Fatal non-accidental injury in South Africa: A Gauteng hospital’s perspective on the incidence and fracture types in post-mortem skeletal surveys

Background In its severest form, non-accidental injury (NAI) in children is fatal. South Africa has been reported to have double the global average of child homicides. Autopsy is the main investigation in fatal NAI with post-mortem skeletal surveys (PMSS) playing an adjunctive role. Whilst fracture patterns associated with NAI in living patients have been established, this has not been investigated in PMSS in South Africa. Objectives To determine the incidence and characteristics of fractures in suspected fatal NAI cases. To calculate the incidence of fractures according to high-, moderate- and low-specificity fracture locations for NAI. Methods A retrospective review of all PMSS performed between 01 January 2012 and 03 December 2018 was conducted at the Charlotte Maxeke Johannesburg Academic Hospital. Results Of the 73 PMSS, 33 (45.2%) demonstrated fractures. No statistical significance in sex was found: 38 (52.1%) were male and 35 (47.9%) were female (p > 0.05). The mean age of those who sustained fractures was 28 months (standard deviation [s.d.]: 21 months). A total of 115 fractures were sustained, of that the top five bones fractured were the ribs 37 (32.2%), parietal bone 13 (11.3%), ulna 13 (11.3%), femur 13 (11.3%), and radius 11 (9.6%). High-specificity fracture locations accounted for 40/133 (30.1%). Conclusion The fracture types in PMSS were similar to those in live skeletal surveys. Our study’s fracture rate was higher in comparison to international studies. The PMSS is a valuable adjunct to autopsy in detecting occult fractures of the limbs. We recommend that PMSS be performed in suspected fatal NAI cases at least in children up to 24 months of age.


Introduction
Child abuse or non-accidental injury (NAI) in its severest form is fatal. 1 The World Health Organization (WHO) defines child abuse and child maltreatment as: [A]ll forms of physical and/or emotional ill-treatment, sexual abuse, neglect or negligent treatment or commercial or other exploitation, resulting in actual or potential harm to the child's health, survival, development or dignity in the context of a relationship of responsibility, trust or power. 2 resource constrained environment in terms of the availability of imaging equipment, radiographers and radiologists with experience in post-mortem imaging. Post-mortem CT as the primary post-mortem imaging modality has shown promise in some international centres; 14,15 however, this is not currently the standard in SA.
The standard views required in PMSS are debated in the literature and varies depending on the individual country or institutional regulations. 12,16,17,18 In SA, the South African Society of Paediatric Imaging (SASPI) stipulates guidelines for live skeletal surveys (LSS). In Charlotte Maxeke Johannesburg Academic Hospital, modified skeletal survey guidelines from the Red Cross War Memorial Children's Hospital (RCWMCH) in Cape Town, which are similar to the SASPI guidelines, are used for both living and deceased patients (see Table 1).
The procedure of performing a PMSS is more cumbersome and time-consuming than performing a LSS. The postmortem condition of the patient (rigor mortis, decomposition, post autopsy changes) provides unique challenges in obtaining and maintaining the diagnostic quality of the PMSS images. Hughes-Roberts et al. found a larger proportion of significant artefacts in the PMSS when compared to the LSS. 9 Despite these factors, PMSS remains the current standard for occult fracture detection and fracture dating in suspected fatal cases of NAI in SA. 19 The patterns of fractures associated with NAI in living patients have been published internationally. 17,18,20 These have been divided into high-, moderate-and low-specificity groups for NAI as follows: high-specificity fractures include classic metaphyseal lesions, rib fractures (especially posteriormedial), scapular fractures, spinous process fractures and sternal fractures. 17,18,20 Moderate-specificity fractures include multiple fractures (especially bilateral), fractures of different ages, vertebral body fractures and subluxations, digital fractures and complex skull fractures. 17,18,20 Low-specificity fractures comprise subperiosteal new bone formation, clavicular fractures, long bone shaft fractures and linear skull fractures. 17,18,20 In contradistinction, only a single South African study was published in 2007 detailing the fracture patterns sustained in living patients at the RCWMCH. 6 The fracture patterns in PMSS in SA have not been investigated to date.
The paucity of research on the fractures sustained in fatal NAI and the debate regarding the validity of the PMSS provided the motivation for our study framework in a South African tertiary hospital setting.
The aim and objectives of this descriptive study were therefore to determine the incidence and characteristics of the fractures in child fatalities suspected of having been subjected to NAI, and to establish how many fractures occurred in high-, moderate-and low-specificity locations for NAI, as per O'Connor et al. 21 The degree of concordance between two readers of varying reporting experience levels was also assessed.

Materials and methods
A retrospective review of all the consecutive PMSS referred to our hospital by the local forensic pathology services performed over a six-year period (01 January 2012 to 03 December 2018) was conducted. The PMSS of all child fatalities with suspected NAI, those who were found to be deceased in suspicious or unknown circumstances or those who had died unexpectedly, were included in the study. The study included all children from the age of birth to adolescence. The WHO definition of adolescence, those aged between 10 and 19 years, was employed. Where the date of birth was not available, estimation of age was made using a bone age atlas. 22 All cases on film (prior to April 2016) and those saved digitally onto the hospital's picture archiving and communication system (PACS) (after 01 April 2016) were assessed. The views were acquired according to the protocol detailed in Table 1.
Each skeletal survey was read independently by a radiology registrar (year three of study) and a paediatric radiologist (four years of experience) blinded to all clinical details except sex and age. A standardised checklist for reporting was used by the readers for data collection, which assessed demographic details, technical aspects and skeletal survey findings including fracture dating (i.e., acute, healing and healed fractures) according to Paddock et al. 17 Statistical analysis was performed with STATA 15. The descriptive statistics: mean, median, standard deviation (s.d.) and interquartile range (IQR) were used to describe the continuous variables including age and the number of fractures per patient. Frequencies and proportions were used to describe the type of fractures and other categorical variables. Incidence rates were calculated for the number of fractures in the study population, first overall as well as by the type of fracture. Associations between categorical variables were tested using the Fisher's exact test. Cohen's Kappa was used to determine the concordance between readers. Tests were evaluated at a 5% level of significance.

Ethical considerations
Ethical approval was granted by the Human Research Ethics Committee (Medical) of the University of the Witwatersrand. Clearance number: M181112. All data were anonymised.
Overall, 115 fractures were sustained. The mean age of those who sustained at least one fracture was 28 months (s.d.: 21 months). A comparison between the cases who sustained fractures according to the different age groups is illustrated in Figure 1. There was no statistical significance when comparing the ages of those who sustained fractures to those who did not (p > 0.05).
When considering the sex and age comparatively, the mean age of those with at least one fracture in girls was 27 months  Table 2).

Reader agreement
When comparing agreement on the incidence of the top five bones fractured, the reader agreement (Kappa values) was as follows: ribs 0.7, parietal bone 0.87, ulna 0.85, femur 0.87 and radius 0.93. The mean reader agreement (Kappa values) for high-, moderate-and low-specificity fracture locations for NAI was 0.73, 0.9 and 0.8, respectively (see Table 3).

Discussion
This study is the first to document the fracture types and characteristics in deceased children being investigated for suspected fatal NAI in a South African population.
The assessment of the adequacy of the PMSS identified a high number with inadequate views, three-quarters of which were due to poor positioning. This is not an unusual finding in post-mortem imaging, as an audit of skeletal surveys in the United Kingdom (UK) three years after the British Society of Paediatric Radiology (BPSR) standards were published, revealed that only 15% included all views. 23 This reiterates the widespread challenges faced and the need for quality assurance when performing PMSS. 9,23 As these studies were conducted over a number of years with various radiographers and radiologists in    Each skeletal survey is approved by a consultant radiologist.
Because of the small number of paediatric radiologists in SA, and in our hospital specifically, not every skeletal survey can be reviewed by this type of specialist. It is imperative that there is adherence to, and consistency of the imaging generated according to the local protocol. This will facilitate obtaining PMSS of sufficient diagnostic quality to aid accurate interpretation by radiologists with varying levels of experience. 23,24,25 Additionally, a significant number of artefacts were recorded; however, these are similar to that reported by Hughes et al. 9 The study findings were compared specifically to two key studies as detailed in Table 4. We found no differences between the age or sex of those who sustained fractures and those without fractures. The mean age of our study population was younger compared to the van As et al. study; however, much younger children were reported by Hughes et al. 6,9 We have established that there is a much higher yield for fractures on PMSS, almost four times than on LSS in the Cape Town population and five times more than on PMSS in the UK. 6,9 This initial finding should be corroborated with future comparative studies on LSS and PMSS in SA. Despite the differences in these two countries findings, we concur that fatal NAI is more common in younger children and the injuries sustained tend to be more severe. 6,9 The characteristics of fractures in our study were concordant with the literature in that the most common regions fractured were the chest, the skull and the limbs. 4,6,7,20,26 Most articles were based on findings in LSS with a few on PMSS. We demonstrated similar findings to those of Van As et al. in that some high-specificity fracture locations for NAI (metaphyseal corner fractures, scapula and sternal fractures) were rare in this series (see Table 3), however high-specificity fracture locations overall accounted for just over a third of the injuries. Only rib fractures were more common in our series compared to Van As et al., whereas skull fractures were in the majority in the latter. This may show a possible difference in fracture patterns sustained per province in SA. The need for future local studies to establish the spectrum of fracture patterns in our population at multiple centres is emphasised.
In some studies, the need for routine PMSS has been debated. 12 McGraw et al. concluded that the addition of a PMSS added valuable additional information to the autopsy findings in cases of suspected fatal NAI. In his study about post-mortem radiography, 92% of the extremity fractures Five-year-old-male with a distal radial diaphyseal fracture with callus formation (white arrow) indicating a healing fracture on an AP view of the right forearm. This is a low-specificity fracture for non-accidental injury.

FIGURE 6:
Two-year-old female with a linear occipital bone fracture (white arrow) on a lateral skull view. This is a low-specificity fracture for non-accidental injury.
identified were metaphyseal corner fractures. 13 These types of fractures are highly specific for NAI and are often difficult to detect on routine autopsy. 10,13 In this study, however we found fewer metaphyseal corner fractures compared to the studies performed by McGraw et al. and Kleinman et al. 13,27 More importantly, we found that overall there was a high incidence of the limb fractures in our PMSS series. Many of the limb fractures would have otherwise been occult at autopsy, since these regions are not routinely dissected. This and the further findings of multiple fractures, fractures at varying stages of healing and specifically the high yield of high-specificity fracture locations on PMSS in our cohort illustrates the utility of PMSS, as a necessary adjunct in the evaluation of fatal NAI in identifying and documenting the extent of injuries for the ensuing medico-legal proceedings.
We assessed fracture dating which has not been investigated in our population previously. Acute fractures made up over half of those who sustained fractures. This possibly indicates the severity of the injuries sustained and that a proportion of patients do not survive to recover from them. The overall number of fractures in multiple stages of healing could allude to the repeated abuse in these cases. Fracture dating in PMSS remains an important observation to make in suspected cases of NAI, as determining the chronicity of fractures contributes to the medico-legal case by providing evidence of previously sustained injury. 17,18 It is important to report on reader agreement to indicate if the assessment criteria are robust and valid in clinical practice. This has not been assessed in our setting before. This study showed substantial to almost perfect agreement between the readers (see Table 3). The substantial agreement was found in the identification of high-specificity fractures. This could be because of the finite details being overlooked by less experienced readers. This finding speaks to the challenges of offering these much-needed services in a country with a higher prevalence of NAI compared to worldwide, with limited paediatric radiology expertise in PMSS. The need for dissemination of this training and expertise to general radiologists and radiology registrars should therefore be encouraged.

Limitations
The most notable limitation to this study is its retrospective nature which may have introduced bias. The study was conducted at a single centre. Furthermore, no comparison of PMSS to LSS, autopsy findings or the final medico-legal investigation outcome with confirmed cases of fatal child abuse was performed.

Conclusion and recommendations
This study analysed the first and largest number of PMSS in SA to date. In comparison to international studies, the study population's fracture rate was notably higher. We have shown that in cases of suspected NAI, the types of fractures found in PMSS were similar to those seen in LSS, and predominantly occurred in those under 24 months of age. Whereas some of the classically described high-specificity fracture locations for NAI in the international literature, were rare. Future local studies are needed to investigate possible differences in the types of fractures sustained according to region.
Post-mortem skeletal surveys in the South African setting is a valuable adjunct to autopsy in detecting occult fractures of the limbs, which formed a large portion of the fractures that would not have been discovered or routinely dissected at autopsy. We therefore recommend PMSS at least be performed in suspected fatal NAI cases in children up to 24 months of age.
Standardised radiographic technique as per the local guidelines and quality assurance of PMSS is vital to obviate factors that affect interpretation. Furthermore, in our resource limited setting, the encouragement and continuous education and training of radiographers regarding imaging technique, general radiologists and radiology registrars on interpreting PMSS remain key to improving the investigation of these cases.