Relationship between social cognition and social behaviour following traumatic brain injury

ABSTRACT

Objective: To present an overview of studies that investigated associations between social cognition functions (social cue perception, empathy, understanding intentions) and social behaviour or social outcome following traumatic brain injury (TBI).

Methods: The literature search was conducted in the Medline, PsycInfo, Cochrane Library and Web of Science databases. Main criteria for selection were that the participants were adult persons with TBI, social cognition as well as social behaviour or social outcome post-TBI was assessed and correlations between social cognition and social behaviour or outcome were reported. Average correlations were calculated based on weighted summation of the correlations from the individual studies.

Results: Of the 511 publications identified in the search, 13 were selected. Ten of these assessed emotion recognition, six assessed understanding of intentions and two assessed empathy. Average correlations between social cognition performance and post-injury social behaviour or outcome were significant for each of the three social cognition functions; poorer performance was associated with poorer outcome. Effect sizes varied from small to moderate (understanding intentions) to moderate (emotion recognition) to large (empathy).

Conclusions: The associations were in line with models of social cognition and proposals that impairments in social cognition may underlie social behaviour difficulties and poor social outcome following TBI.

KEYWORDS:

• Traumatic brain injury
• social behaviour
• social outcome
• social cognition

Introduction

Changes in behaviour following traumatic brain injury (TBI) are not rare and prevalence rates of 25–88% of persons with moderate or severe TBI have been reported, with higher prevalence rates associated with more severe TBI (1–3). Changes in emotional and social behaviour can include indifference, egocentric behaviour, emotional lability, poor social judgement and communication, aggression, impulsive, disinhibited or irritable behaviour (4,5). Such changes can have several adverse consequences for persons with TBI and their family. Many persons with TBI may fail to return to work or maintain meaningful social relationships as a result of these changes (6,7). Survivors of TBI may find it more difficult to establish and maintain close personal relationships, friendships or relationships with work colleagues, which can contribute to failure to reintegrate back into the workplace (8,9). For relatives of persons with TBI, these behavioural changes are often a greater burden than the physical or cognitive impairments (10,11), even many years after the injury (12). For relatives and caregivers of patients with TBI, changes in social behaviour can contribute to caregiver burden (13).

Although the adverse consequences of changes in emotional and social behaviour are well documented, relatively little is known about the impairments that may underlie and predict these changes. In terms of the WHO’s International Classification of Functioning, Disability and Health, which can be useful as framework for operationalizing outcome, the question is what links impairments at the body function level with functioning at the activities and participation level (14). Models of psychosocial outcome following TBI include neuropsychological deficits as factors directly contributing to post-injury behaviour (15,16). Outcome research has originally focussed primarily on cognitive functions such as memory, processing speed or attention, as possible predictors of outcome. Impairments in cognitive functions occur very frequently following TBI. However, associations between impairments in cognitive functions and social outcome and behaviour following TBI are fairly weak (8,17) and good recovery in the cognitive domain is no guarantee for good social outcome (18). Potentially more relevant for social outcome could be social cognition, which refers to those functions involved in recognizing and responding to social information. Models of social cognition have proposed three processes in social cognition that would underlie adequate social functioning (19–21).These processes include (1) perception of social cues; (2) empathy or being able to understand other people’s experience; (3) understanding other people’s intentions. Impairments in any of these functions could result in maladaptive social behaviour and poor social outcome. Two of the models of social cognition were originally proposed to account for poor social functioning in persons with schizophrenia (20,21). Research in schizophrenia has had a longer tradition of investigating associations between social cognition and social functioning than research in TBI. Links between deficits in social cognition, including facial affect recognition and understanding intentions, and difficulties in social behaviour or social integration were found in persons with schizophrenia, although not consistently by all studies (22,23).

Deficits in the social cognition functions proposed by the models mentioned above have also been reported in persons with TBI, in particular moderate-to-severe TBI. Impaired recognition of social cues following TBI has been reported in several studies. The social cues assessed were primarily expressions of emotions (24–26). A review by Babbage et al. (27) estimated that 39% of persons with moderate-to-severe TBI experience impairments in emotion recognition. Reduced levels of empathy have also been found after moderate-to-severe TBI (5,28) and lack of empathy in the person with TBI had a detrimental effect on life satisfaction of their relatives and caregivers (29). Understanding other people’s intentions and beliefs, also referred to as theory of mind (ToM), can be impaired following TBI (30–33). Martin-Rodriguez and Leon-Carrion (34) reviewed more than 20 studies that assessed ToM in patients with acquired brain damage. ToM impairments were found on a variety of different tasks with medium-to-large effect sizes.

While there have been numerous studies reporting impairments in the functions proposed as prerequisites for adequate social behaviour following TBI (for review see also McDonald (35)), there have been far fewer attempts to investigate the association between impairments in these functions and post-injury behaviour. Studies reporting impairments in social cognition tended to assume that these impairments contribute to the changes in behaviour and poor social outcome following TBI, without actually testing that assumption. If impairments in social cognition could result in maladaptive social behaviour, as models of social cognition propose, one would expect associations between social cognition ability and post-injury behaviour. Overall, measures of social cognition appeared sensitive to the effects of brain injury, but are these measures also sensitive to the functions relevant for social behaviour and outcome? Despite the face validity of many tests of social cognition, their ecological validity is largely unknown (36).

The aim of the current paper was to review studies that attempted to link impairments in social cognition functions with social behaviour and social outcome following TBI. If social cognition functions play a role in social behaviour, performance on social cognition tasks should be associated with social behaviour or social functioning, while acknowledging of course that other factors than social cognition can also influence post-injury behaviour and social functioning.

Method

The literature search included the following databases: Medline, PsycInfo, Cochrane Library and Web of Science. Search terms used were ‘traumatic brain injury’ or ‘head injury’ or ‘acquired brain injury’ in combination with ‘social behaviour’ or ‘social functioning’ or ‘social outcome’ or ‘social integration’ or ‘social communication’. These terms were combined with ‘social cognition’ or ‘emotion recognition’ or ‘empathy’ or ‘intention’ or ‘theory of mind’. Main inclusion criteria were that the studies were published in peer-reviewed journals, the participants were adult persons with TBI, social cognition assessment included at least one of the three domains proposed (social cue perception, empathy, understanding intentions), post-injury behaviour or social functioning was assessed using self-ratings or informant ratings and associations were reported between performance on the social cognition tasks and ratings of behaviour within the TBI group. Ratings of behaviour comprised a range of measures assessing interpersonal behaviour, adequate social interactions and social integration. Informant ratings are typically recommended for assessment of behaviour in persons with brain injuries, due to the risk of poor insight with patient self-report. Therefore, in case both self-report and informant ratings were available, the informant ratings were selected for this review.

In case multiple studies were identified that assessed the same social cognition domain, an average correlation was calculated based on the correlations between social cognition and post-TBI behaviour reported in the individual studies. The average correlation within a domain was calculated using meta-analytic computations. In line with recommendations by Borenstein, Hedges, Higgins and Rothstein (37), correlations (r) were first transformed to Fisher’s z scores. These scores were weighted by the inverse of the within-study variance plus the between-study variance under the random effects model. A weighted mean was calculated by dividing the sum of the weighted z scores of the individual studies by the sum of the weights. This weighted average Fisher’s z score was subsequently transformed back to a weighted average correlation (37). The sign of the individual correlations (positive or negative) was not taken into account when calculating the average correlation, only the magnitude of the correlations. The average correlation therefore represented the effect size of the association for that domain. In case multiple outcomes within the same domain were reported in the same paper (e.g. separate correlations for two measures of emotion recognition), only one correlation was included in calculation of the average correlation. In such cases, a conservative approach was taken and the lowest correlation was selected.

Results

The total number of studies returned in the search was 511. After removal of duplicates and papers that did not meet the inclusion criteria concerning adult persons with TBI or associations between social cognition and post-injury behaviour, 13 papers were selected. The selected studies varied in the number of social cognition domains assessed, the measures used to assess social cognition, the aspects over post-injury behaviour assessed and the rating scales used. The selected studies were grouped into the three social cognition domains mentioned above and presented accordingly. Studies that assessed more than one domain are presented under each of those domains.

Perception of social cues

Ten studies were identified that examined correlations between perception of social cues and post-injury behaviour in persons with TBI. In all studies, the social cues assessed were expressions of emotions and recognition of emotions in the face was the most common task. Four studies presented facial expressions of emotions only, four presented emotions expressed in the tone-of-voice, in addition to facial expressions and two studies presented emotions expressed in both face and body postures (see Table 1). Post-TBI social behaviour and outcome was assessed with a range of different measures representing different aspects of functioning, including social communication, social integration (e.g. contact with friends), socially appropriate behaviour and social outcome (e.g. return to productive activities). Ratings could be based on self-report by the person with TBI or informant report by family members or close friends of the person with TBI. Six of the 10 studies reported significant (p < 0.05) associations between emotion recognition in persons with TBI and ratings of their post-injury behaviour in the expected direction; better emotion recognition was associated with fewer behavioural problems. Four studies reported no significant correlations (p > 0.05), even though the correlations were in the expected direction. The selected studies are described below.

Table 1. Overview of studies reporting correlations between perception of social cues and social behaviour following TBI.

Study	Sample size (n)	Social cognition measures	Ratings of post-injury behaviour	Correlation (r)	Results	Relative weight (%)
Knox & Douglas (38)	13	Recognizing emotions in faces and body postures	R-CHART occupation and social integration (self-report)	0.65	Better emotion recognition - better social outcome	3
May et al. (39)	40	Recognizing emotions in faces	DEX social behaviour (proxy)	−0.43	Better emotion recognition - fewer difficulties in social behaviour	11
Milders et al. (40)	17	Recognizing emotions in faces and tone-of-voice	NBAP (proxy)	0.20	No significant correlation emotion recognition – social and emotional behaviour	4
Milders et al. (36)	33	Recognizing emotions in faces and tone-of-voice	NBAP pragnosia - effective communication (proxy)	−0.41	Better emotion recognition - fewer difficulties in effective communication	8
Osborne-Crowley & McDonald, (41)	23	Recognizing emotions in dynamic face and body postures	SPRS interpersonal relationships (proxy)	−0.34	No significant correlation emotion recognition – changes in interpersonal relationships	6
Ryan et al. (37)	34	Recognizing emotions in faces and tone-of-voice	ABCL externalizing behaviour (proxy)	−0.54	Better emotion recognition in tone-of voice - fewer behavioural problems	9
Saxton et al. (42)	24	Recognizing emotions in faces	KBCI interpersonal difficulties (self-report)	0.15	No significant association between emotion recognition and behaviour	6
Spikman et al. (43)	51	Recognizing emotions in faces	DEX (proxy)	−0.38	Better emotion recognition - fewer behavioural problems	14
Struchen et al. (35)	121	Recognizing emotions in faces and tone-of-voice	CHART-SF social integration (self-report)	0.26	Better emotion recognition - better social relationships	34
Ubukata et al. (44)	20	Recognizing emotions in faces	R-CHART social integration (self-report)	0.21	No significant association between emotion recognition and social outcome	5
Total n	378		Average r	0.36		100%

ABCL: Adult Behaviour Check-List; CHART-SF: Craig Handicap Assessment and Reporting Technique – Short form; R-CHART: Revised-Craig Handicap Assessment and Reporting Technique; DEX: Dysexecutive Questionnaire; KBCI: Key Behaviours Change Inventory; NBAP: Neuropsychology Behaviour and Affect Profile; SPRS: Sydney Psychosocial Reintegration Scale.

Struchen et al. (40) included by far the largest sample with 121 persons with TBI, most of whom had suffered severe TBI. Both recognition of facial emotions and emotions in the tone-of-voice were assessed and performance on both tasks was associated with self-ratings of social outcome (CHART-SF); better emotion recognition was associated with better ability to maintain social relationships and to participate in work or other productive activities. Only the lowest of the four correlations reported by Struchen et al. (40) is presented in Table 1 and was included in calculation of the average correlation. Milders et al. (43) also assessed recognition of emotions expressed in faces and voices and found a significant association between an emotion sum score, containing both expressions in the face and in voice, and informant ratings of effective social communication; better emotion recognition was associated with more effective communication. Ryan et al. (42) reported associations between recognition of emotions expressed in the tone-of-voice, but not recognition of facial emotions, and informant ratings of social communication (r = −0.54) as well as informant ratings of externalizing behaviour (rule breaking, aggressive or intrusive behaviour; r = −0.56); poorer emotion recognition was associated with more communication difficulties and more externalizing behaviour. The lower of the two correlations is presented in Table 1 and included in calculation of the average score. Milders et al. (44) assessed emotion recognition in faces and in tone-of-voice but found no significant association between either task or proxy ratings of post-injury emotional and social behaviour. Spikman et al. (41), Saxton et al. (38) and Ubukata et al. (39) assessed emotion recognition using facial expressions only. Spikman et al. (41) reported an association between emotion recognition performance in persons with TBI and proxy ratings of post-injury behaviour; those persons with TBI who were worse at emotion recognition showed more impulsive behaviour or problems in social behaviour. However, neither Saxton et al. (38) nor Ubukata et al. (39) found significant correlations between facial affect recognition and self-reported post-injury interpersonal and communication difficulties.

Finally, three studies assessed emotion recognition using dynamic stimuli. Osborne-Crowley and McDonald (45) presented emotions expressed in dynamic facial expressions and body posture in video vignettes (The Awareness of Social Inference Test – TASIT). Despite naturalistic stimuli, emotion recognition performance was not significantly associated with proxy ratings of post-injury behaviour concerning interpersonal relationships. Knox and Douglas (46) also used the TASIT but in their study emotion recognition was positively associated with self-reported social outcome. May et al. (47) used two tasks to assess emotion recognition, one presenting static facial expressions and one presenting dynamic facial expressions. A sum score containing performance on both expression tasks was associated with proxy ratings of post-injury social behaviour (r = −0.43) and social outcome, in particular return to work and productive activities (r = 0.43). Better emotion recognition was associated with fewer behavioural problems and better outcome. Only the correlation between emotion recognition and social behaviour is presented in Table 1 and was included in calculation of the average correlation.

Of the six studies examining associations between emotion recognition and post-injury behaviour based on proxy ratings, four also collected self-ratings on the same ratings scales. In three of these studies (41–43), the correlations were weaker with self-report than with proxy ratings and failed to reach significance. The fourth study (44) did not report correlations based on self-ratings.

An average correlation based on the correlations from the individual studies mentioned above and listed in Table 1 was calculated using the procedure outlined in the ‘Methods’ section. The average correlation was r = 0.35 (95% confidence interval: r = 0.25–0.43), which corresponds to a medium effect size. This correlation was significantly different from zero (z = 6.63, p < 0.0001), suggesting that when averaging results across 10 studies with a combined sample of 378 persons with TBI, recognition of emotional expressions was associated with social behaviour following TBI.

Empathy

Although empathy in persons with TBI has been assessed in various studies (35), only two studies were identified that reported associations between empathy ratings and post-TBI behaviour. De Souza et al. (28) found a significant correlation (r = −0.57) between self-report empathy ratings on a commonly used empathy scale, the Balanced Emotional Empathy Scale, and proxy ratings of post-injury behaviour on the Current Behaviour Scale. Persons with TBI who showed more disorders in drive and control, in particular loss of motivation, also reported lower levels of emotional empathy. Saxton et al. (38) reported empathy self-ratings on the Interpersonal Reactivity Index to be significantly correlated with self-ratings of changes in interpersonal behaviours (r = −0.56). Higher levels of interpersonal difficulties were associated with lower levels of empathy. As only two studies were identified in the empathy domain, no average correlation was calculated. All associations reported were significant and even the lowest correlation reported (r = −0.56) corresponds to a large effect size.

Understanding other people’s intentions and ToM

Six studies were identified that correlated ratings of behaviour following TBI with the ability to understand intentions or ToM (see Table 2). Each study used different measures to assess ToM or intention understanding. Byom and Turkstra (48) counted the number of mental state terms (MSTs) used in discourse as an indication of ToM. MSTs are words that refer to thoughts, feeling or desires. Persons with moderate-to-severe TBI used fewer MST than healthy controls and the frequency of MST was associated with the quality of social communication, as rated by an independent observer. More MSTs in discourse were associated with socially more acceptable communication. Ubukata et al. (39) found no significant correlation between ToM perception, as assessed with the Mind in the Eyes test and the Faux Pas test, and social outcome. Milders et al. (44) also used the Faux Pas test to assess ToM. Performance on the Faux Pas task was associated with proxy ratings of difficulties in social and emotional behaviour following TBI, although the association just failed to reach significance. However, another study (43) using a ToM sum score, containing the Faux Pas test and a visual ToM test (the Cartoon test), found no significant association with proxy ratings of post-TBI social and emotional behaviour. Similarly, May et al. (47) found no significant association between proxy ratings of social behaviour following TBI and performance on four tasks of intention understanding, including the Faux Pas task, was summarized into a single sum score. On the other hand, Struchen et al. (49) reported an association between the ability to identify inappropriate behaviour in video vignettes of social situations and self-ratings of social integration in a substantial sample (n = 184) of persons who had sustained TBI at least 6 months earlier. Of the four studies examining the association between understanding intentions and post injury behaviour based on proxy ratings, two also collected self-ratings (43,44). In one of these studies (43), the correlation with self-ratings was not significant, similar as with the proxy ratings. The other study (44) did not report correlations based on self-ratings.

Table 2. Overview of studies reporting correlations between understanding of intention or theory-of-mind and social behaviour following TBI.

Study	Sample size (n)	Intention/theory-of-mind	Ratings of post-injury behaviour	Correlation (r)	Results	Relative weight (%)
Byom & Turkstra (45)	20	Mental state words in discourse	Quality of social communication (independent raters)	−0.46	More mental state words – more socially acceptable communication	6
May et al. (39)	40	Faux Pas test, Hinting test, ToM Cartoon test, Cartoon Predictions test	DEX social behaviour (proxy)	−0.25	No significant association intention understanding sum score - difficulties in social behaviour12	12
Milders et al. (40)	17	Faux Pas test	NBAP (proxy)	−0.61	Better performance on Faux Pas - fewer behavioural problems (trend)	5
Milders et al. (36)	33	Faux Pas test, Cartoon test	NBAP (proxy)	0.03	No significant association intention understanding sum score - behavioural problems	10
Struchen et al. (46)	184	AIPSS problem identification	CIQ social integration (self-report)	0.15	Better understanding social situations – better social outcome	61
Ubukata et al. (44)	20	Mind in the Eyes tests	R-CHART social integration (self-report)	0.10	No significant association intention understanding and social outcome	6
Total n	314		Average r	0.24		100%

R-CHART: Revised-Craig Handicap Assessment and Reporting Technique; DEX: Dysexecutive Questionnaire; NBAP: Neuropsychology Behaviour and Affect Profile; AIPSS: Assessment of Interpersonal Problem Solving Skills; CIQ: Community Integration Questionnaire.

The correlations from the six individual studies were averaged into a single correlation, using the procedure outlined in the ‘Methods’ section. The average correlation was r = 0.24 (95% confidence interval r = 0.03–0.44), which corresponds to a small-to-medium effect. Despite the fact that several individual studies failed to find significant associations between intention understanding and ratings of behaviour, the average correlation across the six studies and a combined sample of 314 persons with TBI was significantly different from zero (z = 2.26, p < 0.05), suggesting an association between understanding intentions or ToM and social behaviour following TBI.

Discussion

The current review of studies into the relationship between social cognition and social behaviour following TBI found associations for the three domains of social cognition examined: social cue perception, empathy and understanding intentions or ToM. Persons with TBI who performed more poorly on tests of social cognition tended to show more maladaptive social behaviour or poorer social outcome and social integration. Although the results of individual studies in two domains, perception of social cues and understanding of intentions, could fail to reach significance, the average correlations based on weighted summation of the individual correlations were significantly above zero. The average correlations for social cue perception and intention understanding corresponded to a medium effect size and a small-to-medium effect size, respectively (50). If these effect sizes would represent population effect sizes, it is not surprising that several studies failed to find significant associations due to insufficient power. Given the modest sample sizes of most studies, their statistical power would have been limited. For empathy only two studies were included, both reporting similar and significant correlations. Both correlations corresponded to a large effect size, but as this effect size was based on a very limited number of studies it should be interpreted with caution. The associations between social cognition and post-TBI behaviour were found despite a wide variety of tasks used to assess emotion recognition, understanding intentions or empathy and different rating scale to assess behaviour. Therefore, the associations do not appear to depend on a particular task or questionnaire.

The results are in line with models of social cognition (19–21) and partially support the models’ proposal that social cognition abilities are important for social functioning and that impairments may result in difficulties with social behaviour and poor social outcome. Three domains of social cognition that he models distinguished were associated, in the predicted direction, with ratings of social behaviour and social outcome following TBI. However, the associations provide no indication of the direction of the causal relationship. The models of social cognition assume that impairments in social cognition functions lead to maladaptive social behaviour. Alternatively, it is also conceivable that the changes in behaviour and the resulting reduction in social interactions affect the ability to recognize social cues and other social cognitive functions. In that case one would expect social cognition impairments to become worse over time, which was not what was found in follow-up studies in persons with TBI. Ietswaart et al. (25) and Milders et al. (33) found that impairments in emotion recognition and understanding intentions were present shortly after moderate-to-severe TBI and remained largely unchanged at 1-year follow-up.

Another alternative explanation for the associations found may be that the correlations between social cognition and post-injury behaviour are spurious and caused by a third factor, such as injury severity. More severe TBI is likely to be accompanied by more severe social cognition impairments (26,42) and more frequent behavioural problems (1–3). A number of studies included in this review explicitly examined the influence of injury severity on the association between social cognition and post-injury behaviour. These studies found that controlling for injury severity had little effect on the association between social behaviour and emotion recognition (47) or empathy (28). Therefore, it is unlikely that injury severity could fully account for the associations reported.

Even though the associations as expressed in average correlations are in line with the models of social cognition, the contribution of social cognition to explain social behaviour or outcome after TBI was modest. At best 31% of the variance in post-TBI behaviour was explained in the case of empathy, based on only two studies, and for social cue perception and intention understanding these percentages were considerably smaller, 13% and 6%, respectively. Clearly additional factors are required to predict social outcome. Such factors could include other cognitive abilities such as self-monitoring, self-control or initiation of action (35,51,52) or environmental factors, personal factors and coping (16).

In terms of implications for clinical practice and rehabilitation, the results of this review showed that evidence was strongest for a link between emotion recognition ability and social outcome after TBI. Therefore, as potential target for interventions aimed at improving social outcome post-TBI, emotion recognition may have merit. Recent studies have shown that rehabilitation of emotion recognition is possible in persons with TBI (53,54), but improvements in emotion recognition were not necessarily accompanied by changes in social behaviour. Possible reasons suggested for these null effects included, insensitivity of the outcome measurements used, which may not be able to detect subtle improvements in social functioning, or lack of generalization of the training effect to the patient’s home environment (53,54).

Limitations of this review were the limited number of studies that met the inclusion criteria and the wide heterogeneity in measures used to assess social cognition and post-injury behaviour. Almost every study employed a different combination of tasks and ratings scales. In addition, outcome rating scales could represent different constructs, e.g. post-injury behaviour or social integration. What the outcome measures did have in common was that all referred to observable actions rather than feelings or thoughts. As a result of the heterogeneity in instruments, comparisons between studies have to be considered with caution and differences in strength of the associations reported in different studies may, in part, be attributed to differences in the instruments used. This point illustrates a more general limitation of research into social cognition following TBI, namely the lack of a standard set of measures. To date there is no agreement on a standard set of social cognition measures, which is a disadvantage for both research into social cognition following brain injury as well as for clinical practice, where changes in social behaviour are frequently observed (55). Another limitation was that several studies included in this review relied on self-reports to assess post-injury behaviour, with the risk that lack of insight affected the assessment. In this context it was interesting to note that studies that collected both proxy and self-ratings tended to find stronger associations between social cognition and post-injury behaviour based on proxy ratings than ratings based on self-ratings. Finally, associations between standard cognitive tests and behaviour were not considered in this review, which could have shown whether the relationship between social cognition and behaviour was stronger than between cognition and behaviour. The main reason for not including cognitive functions was that most of the studies included in this review did not report cognitive performance or associations between cognitive functions and post-injury behaviour.

In sum, studies into social cognition impairments following TBI tend to assume that such impairments are relevant for social functioning and social outcome, but few studies have actually investigated this putative link. The current review of studies that examined the relationship between social cognition functions and social behaviour or social outcome following TBI showed significant associations for the three functions considered, perception of social cues, understanding intentions and empathy. The findings are in line with models of social cognition, although the strengths of the associations were modest and the associations were based on a limited number of studies. To strengthen evidence for an association between social cognition and behaviour, further studies would ideally present the same set of instruments in a large sample of persons with TBI.

Declaration of interest

The author report no conflicts of interest.