubMedView Article
              20. Thomas SA, Rosenfield NS, Leventhal JM, Markowitz RI: Long-bone fractures in young children: distinguishing accidental injuries from child abuse. Pediatrics 1991,88(3):471–476.PubMed
              21. Kowal-Vern A, Paxton TP, Ros SP, Lietz H, Fitzgerald M, Gamelli RL: Fractures in the under-3-year-old age cohort. Clin Pediatr 1992,31(11):653–659. 10.1177/000992289203101103View Article
              22. Blakemore LC, Loder RT, Hensinger RN: Role of intentional abuse in children 1 to 5 years old with isolated femoral shaft fractures. J Pediatr Orthop 1996,16(5):585–588. 10.1097/01241398-199609000-00007PubMedView Article
              23. Hinton RY, Lincoln A, Crockett MM, Sponseller P, Smith G: Fractures of the femoral shaft in children. Incidence, mechanisms, and sociodemographic risk factors. J Bone Joint Surg Am 1999,81(4):500–509.PubMed
              24. Rex C, Kay PR: Features of femoral fractures in nonaccidental injury. J Pediatr Orthop 2000,20(3):411–413.PubMed
              25. Scherl SA, Miller L, Lively N, Russinoff S, Sullivan CM, Tornetta P 3rd: Accidental and nonaccidental femur fractures in children. Clin Orthop Relat Res 2000, 376:96–105.PubMedView Article
              26. Schwend RM, Werth C, Johnston A: Femur shaft fractures in toddlers and young children: rarely from child abuse. J Pediatr Orthop 2000,20(4):475–481.PubMed
              27. Coffey C, Haley K, Hayes J, Groner JI: The risk of child abuse in infants and toddlers with lower extremity injuries. J Pediatr Surg 2005,40(1):120–123. 10.1016/j.jpedsurg.2004.09.003PubMedView Article
              28. Arkader A, Friedman JE, Warner WC Jr, Wells L: Complete distal femoral metaphyseal fractures: a harbinger of child abuse before walking age. J Pediatr Orthop 2007,27(7):751–753. 10.1097/BPO.0b013e3181558b13PubMedView Article
              29. Leventhal JM, Larson IA, Abdoo D, Singaracharlu S, Takizawa C, Miller C, Goodman TR, Schwartz D, Grasso S, Ellingson K: Are abusive fractures in young children becoming less common? Changes over 24 years. Child Abuse Negl 2007,31(3):311–322. 10.1016/j.chiabu.2006.07.004PubMedView Article
              30. Hui C, Joughin E, Goldstein S, Cooper N, Harder J, Kiefer G, Parsons D, Howard J: Femoral fractures in children younger than three years: the role of nonaccidental injury. J Pediatr Orthop 2008,28(3):297–302. 10.1097/BPO.0b013e3181653bf9PubMedView Article
              31. Pierce MC, Bertocci GE, Janosky JE, Aguel F, Deemer E, Moreland M, Boal DK, Garcia S, Herr S, Zuckerbraun N, Vogeley E: Femur fractures resulting from stair falls among children: an injury plausibility model. Pediatrics 2005,115(6):1712–1722. 10.1542/peds.2004-0614PubMedView Article
              32. Shrader MW, Bernat NM, Segal LS: Suspected nonaccidental trauma and femoral shaft fractures in children. Orthopedics 2011,34(5):01477447–20110317.
              33. Jeerathanyasakun Y, Hiranyavanitch P, Bhummichitra D, Sukswai P, Kovitvanitcha D, Thumkunanon V: Causes of femoral shaft fracture in children under five years of age. J Med Assoc Thai 2003,86(3):S661-S666.PubMed
              34. Heideken J, Svensson T, Blomqvist P, Haglund-Akerlind Y, Janarv PM: Incidence and trends in femur shaft fractures in Swedish children between 1987 and 2005. J Pediatr Orthop 2011,31(5):512–519. 10.1097/BPO.0b013e31821f9027PubMedView Article
              35. Rewers A, Hedegaard H, Lezotte D, Meng K, Battan FK, Emery K, Hamman RF: Childhood femur fractures, associated injuries, and sociodemographic risk factors: a population-based study. Pediatrics 2005,115(5):e543-e552. 10.1542/peds.2004-1064PubMedView Article
              36. Banaszkiewicz PA, Scotland TR, Myerscough EJ: Fractures in children younger than age 1 year: importance of collaboration with child protection services. J Pediatr Orthop 2002,22(6):740–744.PubMed
              37. National Association of Children’s Hospitals and Related Insititutions: 2005 survey findings: children’s hospitals child abuse services. In NACHRI Edited by: Alexandra VA. 2006, 1–16.
              38. Levi BH, Brown G: Reasonable suspicion: a study of Pennsylvania pediatricians regarding child abuse. Pediatrics 2005,116(1):e5-e12. 10.1542/peds.2004-2649PubMedView Article
              39. Lichenstein S, Fischoff B, Phillips L: Calibration of probabilities: The state of the art to 1980. In Judgment Under Uncertainty: Heuristics and Biases. Edited by: Kahneman D, SLovic P, Tversky A. Cambridge, UK: Cambridge University Press; 1982:307–334.
              40. Pre-publication history

                1. The pre-publication history for this paper can be accessed here: http://​www.​biomedcentral.​com/​1471-2431/​14/​169/​prepub

              Copyright

              © Wood et al.; licensee BioMed Central Ltd. 2014

              This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://​creativecommons.​org/​licenses/​by/​2.​0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://​creativecommons.​org/​publicdomain/​zero/​1.​0/​) applies to the data made available in this article, unless otherwise stated.