Based on a large cohort of infants, we investigated whether it was possible to predict diagnosis of psychiatric disorders from analysis of mother-infant joint attention behaviours in social-communicative interaction at 12 months.
Specifically, we examined three JA behaviours: shared look rate (count/min), shared attention rate (count/min)
and shared attention intensity (% time). There was no evidence that JA at one year strongly predicts psychopathology at age seven. None of the three JA measures showed a significant association with the primary outcome of case–control status. Only shared look rate predicted any of the exploratory sub-diagnosis outcomes and was found to be positively associated with later oppositional-conduct disorders.
It is possible that other associations exist but were not detected in this study. There are a variety of explanations for this. Firstly, the methodological shortcomings of our study need to be considered. Our study was powered to detect only strong effects which was something that we could not modify in this exploratory study. With regards to the sub-diagnoses, some of these had very small sample sizes, so associations would have had to be very strong indeed to have been detected. Potential existence of weaker associations, not detectable by the present exploratory study, may be due to the case group being too broadly defined, so that true associations between JA measures and sub-diagnoses might be hidden due to being combined in the case group with diagnoses with no association or opposite associations. However, this explanation is not well supported by the results of the tests for association between the three joint attention measures and the five sub-diagnoses, since only one of the 15 tests – between the rate of shared looks and any oppositional-conduct disorder – was significant, which is close to the number expected due to chance alone.
The angle of the camera recording the videos is a potential limitation as discussed earlier in the methods section. Additionally, it is possible that the structure of the book situation (caregiver asked to share a picture book with their infant and engage their child in this activity as they would at home) reduced the social demand of the context and modulated the child’s activity and behaviour masking possible associations between JA behaviours and later diagnosis. Therefore, the scaffolding measures adopted during the task by the caregiver is creating a more controlled and limiting environment than say a free play situation. Nevertheless, the observers were able to assess variation in JA levels between videos, and do this with high inter-rater reliability. However, it may simply be the case that there really are no other associations or significant predictors of later diagnoses observable from the JA behaviours and the single significant result was due to chance. There is also the issue of accuracy of the DAWBA version which was used in identifying psychopathology in the present study. It has been considered limited in its ability to identify autism – the five cases of PDD were not identified as having an ASD by the clinicians. The PDD diagnoses were not made using the specific section of the DAWBA which was developed later, instead the diagnoses were made incidentally from other questions. In sum, we found no evidence that JA at one year strongly predicts psychopathology as a whole at age seven.
From previous literature, it is possible that the reason no association was found between JA and psychopathology in the present early sample is that psychiatric disorders, including ADHD, have been found to take different trajectories from infancy to adolescence [50, 51]. Shared attention also failed to predict autism (one of the five types of pervasive developmental disorders) which may be due to the caregiver compensating for the infant’s behaviour . Previous studies have found shared attention or JA difficulties in infancy  and it is even argued to be one of the earliest signs of autism [31, 32]. The small sample size of only five cases may be a reason for the lack of significant findings with respect to the prediction of later diagnosis of pervasive development disorder. It is also possible that joint attention between caregivers and infants can neither explain nor predict later psychopathology, so that the association between shared look rate and any oppositional-conduct disorder was a chance result. Additionally, there may be cases within the control group and vice versa [52, 53], since both under-diagnosis and over-diagnosis routinely occur in ADHD [54, 55]. Surprisingly our study found that, of the three joint attention behaviours, only shared look was positively associated with later oppositional-conduct disorders and none of the joint attention behaviours predicted or were associated with ADHD. A potential interpretation of this finding is that the mothers of the infants later diagnosed with ADHD were perhaps (even on an unconscious level) controlling the infants impulsivity and “watching” the infants behaviour. It may also be an indication of extroversion in both the mother and infant or infant alone. Previous research however has shown in older children aged between 4–8 years that, compared with controls, children with oppositional defiant disorder expressed lower levels of affection back towards their mothers; those with high levels of callous-unemotional traits showed significantly lower levels of affection than the children lacking these traits. The former group exhibited lower levels of eye contact toward their mothers. These impairments were found to be independent of maternal behaviour. No group difference in affection and eye contact expressed by the mothers was found .
We have presented a case–control study nested within a prospective longitudinal cohort study. The longitudinal nature of the study is one of its main strengths. Previous studies have either been retrospective or have sampled high risk referred children or siblings of affected individuals. Retrospective studies are limited in that they are primarily based on parental reports which are often biased and subject to recall/memory problems. Here we report the first study of the utility of measures of joint attention in early mother-infant interaction in predicting later onset of childhood psychopathology, based on a large cohort of infants from the ALSPAC community-based cohort. Another important strength of the present study is that all the children in the study received an independent psychiatric assessment at age seven years using the DAWBA . Lastly, we made a partial adjustment for caregiver psychopathology via the maternal depression rating which is important to strengthen the conclusions we draw from our findings given that there is much evidence strongly indicating the impact of maternal psychopathology on infant cognitive and psychological development [56, 57] and behaviour . On the other hand, one study found that maternal depression (whether prior to the birth, postpartum, or at nine months) had little impact on JA between a caregiver and nine month old infants and therefore, relationships between JA and maternal behaviour reflect infants’ social interactions with their mothers, not depression per se[59, 60].
A future study could also improve the quality of video recording to ensure that parent and child faces are always in optimum view. A new larger cohort, comprising ‘at-risk’ infants using more task conditions (i.e. play, feed, etc.), could be implemented using a placement of video equipment which would enable the capturing of more information. These more naturalistic settings might reveal more than the constrained setting of TIM and a previous study has found it to be an effective method . It remains to be established whether analyses of this kind can contribute to the development of screening instruments for disorders amenable to early intervention .