Questionnaire‐Based Experimental Study of the Diagnostic Process in Suspected Shaken Baby Syndrome

ABSTRACT

Objective

The objective of this study was to examine the correctness of the statement that the diagnosis of shaken baby syndrome (SBS) is a “medical conclusion.”

Design

If the outcome of an event is considered bad, people tend to attribute intent or responsibility. Against this backdrop, a randomized, blinded trial was applied using two case‐based questionnaires of possible violent shaking of an infant, identical except for the outcome: fatal or nonfatal.

Setting

We used a report of a 2‐month‐old baby who had suddenly stopped breathing and was subsequently shaken by the father with the intention to resuscitate him. After admission, encephalopathy, subdural hemorrhages, and retinal hemorrhages were detected.

Participants

A total of 1269 randomly selected physicians received the questionnaire, whose distribution was randomized and blinded. The participants were not informed about the existence of two versions of the questionnaire.

Results

The participants who responded to the fatal version considered in a significantly higher proportion (79% [95% CI: 73−85]) that shaking caused the triad findings than those who responded to the nonfatal version (68% [95% CI: 61−75]) (p = 0.01). When pediatricians and ophthalmologists are merged, the corresponding proportions were (91% [95% CI: 86−96]) versus (74% [95% CI: 66−82]) (p = 0.001). Radiologists and forensic pathologists did not distinguish significantly between fatal and nonfatal outcomes.

Conclusion

The study indicates that among pediatricians and ophthalmologists, the diagnostic process in suspected SBS is more value‐based than evidence‐based. As these two specialties dominate the SBS diagnostic procedure, the SBS diagnosis is, in this sense, not strictly a “medical conclusion.”

1. Introduction

Studies of scientific controversies may lead to a better understanding of how the scientific community is functioning. Historical debates are often easier to study because those involved—both proponents and opponents—are no longer alive or active. The scientific arguments of ongoing controversies are sometimes more difficult to study. Even if one theory is assessed as more plausible than another [1], scientific arguments are not always sufficient to convince its opponents [2].

The scientific controversy investigated here is about the accuracy of the diagnostic procedure regarding shaken baby syndrome (SBS). When contemporary acceptable causes have been excluded, this syndrome defines shaking as the cause of symptoms of encephalopathy, subdural hemorrhage (SDH), and retinal hemorrhages (RH)—the triad—if signs of impact are lacking. Encephalopathy has been assumed to result from disrupted cerebral nerve fibers, SDH from disrupted bridging veins, and RH from traction forces between the vitreous body and the retina [3]. Importantly, if the triad is identified in an infant and acceptable causes are ruled out, it is usually concluded that the infant must have been shaken [4]. The same conclusion is reached even if the caregiver denies that shaking took place or states that shaking was applied after the baby's sudden collapse as a resuscitation attempt [4, 5].

This pathogenesis (according to the traditional SBS theory) was the uncontested mechanistic theory until the early 2000s, when another theory regarding the pathogenesis of triad findings was presented [6, 7, 8, 9]. According to this non‐shaking theory, triad findings are suggested to be caused by hypoxia due to a respiratory arrest, which in turn could be caused by, for example, aspiration or immaturity of the infant's brain [8, 10]. Moreover, the diagnostic accuracy of triad findings to predict that shaking caused these findings was later found to be insufficient (i.e., very low scientific evidence) because almost all studies were based on circular reasoning and hence hampered by a high risk of bias [11].

These factors have generated a heated scientific controversy, in part due to the vast differences in social, societal, and legal consequences of the two main competing theories. Following the traditional SBS theory, the consequences are that a caregiver will be held responsible and that the infant will be removed from his or her family, but following a non‐shaking theory, there will probably be no such consequences.

Studies have indicated that the diagnostic approach to SBS (which, after 2009, was conflated into the broader and heterogeneous group of abusive head trauma [AHT] [12, 13]), is unreliable. For example, studies of “do not shake a crying baby” preventive programs conducted before 2009 indicated a significant decrease of AHT, whereas similar preventive studies conducted after 2009 did not [14]. Moreover, in one study, the effect of a “do not shake a crying baby” program included three sham programs not associated with SBS or AHT [15]. The “do not shake a crying baby” program did not result in a decreased AHT incidence, whereas all three sham programs resulted in a highly significant decrease (p < 0.001) [15, 16]. These unexpected results further suggest that the diagnostic procedures used to substantiate SBS are unreliable.

Despite the ongoing scientific controversy, several pediatric, radiology, and neuroradiology societies from the USA, Japan, and Europe supported a consensus document in which the traditional SBS theory with its legal consequences was endorsed. This consensus document claimed, among other things, that the diagnostic procedure of suspected SBS/AHT is a “medical conclusion,” leaving no room for value judgments in the process. It was also stated that “the diagnosis of AHT is not one of intent or murder” [17]. The authors of this consensus document not only stated that the traditional SBS theory and the associated diagnostic accuracy was evidence‐based but also claimed that a systematic review of the diagnostic accuracy of SBS [11] had high risk of bias and hence should be ignored. On this background, they argued that the diagnosis of SBS was based on “genuine” evidence and accordingly nothing but a “medical conclusion” [17].

The point of departure of the present study was that if the SBS diagnosis is a strictly “medical conclusion, not a legal determination of intent,” this means that the diagnostic procedure is not influenced by the outcome, fatal or not. Following the traditional SBS theory, however, the legal consequences will be very different depending on whether the infant dies or recovers.

To assess the diagnostic procedure in suspected SBS, we conducted an experimental study to test the hypothesis that concerned physicians' “medical conclusion” depends on the outcome, the so‐called Knobe effect [18, 19, 20]. The Knobe effect reflects people's inclination to ascribe intentions if the consequence (outcome) of an action is considered bad, including unintentional side effects. If, for example, a physician provides drugs to relieve a patient's pain at the end of life, and even though the foreseen side effect (consequence) is that the patient's life might be shortened, people (even colleagues) are inclined to ascribe an intention to shorten the patient's life.

The research question was: Is the SBS diagnosis influenced by the outcome—fatal or nonfatal—reflecting a stronger inclination to ascribe intention and responsibility when the consequence is fatal [20]? Does this indicate that the SBS diagnosis is value‐based rather than evidence‐based?

2. Ethics Approval and Consent to Participate

Because this was a questionnaire‐based study, information was provided in a cover letter clarifying that participation was voluntary (see Supporting Information text). In order not to jeopardize the validity of the study, the participants (all physicians licensed to practice medicine, working in one of four engaged medical specialties) were not informed that the study design involved two versions of the questionnaire (fatal or nonfatal outcome). Those who responded were assumed to consent to participate in the study. Further, those who were interested in the results of the study were invited to request these from the authors. Participants were also informed that the study would not be subject to the Ethics Research Act because it involved no intervention on a research person and no processing of sensitive personal data as specified in the Swedish Ethics Research Act. Hence, the Swedish Ethics Authority provided only an advisory statement, declaring that there were no ethical obstacles to conducting the project (Nr: 2022‐05281‐01), and accordingly waived full informed consent.

3. Methods

We used two questionnaires where the only difference was the outcome: fatal or nonfatal. The distribution of the questionnaire was conducted as randomized and blinded. Blinded in this context means that the participants were not informed that the design included two versions of the questionnaire.

Instead of sending the questionnaires to individual concerned physicians, we also considered approaching concerned multidisciplinary teams with the questionnaire. We found this was not an adequate option because there are few such teams in Sweden, and because the goal of the study was to avoid the well‐known risks of groupthink and group obligations within the teams [5]. The composition of multidisciplinary teams might differ from country to country, and in, for example, the US, they might include pediatric neurologists, neurosurgeons, intensivists, and so forth. We included solely pediatricians, ophthalmologists, radiologists, and forensic pathologists, as these are the specialties involved in the diagnostic procedure in suspected SBS—forensic pathologists, however, on request of legal authorities, separate from healthcare. In Sweden, child abuse pediatrics does not exist as a formal subspeciality, and hence it was necessary to include a rather large proportion of participants from the concerned specialties.

The design of the study was inspired by Nichols and Knobe [18] and based on people's inclination to ascribe intention when the consequence of an action is considered bad, even though the original intention was benevolent—that is, to resuscitate the infant [18].

The two versions of the questionnaire were equally distributed to randomly selected participants, received from a commercial company (IQVIA, Oslo, Norway, https://www.iqvia.com/) and including 400 pediatricians, 400 ophthalmologists, and 400 radiologists. All 69 forensic pathologists came from the National Board of Forensic Medicine. The participants randomly received either the questionnaire with a fatal outcome or the nonfatal questionnaire version. Of the 1269 selected participants, 1244 were accessible (i.e., received the mailed questionnaire), and a total of 391 responded to the questionnaire. One respondent, however, did not specify a specialty.

The two versions of the questionnaire were randomly distributed and tested in a pilot study among a group of medical students (n = 61) at Karolinska Institutet, Stockholm, before the present study was conducted. This pilot study was also used when estimating the number of participants in the present study. Moreover, because the incidence of suspected SBS cases is very low in Sweden [21, 22], a large study population was chosen to include as many pediatricians, ophthalmologists, and radiologists as possible with at least some experience of handling suspected child abuse. Accordingly, a rather low response rate was expected among those who had no experience managing SBS cases or had not been trained in this area. Apart from including all forensic pathologists in the country (N = 69), we aimed to include at least 100 respondents from each of the randomized three larger specialties. Of 1249 accessible participants from all four specialties, 391 (31.3%) responded.

Response rates and ratios of respondents to the two versions of the questionnaire are summarized in Table 1.

Table 1.

Background variables. Number of respondents in each specialty, response rate, ratio of those who had responded to the nonfatal/fatal version of the vignette, median age, and sex of the respondents.

Number of respondents/accessible participants	Response rate (95% CI)	Ratio of responses to nonfatal/fatal version (%)	Median age (years) with interquartile range)	Ratio female/male respondents (%)
All respondents 391/1249	31% (26−36)	49/51	48 (40−58)	51/49
Pediatrics 129/391	33% (28−48)	47/54	51 (42−59)	60/40
Ophthalmology 119/396	30% (25−35)	59/41	47 (38−57)	48/52
Radiology 114/393	29% (25−33)	47/53	48 (40−59)	43/57
Forensic pathology 28/64	44% (32−56)	54/46	44 (35−48)	50/50

The response options to each statement were “totally agree,” “agree to a large extent,” ”disagree to a large extent,” and “totally disagree.” Accordingly, we did not apply a Likert scale usually associated with ordinal data, and in this manner the results were transformed into nominal data and presented as proportions of those who agreed (totally and to a large extent) with a 95% CI. The reason we took this approach is that it is more informative to compare proportions of those who agreed (totally and to a large extent) with the actual statement with a 95% CI. This approach is more informative than solely presenting p values, because when comparing proportions, it is possible to identify the size of the difference—for example, 30%—and not just a p < 0.05. In addition, we also applied a chi‐square hypothesis test. As the significance level, we chose p < 0.05. The registration and statistical calculation of p values was conducted by the Epi Info software (version 6.04d, January 2001, Centers for Disease Control and Prevention [23]). When calculating the 95% CI, we used the following formula: P ± $1.96\sqrt{\frac{P(1-P)}{n}}$, where P is the proportion of those who agreed, (1−P) is the proportion of those who disagreed, and n is the total number of the actual subgroup.

The questionnaires were distributed in mid‐January 2023 by surface mail, and two reminders were distributed 2 and 4 weeks later, respectively. The last response was included in late June 2023. When the responses were registered, the envelope with the code number was discarded, and it was no longer possible to identify the respondents. In this manner, total anonymity was guaranteed.

To inform participants who were familiar only with traditional SBS theory and its social and legal consequences, we added a short introduction letter in which we presented the scientific controversy between the traditional SBS theory and the main competing non‐shaking theory (see Supporting Information text). The same introduction text was sent to all participants independent of whether they responded to the fatal or the nonfatal version of the questionnaire.

The questionnaire was based on a vignette of a typical 2‐month‐old boy with a triad of findings (but without signs of relevant trauma) where the father stated that he had shaken the baby, but only after its collapse from a respiratory arrest—that is, the shaking was a resuscitation attempt (see Supporting Information text).

One question was whether the participants agreed that the triad findings were caused by (a) shaking or by (b) hypoxia.

3.1. Outcome Measures

The main outcome measure was the possible difference between the participants' preferences of valid theory when randomly and blindly exposed to the fatal and the nonfatal outcome of the infant.

The CONSORT 2017 checklist was followed when applicable [24].

4. Results

Pediatricians formed the largest group of respondents (n = 129), followed by ophthalmologists (n = 119), radiologists (n = 114), and forensic pathologists (n = 28) (Table 1). The sex distribution among the respondents in the four specialties was similar (Table 1).

There was no statistically significant difference in the response pattern between those who had “some” experience (70%) and those who had “no” personal experience (30%) of diagnosing suspected SBS cases regarding whether they agreed or disagreed that shaking or hypoxia caused the triad findings. As can be deduced from the interquartile range age, the study groups included not only board‐certified specialists but also some residents and retired but still actively working specialists. The sex ratio (female/male) within each random sample of each specialty did not differ statistically from the whole group of physicians in the respective specialty.

4.1. Fatal Versus Nonfatal Outcome

Generally, a significantly higher proportion of participants who responded to the fatal version of the questionnaire considered that shaking caused the infant's triad findings (79% [95% CI: 73−85]) compared with those who responded to the nonfatal version (68% [95% CI: 61−75]) (p = 0.01) (Table 2A).

Table 2A.

Main outcome. Proportions of respondents who agreed that the triad findings were caused by shaking, by nonfatal and fatal outcomes, by medical specialty, and by age.

Specialty n = number in each subgroup	Agreed that shaking caused the triad findings (95% CI)		χ 2 test
	Nonfatal outcome	Fatal outcome	p value
All participants (n = 192 vs. 185)	68% (61−75)	79% (73−85)	0.01
Pediatricians (n = 58 vs. 68)	69% (57−81)	91% (84−98)	0.002
Ophthalmologists (n = 67 vs. 48)	79% (69−89)	90% (82−98)	0.1
Pediatricians and ophthalmologists (n = 125 vs. 116)	74% (66−82)	91% (86−96)	0.001
Radiologists (n = 52 vs. 56)	62% (49−75)	68% (56−80)	0.5
Forensic pathologists (n = 14 vs. 13)	29% (5−53)	23% (0−46)	0.7
Radiologists and forensic pathologists (n = 66 vs. 69)	55% (43−67)	59% (47−71)	0.6

Note: Proportions are presented with a 95% CI. p values were calculated with χ ² test. CI not overlapping each other indicates that if a hypothesis test was conducted, it would have been < 0.05. The meaning of “versus” indicates the difference between the number of nonfatal responses and the number of fatal responses.

The whole group of pediatricians displayed a significant difference in this respect (69% [95% CI: 57−81] vs. 91% [95% CI: 84−98]) (p = 0.002). There was no statistically significant difference between the fatal and nonfatal outcome regarding the whole group of ophthalmologists, but when merged with the pediatricians a significantly higher proportion of the combined groups agreed that shaking caused the triad findings when responding to the fatal version (91% [95% CI: 86−96] vs. 74% [95% CI: 66−82] in the nonfatal version [p = 0.001]).

There was no difference among radiologists or forensic pathologists regarding fatal versus nonfatal outcome, and when merged, 59% (95% CI: 47−71) agreed that shaking caused the triad findings in the fatal version and 55% (95% CI: 43−67) (p = 0.6) in the nonfatal version.

Although there was no statistically significant difference regarding the fatal or nonfatal versions among those who agreed that hypoxia caused the infant's triad findings, fewer agreed that hypoxia caused the triad findings when the outcome was fatal. The proportional difference was also low in a comparison of the fatal and the nonfatal versions (Table 2B).

Table 2B.

Proportions of respondents who agreed that the triad findings were caused by hypoxia, by nonfatal and fatal outcomes and by medical specialty. Proportions are presented with a 95% CI.

Specialty	Agreed that hypoxia caused the symptoms and signs (95% CI)		Fisher's exact test
	Nonfatal outcome	Fatal outcome	p value
All participants (n = 187 vs. 182)	23% (16–28)	17% (11–23)	0.2
Pediatricians (n = 57 vs. 66)	23% (13–33)	11% (3–19)	0.07
Ophthalmologists (n = 65 vs. 48)	22% (12–32)	15% (5–25)	0.4
Pediatricians and ophthal‐mologists (n = 122 vs. 114)	22% (15–29)	12% (6–18)	0.05
Radiologists (n = 55 vs. 50)	20% (9–31)	24% (12–36)	0.7
Forensic pathologists (n = 13 vs. 14)	36% (10–62)	23% (1–45)	0.5
Radiologists and forensic pathologists (n = 64 vs. 68)	23% (13–33)	24% (14–34)	1.0

Note: p values were calculated with χ ² test. The meaning of “versus” indicates the difference between the number of nonfatal responses and the number of fatal responses.

5. Discussion

The results of the present study suggest that pediatricians and ophthalmologists are more inclined to ascribe intention when the outcome is fatal, even though the intention was described as benevolent—that is, as a revival attempt. The reason there was a significant difference between the fatal and nonfatal outcome for the combined pediatrician–ophthalmologist group, but not among radiologists and forensic pathologists, might be because both pediatricians and ophthalmologists are those who meet with and talk to the parents. Radiologists have no direct contact with the parents, and forensic pathologists base their opinion mainly on hospital records and police investigations (and, in fatal cases, on autopsy findings). Pediatricians also have the first contact with the parents and the baby and observe their interaction; they also initiate brain imaging and, if an SDH is identified, an ophthalmologic examination. As RH often have been considered specific to shaking, ophthalmologists often participate in the team's determination of the SBS diagnosis.

The results further suggest that medical experts within these specialties are apparently not immune to the Knobe effect, that is, to ascribe malicious intent even when the intent of the action is described as a benevolent revival attempt. We chose to test the hypothesis that the SBS diagnosis is value‐based rather than evidence‐based, and the results indicate that our hypothesis has obtained support among pediatricians and ophthalmologists, who usually dominate the diagnostic determination of SBS. In this sense, it seems reasonable to conclude that the SBS diagnosis is not a purely “medical conclusion.”

Apart from the forensic pathologists, who were all (N = 69) included, the samples were randomly selected, but the distribution of the two versions of the questionnaire was randomized. Although selection bias cannot be fully excluded, we found no difference regarding sex between responders and nonresponders. Moreover, the absence of significant differences regarding sex ratios between responders and the whole samples of each of the four specialties indicates that the responders were representative of their specialty.

5.1. Fatal Versus Nonfatal Outcome

Compared with those who responded to the nonfatal version of the questionnaire, significantly more of those who responded to the fatal version stated that the triad findings were caused by violent shaking. This discrepancy was most obvious among the combination of pediatricians and ophthalmologists. Contrary to what would be expected if the diagnostic procedure were an uncontroversial and strictly “medical conclusion,” these results indicate that the outcome was influenced by the Knobe effect [18, 19, 20]. Generally, as expert witnesses in the courtroom, physicians are supposed to consider and present evidence‐based, not eminence‐based and value‐based, assessments and diagnoses. The large proportional differences demonstrated in the present study between the fatal and the nonfatal versions, especially among the merged group of pediatricians and ophthalmologists, indicate that the SBS diagnostic procedure is unreliable, particularly within these medical specialties.

A 2015 study from 10 leading US pediatric hospitals demonstrated that a large majority of respondents felt that shaking with or without impact would be likely or highly likely to result in the triad findings of SDH, severe RH, and coma or death [25]. About 83% felt that shaking (without impact) was likely or highly likely to produce SDH, ~94% felt that shaking (without impact) was likely or highly likely to produce RH, and ~80% felt that shaking (without impact) was likely or highly likely to produce coma or death [25].

Among radiologists, 88% considered that the SBS diagnosis was valid, as did 98% of the ophthalmologists and 93% of the child abuse pediatricians, but only 41% of the pathologists did so [25]. Very few considered that hypoxia caused the triad findings. There seem to be big differences between some of the specialties in the US versus Sweden (and/or in 2015 vs. 2023), but not with respect to child abuse pediatricians, ophthalmologists, and forensic pathologists [25].

The indication that the responses of pediatricians and ophthalmologists seem to be more value‐based than evidence‐based might be associated with the traditional SBS theory in which the ethical principle “First of all, protect the child!” is dominating, although disguised [26]. In the ethical dilemma between evidence‐based diagnostic procedures and value‐based principles, the present results suggest that most pediatricians and ophthalmologists found that the least bad choice was to prioritize the ethical principle [26]. The results also suggest that when the outcome is considered bad, the diagnostic procedure of SBS is not just a “medical conclusion.”

5.2. Strengths and Limitations

The present study is the first within the SBS area with the applied experimental design—a randomized and blinded trial—and the results will hopefully lead to a more constructive discussion of the diagnostic procedure in suspected SBS cases. It will, however, be somewhat difficult to replicate this study because participants might have knowledge about the design and hence make future participants unblinded in a similar study. If the participants are aware of the design, that is, are unblinded, this would jeopardize the validity of such a study.

The response rate was rather low, but this was expected as a smaller share of the included specialties handle cases of suspected SBS/AHT. Some participants (nonresponders) returned the questionnaire to the sender, stating that they had no experience or knowledge about the issue. Particularly, radiologists commented that it was not their job to make the SBS diagnosis, but only to “raise a red flag” if an SDH is identified. Some radiologists also stated that they had been taught during their specialist training that SDH and RH indicated violent shaking, SBS.

A majority of the participants (70%) stated that they had experience managing suspected SBS cases, and the remaining 30% stated that they had no such experience. As there was no statistical difference in response pattern between the two groups, and because guidelines, education, and training in how to manage suspected SBS cases have been provided at least since 2008 in Sweden, we also included the latter group in our study sample. The influence of educational programs and guidelines deserves to be further scrutinized, as indicated in the present study and by a previous study [27].

6. Conclusions

Proponents of the traditional SBS theory declared in 2018 that the AHT diagnosis, which since 2009 has included SBS, is a “medical conclusion” and “not a legal determination of the intent of the perpetrator.” The results of the present study indicate that the concerned pediatricians and ophthalmologists were influenced by the Knobe effect, which in turn indicates that the SBS diagnosis is more value‐based than evidence‐based. Accordingly, it is not reasonable to claim that the SBS diagnosis is a strictly “medical conclusion.”

Author Contributions

N. Lynøe: conceptualization, methodology, data curation, software, formal analysis, writing – original draft. A. Castor: conceptualization, methodology, validation, writing – review and editing, software. N. Juth: conceptualization, methodology, writing – review and editing, supervision. A. Eriksson: writing – review and editing, supervision, validation. N.L. and A.E. participated in the expert panel when conducting the SBU report during 2014−2018.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Final supporting file 10 July 2025.docx.