Transfer of knowledge to diagnose infant abuse and its incidence – a time-series analysis from Sweden

A high diagnostic precision in the detection of child abuse is of utmost societal importance to keep both false positives and false negatives at a minimum [1], as the consequences can be severe in both cases. The evolving field of knowledge on how to detect infant abuse embraces the developments of modern medicine. Battered-child syndrome was first described by the paediatrician Kempe (1962) as exhibiting external signs of injury, fractures (also metaphyseal and ribs), subdural haemorrhage (SDH) and commonly being associated with parental psychiatric factors. Reference was also made to findings described by the radiologist Caffey in 1940s [2]. The neurosurgeon Guthkelch (1971) hypothesised that repeated acceleration/deceleration caused SDH and retinal haemorrhage (RH) in cases of suspected battered-child syndrome without external signs of trauma [3]. This was described by Caffey (1974) as the whiplash shaken infant syndrome [4]. Later, the term shaken baby syndrome/abusive head trauma (SBS/AHT) was coined, described as part of evidence-based medicine and supported by numerous studies (case reports, case-series, case-control studies and consensus/expert opinions) [5, 6]. Algorithms to predict abuse from radiological or ocular findings [7] were provided, to be implemented into medical and legal practice [6].

Transfer of research findings and innovations into medical practice can be distinguished into phases: diffusion, dissemination, implementation and sustainability until obsolescence [8]. Important implementation drivers promoting uptake of research findings into practice include: capacity building and institutionalisation [9], champions and leadership [8, 10], interdisciplinary approaches [11] and a shared agenda between different stakeholders [8, 12]. However, diagnostic work is seldom a straightforward endeavour of applying evidence; professionals’ understanding, established routines and scientific controversies may influence the uptake of new research findings into medical practice [13].

Disruptive reasoning and findings have challenged the evidence on the SBS/AHT paradigm. The proposed shaking mechanism causing SDH and RH has been questioned [14‐20], and differential diagnoses for rib fractures and metaphyseal lesions have been presented [21]. Further, a major bias, circular reasoning, has been identified by two systematic reviews – concluding that there is insufficient scientific evidence to assess the diagnostic accuracy of the triad findings (SDH, RH and encephalopathy) in identifying traumatic shaking [22], and to determine if rib fractures or metaphyseal lesions are caused by abuse [23]. Thus, the state of knowledge on the SBS/AHT paradigm has been, and still is, subject to scientific controversy and can be described as undergoing a paradigm shift [24‐27].

The rationale for this study was to advance the understanding of how health services are implementing evolving research findings in the field of child protection. We have previously reported on an association between SBS/AHT criteria and infant maltreatment diagnosis, and a 10-fold increased incidence of maltreatment diagnosis among infants born 1987–2014 [28]. There is a lack of studies exploring the relationship between scientific/medical controversies on evidence, which play out in scientific journals and grey literature, and the practices of the medical professionals who manage the controversial matters in daily medical practice.

The objective was to analyse the transfer of knowledge on how to detect physical abuse, especially SBS/AHT, and its association to trends in infant abuse diagnoses (maltreatment and assault) during the years 1987 to 2019.

This was a population-based study with a quasi-experimental retrospective design departing from a literature review.

The incidences (3-year moving average) were rather stable for both maltreatment and assault diagnoses during the late 1980s and 1990s. Both maltreatment and assault increased around the turn of the century. Assault levelled out, while maltreatment peaked around 2010–2013 (Fig. 1). In all, 1150 cases were found in 1987–2019, 920 had any diagnosis of maltreatment and 230 had a diagnosis of assault; 45 had both. The overall incidence rate was 32.23 per 100,000 infants; the incidence rate of any maltreatment diagnosis was 25.78 and that for any assault diagnosis was 7.43 (Table 1). A three-fold increase was observed in 2001–2019 compared with in 1987–2000 for both maltreatment and assault, 36.15 and 10.33 versus 11.29 and 3.36, respectively (Table 1). By 2001, when outpatient specialised care was included in the NPR, the incidences of maltreatment and assault were 41.46 and 15.28, respectively; in 2000, they were 11.02 and 2.20, respectively.

The overall mortality rate was 1.40 per 100,000 infants; for maltreatment, it was 0.08 (n = 3), and for assault, it was 2.16 (n = 47). The mortality rate declined by 50% in 2001–2019 compared with in 1987–2000 due to a decline in assault deaths, while maltreatment deaths remained unchanged (Table 1).

Table 2 and Fig. 2 show the transfer of knowledge (diffusion, dissemination and implementation) by year in 1990–2019.

The 1990s was a decade of continuous diffusion and dissemination of general knowledge on child abuse and specifically battered-child syndrome through research, manuals and textbooks. The first mention of SBS came in a short overview in 1994 and SBS was further addressed in a textbook in 1999.

The following two decades were characterised by an increasing transfer of knowledge on child abuse, through diffusion, dissemination and implementation, with a focus on knowledge supportive of SBS (AHT after 2009). Diffusion of new research findings on SBS from the Swedish setting was rare; rather, transfer of knowledge was based on US and UK studies, estimating that close to 100 infants were physically abused annually in Sweden, and only half of them had external signs of abuse, with 1/3 assumed to have been abused by shaking, which allegedly carries the greatest risk of going undetected (Additional file 3 2008:III;4). Dissemination was conducted through textbooks, comments, manuals, conferences (with participation of US and UK representatives) and teaching, emphasising the multidisciplinary diagnostic approach: ‘The radiologist and the ophthalmologist must be prepared to propose in plain text the diagnosis abuse, even though from a clinical view it may seem unlikely.’ (Additional file 3, 2006–2010:II:17). A multitude of interventions supportive of general knowledge on abuse and SBS/AHT were implemented: 1) formation of Children’s Houses (social service leading investigations of abuse in collaboration with healthcare, forensic medicine, police and judiciary); 2) formation of hospital-based Child Protection Teams (CPTs) led by paediatricians (2002–); 3) formation of a sub-association for Child Protection within the Swedish Paediatric Society (2009); 4) guidelines and manuals from Stockholm Healthcare Region and Karolinska Institutet (2008); 5) mental vaccination campaigns against shaking, supported by a national survey in 2011 showing nil shaking. Champions led the implementation activities 2) – 5), as well actively promoting supportive SBS/AHT knowledge by diffusion and dissemination. The years up until 2007 can be characterised by more diffusion and dissemination of knowledge about SBS/AHT, with few implementation activities. The period 2008 to 2013 can be characterised by more intensified implementation.

The period from 2014 can be described as the phase of achieving sustainability in the transfer of knowledge of SBS/AHT, with formations of more CPTs and guidelines issued by the National Board of Health and Welfare (2014) and the Swedish professional societies of paediatric (2018), paediatric radiology (2018) and ophthalmology (2019), However, disruptive knowledge started to emerge. These included acquittals by the Swedish Supreme Court (2014), in which expert witnesses stated that ‘the scientific support for the diagnosis of shaking (AHT, SBS) is unclear’, and the Swedish Supreme Administrative Court (2018), both concluding that certain medical findings indicating shaking violence were not enough for conviction without other findings or circumstances supporting abuse. Foremost in the dissemination of knowledge was the systematic literature review by SBU (2016) concluding: ‘There is insufficient scientific evidence on which to assess the diagnostic accuracy of the triad in identifying traumatic shaking (very low quality evidence).’ The Supreme Court decision in 2014 was questioned and the SBU report was much criticised by representatives of the Swedish Paediatric Society. Diffusion of disruptive knowledge on SBS/AHT was presented by other government agencies and scholars, in seminars, conferences, through new research findings derived from the Swedish setting, and by two paediatricians (2005 and 2017).

Table 3 shows the effect estimates of the incidence rate ratios in the subcategories maltreatment and assault. The change from version 9 to version 10 of the International Classification of Diseases 1997 did not have an effect on the incidence rates of either maltreatment or assault. The inclusion of outpatient specialised care in the registry in 2001 had substantial effects on the incidence rate ratios of the diagnoses of maltreatment and assault, 3.66 (95% CI 1.89–7.78) and 6.93 (95% CI 1.94–44.13). The effect estimate of the period of intensified intervention (2008–2013) was 1.63 (95% CI 1.34–1.98) for maltreatment, compared with the baseline intervention period (2002–2007), while no effect was observed for assault. During the phase of sustainability until obsolescence in 2014–2019 only maltreatment, not assault, decreased compared with in 2008–2013, 0.84 (95% CI 0.71–0.99), while the decrease was more pronounced in 2017–2019, 0.75 (95% CI 0.61–0.92).

Table 4 shows risk estimates of incidence rate ratios in the subcategories maltreatment and assault by university healthcare region during 1987–2014. Adjustments by region were statistically significantly different for maltreatment, with two- to five-fold increased incidence rate ratios compared with the reference period 1987–2001, except for the year 2005. Adjusted rate ratios by region for assault showed no differences.

Table 5 shows effect estimates of implementation of neuroimaging in investigations. Adjustment by neuroimaging showed no effect on the incidence rate ratios for maltreatment or assault for the years 2007–2014 compared with the years 2005–2006.

This study shows an increase in maltreatment diagnosis associated with transfer of supportive knowledge of SBS/AHT, while a decline started when disruptive transfer of knowledge began.

How can a two-fold increase of maltreatment diagnoses and then a decline, while assault diagnoses were stable, be interpreted in relation to other countries? Gilbert et al. (2011) performed a comparative analysis of infant diagnosis of maltreatment syndrome or assault (hospital admissions) and homicide and violent deaths in Sweden, England, Western Australia, New Zealand, Manitoba/Canada and the US. The ranges per 100,000 were 11.5 to 118 and 1 to 10.3, respectively [31]. Sweden was at the lower end of the range in both categories. The results of this study up to the year 2000 confirmed these results. However, when adding outpatient diagnoses, substantial increases of both maltreatment and assault diagnoses were noted. The steep increase in maltreatment diagnoses caused Sweden to have incidences comparable to those of England and New Zealand [31].

This increase is remarkable and several interpretations are possible in respect of false positives and false negatives: true increase, hidden cases or overdiagnosis. An actual increase is less likely, as no increase in assault diagnoses was noted and mortality for both maltreatment and assault declined compared with in previous years. A socioeconomic change like immigration, with the mother being foreign-born, had an impact during the years 2008–2014 compared with 1997–2007, but is unlikely to have had a major importance [28]. As mentioned, there was a strong correlation between SBS/AHT criteria and maltreatment diagnosis [28]. Thus, hidden cases or overdiagnosis are likely as alternative explanations. Can the implementation of supportive and disruptive knowledge of the SBS/AHT paradigm into healthcare services support these alternative explanations?

The hidden case hypothesis was claimed to achieve incidences of about 100 per 100,000 through application of the SBS/AHT guidelines. Provided that the SBS/AHT paradigm is an innovation based on solid scientific evidence [6], a large amount of hidden cases of abuse were detected in the Swedish setting, and the experience during 2002–2013 can be considered a success story of implementation of best practice for child protection [8]. The bridging of the knowledge gap for SBS/AHT was characterised by transfer of knowledge through diffusion and dissemination, with reference to US and UK scientific publications, while few research findings from the Swedish setting were presented. Several implementation drivers characterised the process of mainstreaming SBS/AHT knowledge. An organisational readiness [8] was evident, which might have been facilitated by previously implemented knowledge on battered-child syndrome. Organisational champions and leadership [8, 10, 11] and a positive managerial attitude toward changing guidelines regarding SBS/AHT in professional societies, healthcare and government agencies [8] can be identified as core factors. Other crucial factors include the system approach of intervention by creation of CPTs and a specific professional association for child protection [11] and capacity-building through teaching and training [9]. The uneven regional detection rates confirm the finding from New Zealand that formation of CPTs quadrupled the incidence of AHT [28]. The interdisciplinary approach of the intervention [11] might also have been a key factor; in this case facilitated by a standard protocol tool containing diagnostic statements from radiologists and ophthalmologists, irrespective of clinical findings. This can be one reason for high compliance, while protocols that rely on people interacting with other people, to change the behaviours of practitioners, generally have very low success rates [13]. The techno-medicine approach (‘only half of them have external signs of abuse’) might have been essential for the efficacy of implementing SBS/AHT knowledge deviating from the clinical understanding of battered-child syndrome.

The efficacy of the implementation of supportive knowledge of SBS/AHT, know-how knowledge, is further indicated by mental vaccination campaigns against shaking, based on hypotheses on the prevention of SBS/AHT. Population surveys in Sweden showed that shaking was reduced from 18% (2006) to almost nil (2011) [28]; however, the increase of maltreatment diagnosis was not impeded during this period. This calls a causal association between shaking and the diagnostic criteria for SBS/AHT into question. There is also a possibility that the mental vaccination campaign was ineffective for prevention of traumatic shaking, since healthcare-seeking of witnessed or admitted shaking may actually have increased during the campaign period [19].

Regarding evidence-based medicine, the overdiagnosis hypothesis is the most likely explanation for the increase of maltreatment diagnosis, supported by the decline observed when disruptive knowledge of the SBS/AHT started to be transferred. Thus, the sustainability of this transfer of knowledge, making the SBS/AHT innovation routine until it reaches obsolescence [8], seems to have been limited, based on the slow decline in the outcome measures during the last 5 years in the study period. The Swedish Supreme Court’s exoneration in 2014 of a father convicted of traumatic shaking might have contributed to this, and was followed by several acquittals in district courts and courts of appeal because of the Supreme Court’s precedence [32]. Similarly, the dissemination of a systematic literature review on the triad of SBS/AHT is likely to have had an impact. In respect of philosophy of science, this challenging of a scientific fact [25] has occurred through transition, with outsiders addressing mainstream questions through new ideas and disruptive findings in a field of science with a decline in publication activity [33]. However, representatives of professional societies have arguied against the Supreme Court’s exoneration and criticised the systematic literature review, even though it emanates from a government agency [34], which usually holds the highest level of credibility among Swedish professionals. Still, the prevailing mainstream thinking seems to be that the diagnostic accuracy of SBS/AHT is a scientific fact, as illustrated by a recent case presentation [35]. This can be interpreted to mean that the construction of SBS/AHT as a scientific fact will take time to change. SBS/AHT is solidly acknowledged worldwide, based on both active elements of knowledge (thought collective) and passive ones (observations) [25], even though new research findings gainsay previous knowledge supportive of SBS/AHT.

Already in 1978, it was discussed that the predictive screening of a protocol for latent markers of child abuse might be low in unselected populations; i.e., misclassification could occur when diagnostic categories, abuse and accidents/diseases, overlapped [36]. The possibility of false positives was again commented on in 1997 by Dr. Sunderland: ‘In the absence of definitive physical signs we may need to profess fallibility. Humility avoids hubris. False diagnoses not only harm children and their families but also devalue the profession.’ [15]. If the scientific foundation of the SBS/AHT paradigm is weak, not evidence-based, the forceful interventions against physical abuse, with a special focus on the SBS/AHT diagnosis, might have caused hundreds of false positive cases of alleged abuse in Sweden alone [37]. This would have had severe consequences for families seeking healthcare for their infants [38]. Hence, in a public health perspective, the findings of this study can be claimed to indicate a situation far from the recommendations of evidence-based child protection from WHO and ISPCAN, to keep false positives at a minimum [1].

Medical practices where best evidence shows no efficacy or harm are common [39]. The half-life of dogma related to surgical practice is estimated at 45 years [40], corresponding to the length of time since the introduction of SBS, although its full implementation came rather late in the Swedish setting. De-implementing obsolete and inappropriate health interventions and procedures is essential for minimising patient harm [39, 41]. The Swedish setting for infant health has experienced two successful de-implementations of harmful practices, strongly supported by professional societies and rapidly executed, first to prevent sudden infant death syndrome [42] and second to prevent coeliac disease [43]. Provided that there were fewer false positive maltreatment diagnoses during the latter study period, the comparatively rather slow decline might be due to obsolescence of clinical guidelines not reconsidered based on disruptive knowledge of SBS/AHT paradigm. Barriers to de-implementation have been recognised among health professionals, leadership and structure, especially if an inappropriate intervention has been delivered for a long period and fear of liability may have been important [41].

From the perspective of implementation science, the findings from this study indicate that the most efficient drivers to change a health practice are the professional societies providing recommendations and guidelines. On the other hand, clinical norms and preferences purportedly in the best interests of the patient may overrule scientific values and new research findings [24]. Hence, lack of readiness of uptake of new research findings to change know-how knowledge among health professionals may be a barrier to evidence-based practice [41].

An increase in maltreatment diagnosis was associated with transfer of supportive knowledge about SBS/AHT by diffusion, dissemination and implementation. During the sustainability phase, disruptive knowledge was presented. A decline in diagnosis occurred toward the end of the study period, which might indicate a burgeoning de-implementation process.