Clinical Radiology 69 (2014) e512ee516
Contents lists available at ScienceDirect
Clinical Radiology journal homepage: www.clinicalradiologyonline.net
Who are we missing? Too few skeletal surveys for children with humeral and femoral fractures S.C. Shelmerdine a, *, R. Das b, M.D. Ingram c, S. Negus b a
Department of Radiology, Great Ormond Street Hospital, London, UK Department of Radiology, St George’s Hospital, Blackshaw Road, London, UK c Department of Radiology, Royal Surrey County Hospital, Guildford, UK b
article in formation Article history: Received 6 July 2014 Received in revised form 13 August 2014 Accepted 20 August 2014
AIM: To determine the potential shortfall in skeletal survey referral for children presenting with an acute non-supracondylar humeral or femoral fracture. MATERIALS AND METHODS: Plain radiograph reports were reviewed retrospectively using the radiology information system database over a 5 year study period (May 2008e2013) in children under 18 months of age who presented with an acute fracture. Subsequent skeletal survey referral was used as a surrogate marker for further investigation of child abuse. Application of robust meta-analysis derived probability data regarding likelihood of child abuse as a cause of non-supracondylar humeral or femoral fracture was applied. An estimation of the expected number of cases of abuse, with shortfall in skeletal survey referrals, was then calculated. RESULTS: There were 288 fractures in 281 children. Three children presented with multiple fractures and were considered separately in the present data. The mean patient age was 10.5 months. Nine (3%) non-supracondylar humeral fractures were identified of which four cases may have been due to non-accidental injury (NAI). One (11%) of these patients was referred for a skeletal survey indicating a potential shortfall of three referrals. Twenty-five (9%) femoral fractures were identified of which 13 cases may have been due to NAI, with six (24%) referrals for skeletal surveys generated. This indicates a potential shortfall of seven referrals. CONCLUSION: The present study serves as a current analysis of practice within a tertiary paediatric referral centre. There appeared to be local under-investigation of NAI. Improved child protection education and awareness programmes have now been introduced. Ó 2014 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Introduction Differentiating non-accidental injury (NAI) from accidental injury (AI) in the paediatric population is challenging and if misdiagnosed, may be fatal.1 The National Society for Protection of Children Charity (NSPCC) data on NAI prevalence have suggested that approximately 7% of all children * Guarantor and correspondent: S.C. Shelmerdine, Department of Radiology, Great Ormond Street Hospital, London, UK. Tel.: þ44 0 7989 984 955. E-mail address: [email protected] (S.C. Shelmerdine).
suffer a degree of serious physical abuse2 during their childhood, with non-ambulatory children having sustained long-bone fractures being at particularly high risk.3e6 Recently published meta-analysis data by Maguire et al.7 (an update to the analysis performed by Kemp et al. in 20088), have calculated positive predictive values (PPVs) from several studies for confirmed or suspected abuse of 50.1% (95% CI: 34.1e66.1) in femoral fractures (excluding road traffic accidents) and 43.8% (95% CI: 27.6e59.9) in humeral fractures (excluding supracondylar injuries) in children aged 0e18 months.
http://dx.doi.org/10.1016/j.crad.2014.08.014 0009-9260/Ó 2014 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
S.C. Shelmerdine et al. / Clinical Radiology 69 (2014) e512ee516
Despite such robust and widely published data, inconsistencies and underreporting of child abuse still widely exists.9,10 The aim of the present study was to determine the possible shortfall in skeletal survey referral for children presenting with an acute non-supracondylar humeral or femoral fracture by applying literature-derived meta-analysis probability data. The hypothesis was that there is local under-investigation of NAI.
Materials and methods Plain radiograph reports were reviewed retrospectively by conducting a search of the radiology information system (RIS, a radiological reporting database) for the terms “fracture”, “bony”, “injury”, “break”, and “accidental” over a 5 year study period (1 May 2008e2013) in children under 18 months of age. The radiographic reports were reviewed by the authors, with all those detailing the presence of an acute fracture included for further analysis. Multiple attendances, demographic data, and subsequent referral for skeletal survey investigations were noted. The incidence of skeletal survey referral was used as a marker for further investigation of child abuse. Children presenting with multiple fractures were considered separately in the present results from children who presented with an isolated solitary fracture due to the much higher probability of NAI in this subset.11 Children were excluded from the study if primary imaging of the fracture was conducted at another hospital (where referral and investigations regarding child abuse may have been instigated for which the authors would not be aware of), related to birth trauma, identified on a postmortem skeletal survey, or in cases of known sudden infant death syndrome (SIDS). The meta-analysis-derived probability data7 for nonsupracondylar and humeral fractures were applied to the present data to calculate the number of suspected fractures, which may have been the result of NAI. From this, the estimated number of shortfalls in referrals for skeletal surveys was determined. In all cases where isolated nonsupracondylar and femoral fractures were sustained and a skeletal survey referral was not performed, the electronic patient records were reviewed to identify whether any clinical decision not to investigate for NAI had been documented (such as a witnessed accidental injury). As the evaluation of the present data was considered to reflect an audit of departmental practice, specific ethical and institutional review board approval was not required.
Results There were 288 fractures in 281 children. The mean patient age was 10.5 months. Three children presented with multiple fractures and were considered separately in the results. Two hundred and seventy-eight fractures were solitary in presentation. Of the isolated fractures identified at presentation, 195 (70%) fractures were identified within a long bone of which 107 (55%) involved the upper limb and
e513
88 (45%) involved the lower limb. Of the remaining 83 fractures that did not occur in the long bones, 44 (16%) clavicular, 19 (7%) skull, and 20 (7%) rib fractures were identified (Table 1). Twenty-seven (10%) patients were referred for a skeletal survey.
Multiple fractures Three patients presented with multiple fractures ranging between three to five fractures in each. Of these patients, one was referred for a skeletal survey. This child initially presented with seizures and was found to have subdural haemorrhages of differing ages at MRI as well as multiple rib fractures on a chest radiograph. A skeletal survey was performed demonstrating further fractures of the femur, tibia, fibula, and several metatarsal bones, without any definite radiographic stigmata of underlying skeletal dysplasia or metabolic bone disease. Social services were contacted and child-safeguarding issues were raised. Of the other two patients not referred for a skeletal survey, one patient was born premature at 23 weeks and noted to have severely osteopaenic bones on multiple radiographs with fractures of the femur, tibia, and fibula. This was thought to be in keeping with the biochemical and clinical diagnosis of metabolic bone disease of prematurity. A short course of vitamin D replacement therapy, cholecalciferol, and phosphate were prescribed. The multiple fractures were treated conservatively. There was no clinical suspicion of NAI. The other patient was also born premature at 28 weeks and found to suffer from metabolic bone disease based on clinical and biochemical investigations. Despite treatment with cholecalciferol and phosphate, the child developed osteopaenia and fractures of the femur, tibia, and fibula. The fractures were treated conservatively. There was no clinical suspicion of NAI at the time.
Table 1 Summary of the type and number of isolated fractures in children aged 0e18 months presenting to St. George’s Hospital, London over a 5-year study period from May 2008e2013. Type of fracture
Non-appendicular fractures Skull Rib Clavicle Pelvis Upper-limb fractures Forearm (radius/ulna) Supracondylar humerus Non-supracondylar forearm Digit/hand Upper limb total Lower-limb fractures Femur Tibia/fibula Foot Lower limb total Grand total of all fractures
No. of fractures Skeletal survey (% of all fracture) performed (% of respective fractures) 19 (7) 20 (7) 44 (16) 0
6 (32) 5 (25) 2 (5) 0
61 (22) 18 (6) 9 (3) 19 (7) 107 (38)
5 1 1 0 7
25 (9) 57 (20) 6 (2) 88 (32) 278
6 (23) 1 (2) 0 7 (8) 27
(8) (6) (11) (7)
e514
S.C. Shelmerdine et al. / Clinical Radiology 69 (2014) e512ee516
Table 2 Summary of long-bone fractures over a 5-year study period with application of positive predictive values (PPVs) from meta-analysis data7 and estimated shortfall in skeletal surveys performed in children aged 0e18 months. Fracture type
No. of fractures
Skeletal survey performed (%)
Meta-analysis PPV for suspected or confirmed abuse (95% CI)
Estimated cases of abusea (with 95% CI data)
Shortfall in skeletal surveys performed (with 95% CI data)
Non-supracondylar humerus Femur
9 25
1 (11) 6 (24)
43.8% (27.6e59.9) 50.1% (34.1e66.1)
4 (2e5) 13 (9e17)
3 (1e4) 7 (3e11)
a
Rounded to nearest digit, from PPV applied to the data.
Upper limb fractures Nine non-supracondylar humeral fractures were identified. Applying the PPV rates from the meta-analysis7 (Table 2), four cases may have been due to NAI. Only one (11%) patient was referred for a skeletal survey indicating a potential shortfall of three referrals (between two to five referrals would be expected if applying the upper and lower limits of the 95% confidence interval probabilities to the present data from the meta-analysis). In cases where a skeletal survey was not generated, documentation could be found in one case where this was felt unnecessary and truly accidental.
Lower limb fractures Twenty-five femoral fractures were identified. Applying the meta-analysis derived probabilities, 13 cases may have been due to NAI. Six (24%) referrals for skeletal surveys were generated indicating a potential shortfall of seven referrals (between nine to 17 referrals would be expected if applying the upper and lower limits of the 95% confidence interval probabilities to the present data). Where a skeletal survey referral was not generated, documentation was found in three cases where this was felt unnecessary on clinical grounds (two cases were felt to be truly accidental, one case was later diagnosed with mild osteogenesis imperfecta and the fracture was thought to be due to the underlying condition).
Discussion The present study reveals a discrepancy between expected and actual skeletal survey referrals for children presenting with acute non-supracondylar humeral or femoral fractures. This discrepancy could be due to missed opportunities from a reduced awareness for NAI referral. To the authors’ knowledge, this is the only study whereby data from a meta-analysis have been used to reflect potential under-investigation of NAI in local practice. The inference is supported by the fact that on review of the electronic patient records, there was documentation that safeguarding issues had been considered and felt unnecessary to pursue in one case of a non-supracondylar humeral fracture and in three cases for femoral fractures. Ideally, all cases where further investigation for NAI was not felt necessary should be documented to make it clear that this was not overlooked during patient assessment. This lack of information in the clinical notes raises concern
regarding missed opportunities for referral, but may also represent a limitation in conducting a retrospective review of medical notes. An alternative explanation for the presumed shortfall in the present results may also reflect a regional variation in abuse within the catchment population compared with those reflected in the meta-analysis. Many of the studies included in the meta-analysis were from North America, with only one from the UK (Scotland). Differences in international child abuse rates16,12,13 may, therefore, affect the degree to which the results can be reliably applied to the study population. However, the lack of documentation regarding safeguarding considerations meant that the estimated shortfall from applied meta-analysis data could not be dismissed as an overestimation as an account could not be given for the reasoning behind non-investigation in all cases. Nevertheless, the vast majority of studies from the meta-analysis were derived from urban paediatric tertiary referral centres14e17 much like the present cohort, presumably covering a similarly large geographical area with a diverse range of cases. As such, the data could be applicable to the clinical situation in the present study. The studies analysed in the meta-analysis by Maguire et al.7 were included to produce robust and reliable outcome data. They comprised only of publications that met strict inclusion criteria and underwent rigorous critical appraisal by at least two specialist reviewers, both involved in child protection. However, applying meta-analytic techniques to observational studies (as oppose to randomized, controlled trials, which do not exist in NAI literature) means combining studies with different populations groups and introduces heterogeneity. The authors tried to overcome this criticism by pooling study data using a random effects model to give a conservative estimate for effect. This is reflected by the large confidence intervals they quote. Nevertheless, the probabilities presented do reflect the best available and most up to date published evidence regarding NAI probability. Even when basing the present data on the most conservative estimations, there was still an estimated shortfall of at least one and three skeletal surveys for nonsupracondylar and femoral fractures, respectively. After discovering the significant shortfall in skeletal survey referrals, it was not deemed appropriate to recall all cases of acute long-bone injuries for re-evaluation for several reasons. The data applied dealt with NAI probability without any definite confirmation of abuse having taken place, and secondly, the retrospective nature of the study meant there was a time lag of at least 1 year or more since clinical presentation to data analysis. This would make it
S.C. Shelmerdine et al. / Clinical Radiology 69 (2014) e512ee516
impractical and near impossible to prove or disprove child abuse had occurred and an agreed local decision was made that education would be the best method in taking the present results forward to aid management of future suspected abuse cases. At our institution, St George’s Hospital, London several clinical governance actions have since been implemented. Presentation of current study findings and the introduction of regular NAI teaching sessions at 3 monthly intervals to emergency department (ED) physicians, nurses, radiologists, and paediatricians have helped ensure all specialties remain up-to-date with current local child protection protocols despite the regular throughput of clinical staff. In the paediatric ED, senior review by either an ED registrar or consultant, of all long-bone fractures in children under 2 years of age now offers added support and experience to junior ED team members. Electronic discharge summaries for paediatric patients now all include a prompt box, reminding clinicians to ensure that safeguarding issues have been assessed prior to the patient leaving the department. This method alerts all healthcare staff to think about the wider social implications regarding the patient’s attendance rather than just their presenting medical complaint. Within the radiology department, registrars and consultants now insert safeguarding recommendations to the end of their reports upon detection of fractures in preambulatory children, reminding clinicians for the need to address child-protection issues. The presence of several reminder systems and safety nets should allow for increased awareness and less anxiety amongst inexperienced staff when asking for a specialist opinion. The present study has several limitations. There can be no reference standard test entirely diagnostic for child abuse, and no fracture alone can be used as sole confirmatory evidence. The retrospective nature of the data collection method meant that accurate documentation within the medical records was not always available, and these issues made assessing the true shortfall in skeletal survey investigation for NAI impossible. The application of probability values from a meta-analysis was, therefore, presumed to represent a valid alternative, yet realistic measure. In addition, the present study may not have examined the full extent of skeletal survey referral discrepancies. Meta-analysis derived data were calculated for patients aged 0e18 months old, which may include children who were both non-ambulatory and in the early ambulatory stages. Although NAI would be considered more suspicious in children who are completely non-ambulatory, the medical notes did not always state the motor development stage of the child and the calculated probabilities from the metanalysis did not include any subset analysis to allow for further differentiation. The small patient numbers mean that applied probabilities may have overestimated the degree of probable NAI. The presence of radiograph interpretation errors was also not accounted for, for example, in cases of missed fractures not mentioned in the radiology report. Nevertheless, addendums added to such radiology reports from discrepancy meetings were included in the RIS
e515
database search and would have been flagged for review during analysis. It is the authors’ intention to repeat the study again within the next year to ensure that the changes that have been implemented have helped to improve the number of appropriate referral rates for skeletal surveys and a greater awareness of child abuse. It is hoped that the results of the present study encourage other institutions to undertake a review of their own practices and strongly recommend continued education of their staff. Increased vigilance in the assessment of long-bone fractures in non-ambulatory children is called for, as a significant number of these may be related to NAI, and the present findings suggests that these may be under-referred. In conclusion, long-bone fractures in children under 18 months of age should raise suspicion for child abuse. Based on national statistics and probability data from metaanalysis, the findings of the present study suggest that fewer than expected skeletal surveys are performed for investigation of NAI. The implication of this finding is whether opportunities have been missed in addressing cases of child abuse at our institution (St. George’s Hospital, London). With increased education and changes in clinical and radiological practice the present authors have tried to improve awareness of this condition. Other institutions should undertake a similar review of their local practice to ensure no child is placed under unnecessary danger.
Acknowledgements The authors acknowledge Mr Rizwan Piracha, Radiology Database Clerk at St George’s Hospital, for his help in conducting computerized database searches of the hospital’s RIS to generate patient lists for the authors to analyse.
References 1. Parton N, Berridge D. Child protection in England. In: Gilbert N, Parton N, Skivenes M, editors. Child protection systems: international trends and orientations. New York: Oxford University Press; 2011. p. 60e88. 2. May-Chahal C, Cawson P. Measuring child maltreatment in the United Kingdom: a study of the prevalence of child abuse and neglect. Child Abuse Negl 2005;29:969e84. 3. 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;293:100e2. 4. Akbarnia B, Torg JS, Kirkpatrick J, et al. Manifestations of the battered child syndrome. J Bone Joint Surg Am 1974;56:1159e66. 5. Karmazyn B, Lewis ME, Jennings SG, et al. The prevalence of uncommon fractures on skeletal surveys performed to evaluate for suspected abuse in 930 children: should practice guidelines change? AJR Am J Roentgenol 2011;197:W159e63. 6. Loder RT, Bookout C. Fracture patterns in battered children. J Orthop Trauma 1991;5:428e33. 7. Maguire S, Cowley L, Mann M, et al. What does the recent literature add to the identification and investigation of fractures in child abuse: an overview of review updates 2005e2013. Evid Based Child Health 2013;8: 2044e57. 8. Kemp A, Dunstan F, Harrison S, et al. Patterns of skeletal fractures in child abuse: systematic review. BMJ 2008;337:a1518. 9. Jones R, Flaherty EG, Binns HJ, et al. Clinicians’ description of factors influencing their reporting of suspected child abuse: report of the Child
e516
10. 11. 12.
13.
S.C. Shelmerdine et al. / Clinical Radiology 69 (2014) e512ee516
Abuse Reporting Experience Study Research Group. Pediatrics 2008;122: 259e66. Sege RD, Flaherty EG. Forty years later d inconsistencies in reporting of child abuse. Arch Dis Child 2008;93:822e4. Offiah A. Radiological features of child maltreatment. Paediatr Child Health 2012;22:483e9. Trocme NM, Tourigny M, MacLaurin B, et al. Major findings from the Canadian incidence study of reported child abuse and neglect. Child Abuse Negl 2003;27:1427e39. Trogan I, Dessypris N, Moustaki M, et al. How common is abuse in Greece? Studying cases with femoral fractures. Arch Dis Child 2001;85:289e92.
14. Baldwin KD, Pandya NK, Wolfgruber H, et al. Femur fractures in the pediatric population: abuse or accidental trauma? Clin Orthop Relat Res 2011;469:798e804. 15. Pandya NK, Baldwin KD, Wolfgruber H, et al. Humerus fractures in the pediatric population: an algorithm to identify abuse. J Pediatr Orthop B 2010;19:535e41. 16. Coffey C, Haley K, Hayes J, et al. The risk of child abuse in infants and toddlers with lower extremity injuries. J Pediatr Surg 2005;40:120e3. 17. Hui C, Joughin E, Goldstein S, et al. Femoral fractures in children younger than three years: the role of nonaccidental injury. J Pediatr Orthop 2008;28:297e302.
Contents lists available at ScienceDirect
Clinical Radiology journal homepage: www.clinicalradiologyonline.net
Who are we missing? Too few skeletal surveys for children with humeral and femoral fractures S.C. Shelmerdine a, *, R. Das b, M.D. Ingram c, S. Negus b a
Department of Radiology, Great Ormond Street Hospital, London, UK Department of Radiology, St George’s Hospital, Blackshaw Road, London, UK c Department of Radiology, Royal Surrey County Hospital, Guildford, UK b
article in formation Article history: Received 6 July 2014 Received in revised form 13 August 2014 Accepted 20 August 2014
AIM: To determine the potential shortfall in skeletal survey referral for children presenting with an acute non-supracondylar humeral or femoral fracture. MATERIALS AND METHODS: Plain radiograph reports were reviewed retrospectively using the radiology information system database over a 5 year study period (May 2008e2013) in children under 18 months of age who presented with an acute fracture. Subsequent skeletal survey referral was used as a surrogate marker for further investigation of child abuse. Application of robust meta-analysis derived probability data regarding likelihood of child abuse as a cause of non-supracondylar humeral or femoral fracture was applied. An estimation of the expected number of cases of abuse, with shortfall in skeletal survey referrals, was then calculated. RESULTS: There were 288 fractures in 281 children. Three children presented with multiple fractures and were considered separately in the present data. The mean patient age was 10.5 months. Nine (3%) non-supracondylar humeral fractures were identified of which four cases may have been due to non-accidental injury (NAI). One (11%) of these patients was referred for a skeletal survey indicating a potential shortfall of three referrals. Twenty-five (9%) femoral fractures were identified of which 13 cases may have been due to NAI, with six (24%) referrals for skeletal surveys generated. This indicates a potential shortfall of seven referrals. CONCLUSION: The present study serves as a current analysis of practice within a tertiary paediatric referral centre. There appeared to be local under-investigation of NAI. Improved child protection education and awareness programmes have now been introduced. Ó 2014 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Introduction Differentiating non-accidental injury (NAI) from accidental injury (AI) in the paediatric population is challenging and if misdiagnosed, may be fatal.1 The National Society for Protection of Children Charity (NSPCC) data on NAI prevalence have suggested that approximately 7% of all children * Guarantor and correspondent: S.C. Shelmerdine, Department of Radiology, Great Ormond Street Hospital, London, UK. Tel.: þ44 0 7989 984 955. E-mail address: [email protected] (S.C. Shelmerdine).
suffer a degree of serious physical abuse2 during their childhood, with non-ambulatory children having sustained long-bone fractures being at particularly high risk.3e6 Recently published meta-analysis data by Maguire et al.7 (an update to the analysis performed by Kemp et al. in 20088), have calculated positive predictive values (PPVs) from several studies for confirmed or suspected abuse of 50.1% (95% CI: 34.1e66.1) in femoral fractures (excluding road traffic accidents) and 43.8% (95% CI: 27.6e59.9) in humeral fractures (excluding supracondylar injuries) in children aged 0e18 months.
http://dx.doi.org/10.1016/j.crad.2014.08.014 0009-9260/Ó 2014 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
S.C. Shelmerdine et al. / Clinical Radiology 69 (2014) e512ee516
Despite such robust and widely published data, inconsistencies and underreporting of child abuse still widely exists.9,10 The aim of the present study was to determine the possible shortfall in skeletal survey referral for children presenting with an acute non-supracondylar humeral or femoral fracture by applying literature-derived meta-analysis probability data. The hypothesis was that there is local under-investigation of NAI.
Materials and methods Plain radiograph reports were reviewed retrospectively by conducting a search of the radiology information system (RIS, a radiological reporting database) for the terms “fracture”, “bony”, “injury”, “break”, and “accidental” over a 5 year study period (1 May 2008e2013) in children under 18 months of age. The radiographic reports were reviewed by the authors, with all those detailing the presence of an acute fracture included for further analysis. Multiple attendances, demographic data, and subsequent referral for skeletal survey investigations were noted. The incidence of skeletal survey referral was used as a marker for further investigation of child abuse. Children presenting with multiple fractures were considered separately in the present results from children who presented with an isolated solitary fracture due to the much higher probability of NAI in this subset.11 Children were excluded from the study if primary imaging of the fracture was conducted at another hospital (where referral and investigations regarding child abuse may have been instigated for which the authors would not be aware of), related to birth trauma, identified on a postmortem skeletal survey, or in cases of known sudden infant death syndrome (SIDS). The meta-analysis-derived probability data7 for nonsupracondylar and humeral fractures were applied to the present data to calculate the number of suspected fractures, which may have been the result of NAI. From this, the estimated number of shortfalls in referrals for skeletal surveys was determined. In all cases where isolated nonsupracondylar and femoral fractures were sustained and a skeletal survey referral was not performed, the electronic patient records were reviewed to identify whether any clinical decision not to investigate for NAI had been documented (such as a witnessed accidental injury). As the evaluation of the present data was considered to reflect an audit of departmental practice, specific ethical and institutional review board approval was not required.
Results There were 288 fractures in 281 children. The mean patient age was 10.5 months. Three children presented with multiple fractures and were considered separately in the results. Two hundred and seventy-eight fractures were solitary in presentation. Of the isolated fractures identified at presentation, 195 (70%) fractures were identified within a long bone of which 107 (55%) involved the upper limb and
e513
88 (45%) involved the lower limb. Of the remaining 83 fractures that did not occur in the long bones, 44 (16%) clavicular, 19 (7%) skull, and 20 (7%) rib fractures were identified (Table 1). Twenty-seven (10%) patients were referred for a skeletal survey.
Multiple fractures Three patients presented with multiple fractures ranging between three to five fractures in each. Of these patients, one was referred for a skeletal survey. This child initially presented with seizures and was found to have subdural haemorrhages of differing ages at MRI as well as multiple rib fractures on a chest radiograph. A skeletal survey was performed demonstrating further fractures of the femur, tibia, fibula, and several metatarsal bones, without any definite radiographic stigmata of underlying skeletal dysplasia or metabolic bone disease. Social services were contacted and child-safeguarding issues were raised. Of the other two patients not referred for a skeletal survey, one patient was born premature at 23 weeks and noted to have severely osteopaenic bones on multiple radiographs with fractures of the femur, tibia, and fibula. This was thought to be in keeping with the biochemical and clinical diagnosis of metabolic bone disease of prematurity. A short course of vitamin D replacement therapy, cholecalciferol, and phosphate were prescribed. The multiple fractures were treated conservatively. There was no clinical suspicion of NAI. The other patient was also born premature at 28 weeks and found to suffer from metabolic bone disease based on clinical and biochemical investigations. Despite treatment with cholecalciferol and phosphate, the child developed osteopaenia and fractures of the femur, tibia, and fibula. The fractures were treated conservatively. There was no clinical suspicion of NAI at the time.
Table 1 Summary of the type and number of isolated fractures in children aged 0e18 months presenting to St. George’s Hospital, London over a 5-year study period from May 2008e2013. Type of fracture
Non-appendicular fractures Skull Rib Clavicle Pelvis Upper-limb fractures Forearm (radius/ulna) Supracondylar humerus Non-supracondylar forearm Digit/hand Upper limb total Lower-limb fractures Femur Tibia/fibula Foot Lower limb total Grand total of all fractures
No. of fractures Skeletal survey (% of all fracture) performed (% of respective fractures) 19 (7) 20 (7) 44 (16) 0
6 (32) 5 (25) 2 (5) 0
61 (22) 18 (6) 9 (3) 19 (7) 107 (38)
5 1 1 0 7
25 (9) 57 (20) 6 (2) 88 (32) 278
6 (23) 1 (2) 0 7 (8) 27
(8) (6) (11) (7)
e514
S.C. Shelmerdine et al. / Clinical Radiology 69 (2014) e512ee516
Table 2 Summary of long-bone fractures over a 5-year study period with application of positive predictive values (PPVs) from meta-analysis data7 and estimated shortfall in skeletal surveys performed in children aged 0e18 months. Fracture type
No. of fractures
Skeletal survey performed (%)
Meta-analysis PPV for suspected or confirmed abuse (95% CI)
Estimated cases of abusea (with 95% CI data)
Shortfall in skeletal surveys performed (with 95% CI data)
Non-supracondylar humerus Femur
9 25
1 (11) 6 (24)
43.8% (27.6e59.9) 50.1% (34.1e66.1)
4 (2e5) 13 (9e17)
3 (1e4) 7 (3e11)
a
Rounded to nearest digit, from PPV applied to the data.
Upper limb fractures Nine non-supracondylar humeral fractures were identified. Applying the PPV rates from the meta-analysis7 (Table 2), four cases may have been due to NAI. Only one (11%) patient was referred for a skeletal survey indicating a potential shortfall of three referrals (between two to five referrals would be expected if applying the upper and lower limits of the 95% confidence interval probabilities to the present data from the meta-analysis). In cases where a skeletal survey was not generated, documentation could be found in one case where this was felt unnecessary and truly accidental.
Lower limb fractures Twenty-five femoral fractures were identified. Applying the meta-analysis derived probabilities, 13 cases may have been due to NAI. Six (24%) referrals for skeletal surveys were generated indicating a potential shortfall of seven referrals (between nine to 17 referrals would be expected if applying the upper and lower limits of the 95% confidence interval probabilities to the present data). Where a skeletal survey referral was not generated, documentation was found in three cases where this was felt unnecessary on clinical grounds (two cases were felt to be truly accidental, one case was later diagnosed with mild osteogenesis imperfecta and the fracture was thought to be due to the underlying condition).
Discussion The present study reveals a discrepancy between expected and actual skeletal survey referrals for children presenting with acute non-supracondylar humeral or femoral fractures. This discrepancy could be due to missed opportunities from a reduced awareness for NAI referral. To the authors’ knowledge, this is the only study whereby data from a meta-analysis have been used to reflect potential under-investigation of NAI in local practice. The inference is supported by the fact that on review of the electronic patient records, there was documentation that safeguarding issues had been considered and felt unnecessary to pursue in one case of a non-supracondylar humeral fracture and in three cases for femoral fractures. Ideally, all cases where further investigation for NAI was not felt necessary should be documented to make it clear that this was not overlooked during patient assessment. This lack of information in the clinical notes raises concern
regarding missed opportunities for referral, but may also represent a limitation in conducting a retrospective review of medical notes. An alternative explanation for the presumed shortfall in the present results may also reflect a regional variation in abuse within the catchment population compared with those reflected in the meta-analysis. Many of the studies included in the meta-analysis were from North America, with only one from the UK (Scotland). Differences in international child abuse rates16,12,13 may, therefore, affect the degree to which the results can be reliably applied to the study population. However, the lack of documentation regarding safeguarding considerations meant that the estimated shortfall from applied meta-analysis data could not be dismissed as an overestimation as an account could not be given for the reasoning behind non-investigation in all cases. Nevertheless, the vast majority of studies from the meta-analysis were derived from urban paediatric tertiary referral centres14e17 much like the present cohort, presumably covering a similarly large geographical area with a diverse range of cases. As such, the data could be applicable to the clinical situation in the present study. The studies analysed in the meta-analysis by Maguire et al.7 were included to produce robust and reliable outcome data. They comprised only of publications that met strict inclusion criteria and underwent rigorous critical appraisal by at least two specialist reviewers, both involved in child protection. However, applying meta-analytic techniques to observational studies (as oppose to randomized, controlled trials, which do not exist in NAI literature) means combining studies with different populations groups and introduces heterogeneity. The authors tried to overcome this criticism by pooling study data using a random effects model to give a conservative estimate for effect. This is reflected by the large confidence intervals they quote. Nevertheless, the probabilities presented do reflect the best available and most up to date published evidence regarding NAI probability. Even when basing the present data on the most conservative estimations, there was still an estimated shortfall of at least one and three skeletal surveys for nonsupracondylar and femoral fractures, respectively. After discovering the significant shortfall in skeletal survey referrals, it was not deemed appropriate to recall all cases of acute long-bone injuries for re-evaluation for several reasons. The data applied dealt with NAI probability without any definite confirmation of abuse having taken place, and secondly, the retrospective nature of the study meant there was a time lag of at least 1 year or more since clinical presentation to data analysis. This would make it
S.C. Shelmerdine et al. / Clinical Radiology 69 (2014) e512ee516
impractical and near impossible to prove or disprove child abuse had occurred and an agreed local decision was made that education would be the best method in taking the present results forward to aid management of future suspected abuse cases. At our institution, St George’s Hospital, London several clinical governance actions have since been implemented. Presentation of current study findings and the introduction of regular NAI teaching sessions at 3 monthly intervals to emergency department (ED) physicians, nurses, radiologists, and paediatricians have helped ensure all specialties remain up-to-date with current local child protection protocols despite the regular throughput of clinical staff. In the paediatric ED, senior review by either an ED registrar or consultant, of all long-bone fractures in children under 2 years of age now offers added support and experience to junior ED team members. Electronic discharge summaries for paediatric patients now all include a prompt box, reminding clinicians to ensure that safeguarding issues have been assessed prior to the patient leaving the department. This method alerts all healthcare staff to think about the wider social implications regarding the patient’s attendance rather than just their presenting medical complaint. Within the radiology department, registrars and consultants now insert safeguarding recommendations to the end of their reports upon detection of fractures in preambulatory children, reminding clinicians for the need to address child-protection issues. The presence of several reminder systems and safety nets should allow for increased awareness and less anxiety amongst inexperienced staff when asking for a specialist opinion. The present study has several limitations. There can be no reference standard test entirely diagnostic for child abuse, and no fracture alone can be used as sole confirmatory evidence. The retrospective nature of the data collection method meant that accurate documentation within the medical records was not always available, and these issues made assessing the true shortfall in skeletal survey investigation for NAI impossible. The application of probability values from a meta-analysis was, therefore, presumed to represent a valid alternative, yet realistic measure. In addition, the present study may not have examined the full extent of skeletal survey referral discrepancies. Meta-analysis derived data were calculated for patients aged 0e18 months old, which may include children who were both non-ambulatory and in the early ambulatory stages. Although NAI would be considered more suspicious in children who are completely non-ambulatory, the medical notes did not always state the motor development stage of the child and the calculated probabilities from the metanalysis did not include any subset analysis to allow for further differentiation. The small patient numbers mean that applied probabilities may have overestimated the degree of probable NAI. The presence of radiograph interpretation errors was also not accounted for, for example, in cases of missed fractures not mentioned in the radiology report. Nevertheless, addendums added to such radiology reports from discrepancy meetings were included in the RIS
e515
database search and would have been flagged for review during analysis. It is the authors’ intention to repeat the study again within the next year to ensure that the changes that have been implemented have helped to improve the number of appropriate referral rates for skeletal surveys and a greater awareness of child abuse. It is hoped that the results of the present study encourage other institutions to undertake a review of their own practices and strongly recommend continued education of their staff. Increased vigilance in the assessment of long-bone fractures in non-ambulatory children is called for, as a significant number of these may be related to NAI, and the present findings suggests that these may be under-referred. In conclusion, long-bone fractures in children under 18 months of age should raise suspicion for child abuse. Based on national statistics and probability data from metaanalysis, the findings of the present study suggest that fewer than expected skeletal surveys are performed for investigation of NAI. The implication of this finding is whether opportunities have been missed in addressing cases of child abuse at our institution (St. George’s Hospital, London). With increased education and changes in clinical and radiological practice the present authors have tried to improve awareness of this condition. Other institutions should undertake a similar review of their local practice to ensure no child is placed under unnecessary danger.
Acknowledgements The authors acknowledge Mr Rizwan Piracha, Radiology Database Clerk at St George’s Hospital, for his help in conducting computerized database searches of the hospital’s RIS to generate patient lists for the authors to analyse.
References 1. Parton N, Berridge D. Child protection in England. In: Gilbert N, Parton N, Skivenes M, editors. Child protection systems: international trends and orientations. New York: Oxford University Press; 2011. p. 60e88. 2. May-Chahal C, Cawson P. Measuring child maltreatment in the United Kingdom: a study of the prevalence of child abuse and neglect. Child Abuse Negl 2005;29:969e84. 3. 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;293:100e2. 4. Akbarnia B, Torg JS, Kirkpatrick J, et al. Manifestations of the battered child syndrome. J Bone Joint Surg Am 1974;56:1159e66. 5. Karmazyn B, Lewis ME, Jennings SG, et al. The prevalence of uncommon fractures on skeletal surveys performed to evaluate for suspected abuse in 930 children: should practice guidelines change? AJR Am J Roentgenol 2011;197:W159e63. 6. Loder RT, Bookout C. Fracture patterns in battered children. J Orthop Trauma 1991;5:428e33. 7. Maguire S, Cowley L, Mann M, et al. What does the recent literature add to the identification and investigation of fractures in child abuse: an overview of review updates 2005e2013. Evid Based Child Health 2013;8: 2044e57. 8. Kemp A, Dunstan F, Harrison S, et al. Patterns of skeletal fractures in child abuse: systematic review. BMJ 2008;337:a1518. 9. Jones R, Flaherty EG, Binns HJ, et al. Clinicians’ description of factors influencing their reporting of suspected child abuse: report of the Child
e516
10. 11. 12.
13.
S.C. Shelmerdine et al. / Clinical Radiology 69 (2014) e512ee516
Abuse Reporting Experience Study Research Group. Pediatrics 2008;122: 259e66. Sege RD, Flaherty EG. Forty years later d inconsistencies in reporting of child abuse. Arch Dis Child 2008;93:822e4. Offiah A. Radiological features of child maltreatment. Paediatr Child Health 2012;22:483e9. Trocme NM, Tourigny M, MacLaurin B, et al. Major findings from the Canadian incidence study of reported child abuse and neglect. Child Abuse Negl 2003;27:1427e39. Trogan I, Dessypris N, Moustaki M, et al. How common is abuse in Greece? Studying cases with femoral fractures. Arch Dis Child 2001;85:289e92.
14. Baldwin KD, Pandya NK, Wolfgruber H, et al. Femur fractures in the pediatric population: abuse or accidental trauma? Clin Orthop Relat Res 2011;469:798e804. 15. Pandya NK, Baldwin KD, Wolfgruber H, et al. Humerus fractures in the pediatric population: an algorithm to identify abuse. J Pediatr Orthop B 2010;19:535e41. 16. Coffey C, Haley K, Hayes J, et al. The risk of child abuse in infants and toddlers with lower extremity injuries. J Pediatr Surg 2005;40:120e3. 17. Hui C, Joughin E, Goldstein S, et al. Femoral fractures in children younger than three years: the role of nonaccidental injury. J Pediatr Orthop 2008;28:297e302.
