Cross-Cultural Comparison of the Contexts Associated with Emotional Outbursts

Towards the goal of identifying the different patterns of contexts associated with outbursts that may be aetiologically relevant, an exploratory study developed the Emotional Outburst Questionnaire and collected responses from caregivers of children and young people (Chung et al., 2022). The development of this new measure was argued as necessary for advancing our understanding of emotional outburst aetiology, as other existing measures collect limited information regarding outbursts (e.g., frequency of outbursts in the Developmental Behaviour Checklist; Einfeld & Tonge, 1995), or were developed for typically developing individuals (e.g., the Multidimensional Assessment of Preschool Disruptive Behavior; Wakschlag et al., 2012, 2014). Using the Emotional Outburst Questionnaire, it was found that the majority of antecedents and setting events of emotional outbursts could be organised into six latent factors (Chung et al., 2022). These factors were labelled as (1) Sensory, which included sensory stimuli (e.g., sudden or loud noises); (2) Cognitive Demand, which contained antecedents that may be cognitively demanding for an individual (e.g., change in expectation); (3) Threat to Self, which consisted of antecedents that may threaten the self-esteem of an individual (e.g., being criticised); (4) Cross-settings, which contained a range of setting events (e.g., public setting, or unfamiliar person); (5) Safety, which included setting events perceived as safe (e.g., private setting); (6) States, which consisted of setting events related to the physiological state of the individual (e.g., tired).

Based on these contextual factors, responses were transdiagnostically classified into three unique clusters, such that people who experienced outbursts in a similar pattern of contexts were grouped together. These three clusters were proposed to each represent an aetiological pathway to emotional outbursts: (1) children and young people with differences in sensory processing may be more susceptible to experience outbursts across antecedents and setting events due to the additional demand of background stimuli (the Sensory Sensitivity pathway); (2) children and young people who mask their distress in settings that lack perceived safety may experience more outbursts upon returning to safe setting events due to the build-up of distress and the cost of masking (the Perceived Safety pathway); (3) children and young people with differences in safety perception may consider more environments to lack perceived safety, thereby increasing the likelihood for outbursts to be experienced in these unsafe setting events (the Perceived Unsafety Pathway; Chung et al., 2022). As this framework of contextual pathways was derived from exploratory and inductive analysis, the validity and replicability of this proposal remain untested. Therefore, the potential generalisability of the previous interpretations requires evaluation through confirmatory analysis with additional data.

The need for additional sampling presents a further opportunity to explore the potential influence of culture on the aetiology of emotional outbursts, as the pathways proposed in the previous study were largely constrained to the context of English-speaking caregivers from the United Kingdom. Cross-cultural studies have commonly investigated the differences in emotional processes between societies that promote the independence of the self from others (e.g., the UK) versus societies that prioritise the interdependence of the self with others (e.g., Brazil; Markus & Kitayama, 1991, 2010). These studies have revealed differing cultural norms in the regulation and expression of emotions (for reviews, see Ford & Mauss, 2015; Friedlmeier et al., 2014), which may lead to fundamental differences in the manifestation and experience of emotional outbursts in individuals across cultures. For example, it has been demonstrated that people from relatively interdependent cultures were more likely to suppress emotions compared to people from relatively independent cultures, a tendency which was motivated by the desire to maintain harmony across interpersonal relationships (Matsumoto et al., 2008; Ramzan & Amjad, 2017). Indeed, this degree of interdependence is apparent in countries such as Brazil, as Brazilian families tend to avoid conflict and place particular emphasis on pleasing family, friends, and the wider community (Dessen & Torres, 2019). Thus, it is possible that individuals from relatively interdependent cultures such as Brazil may be more likely to suppress their distress in response to antecedents, especially within contexts that involve other people (e.g., in public), thereby increasing the likelihood of emotional outbursts manifesting via the Perceived Safety pathway. Additionally, families of individuals with ASD in Brazil have reported a lack of support to socially integrate individuals with ASD into the wider community; and inadequate measures to minimise the exposure to intimidating environments for individuals with ASD (Gomes et al., 2015; Weissheimer et al., 2021). Consequently, these two factors may contribute to the lack of perceived safety in contexts involving the wider community (Dammeyer & Chapman, 2018), further contributing to the experience of distress in these environments.

In addition to cultural differences in how individuals respond to antecedents, there may be further cultural differences in the caregivers’ response that can further modulate the likelihood of emotional outbursts in different contexts. For example, cultural differences in the beliefs around the acceptability of expressing negative emotions appeared to influence how caregivers from Brazil and other countries preferred to respond in situations that elicited such emotions in their children (Corapci et al., 2018; Mograbi et al., 2018; Rimes & Chalder, 2010). Cultural differences in the experience of stigma for caregivers of individuals with ASD and other neurodevelopmental disorders may further influence caregiver response to negative emotion expression (Kinnear et al., 2016). For example, a pioneering study covering a large sample of caregivers of people with ASD across Latin America reported that Brazilian caregivers experienced significantly more stigma compared to caregivers from other Latin American countries, which could be related to the greater frustration Brazilian caregivers experienced when seeking support and the lower verbal ability of people with ASD in the Brazilian sample (Paula et al., 2020). Therefore, caregiver response to emotional outbursts may be differentiated by cultural variations in beliefs around negative emotion expression and the experience of stigma, which may include context-specific differences in the caregiver response (e.g., at home versus in public) that can conceivably promote or suppress the experience of emotional outbursts by individuals in different contexts.

As such, the consideration of culture is critical for the proposed framework, as culture itself could be deemed as an overarching level of context that may directly influence how setting events and antecedents may be organised into latent factors, and how these factors may in turn relate to the aetiology of emotional outbursts (Bronfenbrenner & Morris, 2006). There may be fundamental differences in the pathways that emerge from the responses of a culturally distinct sample, in terms of the distinguishing contexts and mechanisms associated with the identified pathways, and the demographic composition of each group. If a pathway were indeed found to be invariant across cultures, this would demonstrate the importance of recognising the significance of the associated contexts when considering the mechanisms underlying current and potential interventions for emotional outbursts. From a measurement perspective, it would be critical to assess the cross-cultural invariance of the measured constructs related to emotional outbursts, as this would inform the potential generalisability of the derived framework. Furthermore, recent work suggested that measures of emotion regulation strategies were largely invariant across samples of adults from the US and India (Van Doren et al., 2021). If a similar degree of invariance were found across cultures for measures of emotional outbursts, this could facilitate future endeavours in expanding the present framework by integrating patterns of emotional regulation strategies in this cross-cultural aetiological account of emotional outbursts.

The present study aimed to investigate whether the contextual clusters of emotional outbursts that emerge from a culturally distinct sample of caregivers in Brazil would be comparable to the previously identified clusters. To facilitate this cross-cultural comparison, the Emotional Outburst Questionnaire was translated and culturally adapted into a Brazilian Portuguese version, and the factor structure of the contextual items in the Brazilian Portuguese version of the questionnaire was validated and compared against the factors derived from the English version.

Caregivers of young people (aged 6–25 years) who experienced emotional outbursts at least once a month were recruited via local autism units for families of children and young people with neurodevelopmental conditions in Brazil. In addition to these units, the questionnaire was also available on social networks of the authors MCTVT and RL for recruitment of eligible caregivers. The age and outburst frequency inclusion criteria matched those of the previous English study (Chung et al., 2022). Out of a total of 359 completed responses, one participant was excluded as the young person was outside the age inclusion criterion. To ensure that the final sample met the criterion for outburst frequency, 27 participants were excluded based on their response to the question regarding general outburst frequency (item 57) on the Emotional Outburst Questionnaire: 20 indicated Less than once a month; six selected Never; one response was missing. Four participants were excluded as their responses were partially missing for items regarding the antecedents and setting events of emotional outbursts (items 58–112).

Caregivers of 107 young people indicated that medication was taken for emotional outbursts (32.7%). Fifty-four families have accessed support for outbursts in terms of a programme, intervention, or training (16.5%), 48 of whom found the support to be effective (88.9%). Of the families who have accessed support, 10 received this support immediately or soon after difficulties with outbursts began (18.5%), whilst the 44 remaining families received support a while after (81.5%). With regards to schooling or employment status, 262 young people were in mainstream schools (80.1%); 51 attended special schools (15.6%); one was in further education (0.3%); one was in higher education (0.3%); four were in employment preparation (1.2%); seven were unemployed (2.1%; one missing). Overall, 194 young people received a statement of special educational needs or have an educational plan at school (59.3%). Twenty-five caregivers reported that the young people they were caring for have experienced early traumatic or adverse events (7.6%; four selected Prefer not to say; 11 missing), which were described to encompass: natural disasters; death or serious injury of someone close to the person; poverty; witnessing abuse or violence; emotional, physical, or sexual abuse; neglect. Supplementary Table 1 compares the demographic information between caregivers from the English (Chung et al., 2022) and Brazilian samples. Differences were present across demographic variables. Compared to the English sample, the Brazilian sample had: a lower mean age; a higher proportion of boys; a higher proportion of diagnoses of ASD and Down’s syndrome; a lower proportion of diagnosis of intellectual disability; a higher proportion of pharmacological intervention use for outbursts; a lower proportion of non-pharmacological intervention use for outbursts; differences in schooling and employment status; and a lower proportion of exposure to trauma.

The Emotional Outburst Questionnaire is an informant-report measure, which contains 133 items pertaining to the characteristics of outbursts, including the antecedents and setting events associated with outbursts, in addition to frequency, duration, intensity, and behaviours observed in the most and least severe outbursts (Chung et al., 2022). The 55 items related to the antecedents and setting events of outbursts were rated on a three-point scale: Not applicable/never/rarely (0–3 times out of 10), Sometimes (4–6 times out of 10), Often/always (7–10 times out of 10). In the English study, most contextual items loaded onto six latent factors, with internal consistency as indicated by Cronbach’s α ranging from 0.68 to 0.84 (Chung et al., 2022).

The translation and cross-cultural adaptation of the Brazilian Portuguese Emotional Outburst Questionnaire was carried out in a previous study (Balbueno, 2021), following the stages recommended by the International Test Commission (2017). The process included: (1) independent translations by two native speakers of Brazilian Portuguese with proficiency in English; (2) synthesis of the two translations and analysis of the synthesis by three experts (two native speakers of Brazilian Portuguese and one native speaker of English) that compared the synthesis with the original versions of the translations using recommendations of the International Test Commission (2017); (3) analysis by the participants of the target audience based on the understanding of the items, clarity and accuracy; (4) back-translation by a native English speaker and final modifications to rectify discrepancies between the original and back-translated versions of the measure (Balbueno, 2021).

Caregivers provided informed consent prior to participating in this study. The majority of participants completed the Brazilian Portuguese version of the Emotional Outburst Questionnaire online on the Research Electronic Data Capture (REDCap) platform (Harris et al., 2009, 2019). Twenty responses were collected in person from caregivers attending a local mental health care centre. Questionnaire and demographic data for the 268 responses to the English version of the questionnaire were accessed via the publicly available repository: https://​osf.​io/​2j47e/​.

The six-factor solution from the previous analyses (Supplementary Fig. 2) was fitted onto the Brazilian responses, which resulted in inadequate model fit based on the chi-square statistic, RMSEA, and SRMR, but acceptable fit based on the CFI [χ² (512) = 2444.080, p < 0.001; CFI = 0.930; RMSEA = 0.108, 90% CI (0.103, 0.112); SRMR = 0.119]. The modification indices of this base model identified that the addition of the item ‘Familiar person’ loading onto the Safety factor would improve model fit by the greatest amount. With the addition of this loading, the model demonstrated similar levels of fit as the base model [χ² (5110 = 2028.165, p < 0.001; CFI = 0.945; RMSEA = 0.095, 90% CI (0.091, 0.100); SRMR = 0.110]. An examination of the modification indices for this model indicated that ‘Familiar setting’ additionally loading onto the Safety factor would further improve model fit. Indeed, addition of this loading improved model fit as indicated by the CFI, RMSEA, and SRMR to levels ranging from acceptable to good [χ² (510) = 1523.689, p < 0.001; CFI = 0.963; RMSEA = 0.078, 90% CI (0.074, 0.083); SRMR = 0.100]. Therefore, the final model differed from the base model in terms of the additional loading of two items (‘Familiar person’ and ‘Familiar setting’) onto the Safety factor (Supplementary Fig. 3). The internal consistency of each factor, as measured by Cronbach’s α was found to be similar to the previous analyses (Sensory = 0.76; Cognitive Demand = 0.81; Threat to Self = 0.82; Cross-settings = 0.80; Safety = 0.85; States = 0.84).

When the responses from the English (N = 268) and Brazilian Portuguese (N = 327) versions of the questionnaire were combined and fitted onto the final model, the CFI, RMSEA, and SRMR indicated acceptable to good fit [χ² (1020) = 2636.512, p < 0.001; CFI = 0.959; RMSEA = 0.073, (0.070, 0.077); SRMR = 0.098], demonstrating configural invariance across the two versions of the questionnaire. As each contextual item comprised of three response options, the threshold invariance model demonstrated no change in fit indices, because the configural invariance and threshold invariance models were statistically equivalent under this condition (Wu & Estabrook, 2016). The fit of the loading invariance model was found to be similar to the previous model [χ² (1050) = 2786.788, p < 0.001; CFI = 0.956; RMSEA = 0.075, (0.071, 0.078); SRMR = 0.099], such that the changes in the fit indices across the models suggested that loading invariance was demonstrated across the two versions of the questionnaire (ΔCFI = 0.003; ΔRMSEA = 0.002; ΔSRMR = 0.001). Taken together, the measurement invariance analysis indicated that the information regarding the contexts of outbursts captured by the two versions of the questionnaire could be represented by an identical factor structure with the same set of factor loadings.

The confirmatory analyses from this study contributed to the validation of a variant of the factor structure underlying the contextual items of the Emotional Outburst Questionnaire, which included the additional loadings from items concerning familiar people and settings onto the Safety factor, which originally consisted of items that would more directly contribute to the perceived safety of an environment (e.g., a private setting or a person that the individual liked; Supplementary Fig. 2 & 3). Although these modifications were primarily data-driven, they nevertheless appeared to be theoretically consistent, as individuals may be more likely to perceive familiar contexts as safe (Brosschot et al., 2018). The measurement invariance of the contextual items in the questionnaire provided further evidence regarding the robustness of the measure across samples and across cultures. Within the field, the ability to compare results across studies of emotional outbursts in people with neurodevelopmental disorders has been limited by the range of approaches that have been used to systematically characterise outbursts in previous studies (e.g., Beauchamp-Châtel et al., 2019; Rice et al., 2018; Tunnicliffe et al., 2014), which has further precluded the consideration of the impact of culture on the aetiology of emotional outbursts. The Emotional Outburst Questionnaire may be able to overcome this barrier and enable direct comparisons across studies utilising the questionnaire and across cultures, but further validation of both the contextual items and the remaining items is warranted.

The adaptation of the questionnaire into Brazilian Portuguese may facilitate future efforts to investigate emotional outbursts experienced by people in Brazil, a population for whom information regarding outbursts is scarce. Indeed, such endeavours will be critical in refining the aetiological account of emotional outbursts into a culturally sensitive framework, as the majority of the previous literature regarding outbursts has been based on a Western perspective. Cross-cultural comparisons between individuals from additional countries may further reveal the relevance of cultural factors, such as the orientation towards social independence versus interdependence, in the manifestation of emotional outbursts. Furthermore, the inclusion of culturally sensitive measures in these comparisons may simultaneously identify sources of between-group differences and capture within-group variability, which may both be relevant to the observed heterogeneity of emotional outbursts. For example, in regard to cross-cultural comparisons of emotional outbursts between children and young people in Brazil and the UK, it may be beneficial to include the Beliefs about Emotions Scale (Rimes & Chalder, 2010), as the measure appeared to capture differences between how adults in the UK and Brazil conceptualise the beliefs of the acceptability of negative emotion expression (Mograbi et al., 2018).

The three clusters derived from the refined factor scores of the Brazilian responses appeared to largely resemble the clusters previously identified in the English study, in which the clusters were distinguished by (1) high mean scores across contextual factors; (2) high mean Safety score; and (3) high mean Cross-settings score. The similarities across the samples indicate that these clusters may represent – regardless of the country – cross-cultural pathways that describe the variable aetiology of emotional outbursts, which could form the basis of pathway-specific interventions. Additional cross-cultural validation of the Emotional Outburst Questionnaire may facilitate the use of the questionnaire as a screening tool to support the development and delivery of these interventions by ensuring that the outbursts of individuals are classified in a culturally sensitive manner. Furthermore, future work should assess the replicability of the cluster structure of responses in samples from additional cultures so that the potential generalisability of the present framework could be appraised. If the overall pattern of pathways were indeed found to be cross-culturally consistent, fine-grained cultural differences could nevertheless add a degree of variability to the overall pattern of contexts associated with each pathway (e.g., Caron et al., 2012; Gilbert, 2014; Hull et al., 2020).

Indeed, several differences between the clusters derived from refined factor scores from the two samples were identified in this study. One notable difference was the relatively high mean Threat to Self score for the Perceived Safety cluster in the responses from caregivers in Brazil. This difference may be related to the motivation for a person to mask their emotions to hide characteristics that may be perceived by others as less socially desirable (e.g., Cook et al., 2021). Therefore, a person who is motivated to maintain a socially desirable impression on others by masking their emotions may also be more likely to react negatively in the form of an emotional outburst if their self-image or self-esteem is threatened. The prominence of the mean Threat to Self score in the Perceived Safety cluster of the Brazil sample appears to be consistent with a cultural perspective, as individuals from Brazil may place more value on how they are perceived by others within their community, and the individuals may also receive more prejudice and stigma from the community that may negatively impact their self-esteem (Dessen & Torres, 2019; Paula et al., 2020).

A second difference was that the mean Cross-settings score for the Perceived Unsafety cluster was not as high in the Brazilian responses compared to that of the corresponding cluster in the responses to the English version of the questionnaire. Given that there is less social integration into the wider community for individuals with ASD in Brazil, contexts associated with low perceived safety may be more widespread, such that the Cross-settings factor of the questionnaire could not adequately capture these contexts in the Brazilian responses (Gomes et al., 2015; Weissheimer et al., 2021). Additionally, it is possible that this difference in mean Cross-settings score stemmed from the subjective nature of safety perception. In support of the importance of the subjectivity of contexts in the manifestation of emotional outbursts, a previous study involving adolescents with ASD from the UK reported that contrary to expectations, one participant regarded unpredictable changes to routines as positive rather than negative (Acker et al., 2018). The authors identified that instead of the unpredictability of situations, the perception of being pressurised or rushed was more salient to the manifestation of emotional outbursts for the young people in the study (Acker et al., 2018). Therefore, the investigation of safety perception in relation to emotional outbursts may benefit from further qualitative work with young people who experience outbursts to explore the salience and cross-cultural differences of this relationship.

In contrast to the English study, which had found associations of autism spectrum disorder with the Sensory Sensitivity cluster and intellectual disability with the Perceived Unsafety cluster (Chung et al., 2022), no such associations were identified in the responses from caregivers in Brazil. This lack of association between cluster membership and diagnostic status may be due to differences in the composition of diagnoses in the two samples, as the proportion of young people with a diagnosis of autism spectrum disorder was greater in the Brazilian sample than the English sample, whereas the proportion of young people with intellectual disability was lower in the Brazilian sample (Supplementary Table 1). Regarding other demographic factors in the present study, young people in the Perceived Safety cluster were observed to be less likely to receive medication for outbursts and more likely to have accessed other forms of social or psychological support. These discrepancies in access to medication and other forms of support may potentially stem from individuals in the Perceived Safety cluster having a higher level of adaptive functioning compared to individuals in other clusters, as the underlying mechanism for the Perceived Safety cluster is hypothesised to involve the suppression of negative emotions outside of safe settings, which may require greater regulatory ability to achieve (Chung et al., 2022). Therefore, young people in the Perceived Safety cluster may be more likely to receive psychological as opposed to pharmacological interventions, as some psychological interventions for challenging behaviours have been found to be more effective for people with greater adaptive functioning (e.g., mindfulness training; for recent review, see Woodcock & Blackwell, 2020). From a cultural perspective, it is also possible that families are less likely to access medication over other forms of support for emotional outbursts because the impact of the outbursts experienced by young people in the Perceived Safety cluster may be perceived to be lower in the context of a relatively interdependent culture, as these outbursts are mostly experienced in safe settings such as at home or in a private environment, rather than in contexts where unfamiliar community members may be present.

In terms of classification using the non-refined factor scores from the Brazilian responses, the distinguishing features of the centroids for the Sensory Sensitivity and Perceived Safety clusters were maintained, but the mean Cross-settings score for the Perceived Unsafety cluster was not prominent. The use of unweighted non-refined factor scores would be preferable to regression-based refined factor scores, as it would allow new responses to be scored and classified without the prerequisite of conducting factor and cluster analyses, which would be beneficial for studies with smaller samples or those focusing on specific subsets of emotional outbursts. However, the present study demonstrated that this procedure may not be feasible, as the non-refined factor scores calculated for responses from Brazilian caregivers could not be reliably classified into existing clusters from the English study. It is possible that the loss of information from the use of unweighted factor scores hinders reliable classification in this way, so a compromise could potentially be achieved through the use of factor scores weighted by factor loadings (DiStefano et al., 2009). Additionally, there may be inherent cross-cultural variability in the centroids of the clusters that prevent cross-cultural classification. Therefore, further within- and cross-cultural comparisons should be conducted to investigate whether standardised cluster centroids could be derived either within or across cultures.

The present study was limited by the diagnostic homogeneity of the young people within the sample, which precluded further investigations into potential diagnostic differences in cluster membership. However, the converging evidence between the English study, which included higher heterogeneity in terms of diagnoses of young people, and the current study suggest that the proposed pathways may indeed be transdiagnostic (Chung et al., 2022). Future studies could address this question more directly by using a between-groups design to compare the patterns of outbursts between individuals with different diagnoses. The broader demographic differences between the two samples should be considered and potentially controlled for in subsequent studies, as the potential effects of these demographic differences could not be accounted for in the present study. For example, investigations focusing on the emotional outbursts experienced by young people in the upper age range of the present study may be beneficial, as they were less represented in the present sample. Furthermore, as the Emotional Outburst Questionnaire is an informant-report measure, the potential effect of caregiver educational level on the accuracy of responses should be examined (e.g., Van Roy et al., 2010). A critical avenue for future work lies in assessing the associations between cluster membership and measures of differences in the proposed underlying mechanisms to further delineate the aetiology of emotional outbursts.