Abstract
Sleep problems have been reliably associated with psychotic experiences in adults and have been suggested as target for intervention. However, the relationship between sleep disorder and psychotic experiences in children has not been extensively studied despite the potential for guiding intervention. The Adolescent Brain Cognitive Development (ABCD) dataset, containing baseline and one-year follow-up data of over 11,000 10-11 year olds, was utilised to investigate this relationship. More specifically, a set of pre-registered multi-level regression models were applied to test whether a) baseline sleep disorder predicts baseline psychotic experiences cross-sectionally; b) baseline sleep disorder predicts psychotic experiences one year later; c) the persistence of sleep disorder predicts the persistence psychotic experiences at one year; d) the remission of sleep disorder predicts the remission of psychotic experiences. After controlling for potential confounders, sleep disorder was associated with psychotic experiences cross-sectionally (OR=1.40, 95% CI 1.20-1.63), at one-year follow-up (OR=1.32, 95% CI 1.11-1.57), and the persistence of sleep disorder predicted the persistence of psychotic experiences (OR=1.72, 95% CI 1.44-2.04). However, remission of sleep problems did not predict remission of psychotic experiences (OR=1.041, 95% CI 0.80-1.35). In all models where an association was found, sleep was one of the two strongest predictors of psychotic experiences (with stimulant medication being the other). The results indicate that sleep problems in children are common and strongly associated with psychotic experiences but the lack of co-remission raises questions about the mechanism of association. However given existing evidence in adults, further investigation and interest in sleep as a preventative mental health intervention in this age group is warranted.
Introduction
Sleep disorders are hypothesised to be a significant causal factor in the development and maintenance of psychosis (Reeve et al., 2015; Waite et al., 2019). Experimental research supports that sleep disruption increases psychotic experiences (Reeve, Emsley, et al., 2018), and that treating sleep problems reduces psychotic experiences in adults (Freeman et al., 2017). While the relationship between sleep and psychotic symptoms is thought to be bidirectional, longitudinal studies have indicated that sleep problems predict later psychotic symptoms to a greater extent than vice versa (Hennig & Lincoln, 2017; Reeve, Nickless, et al., 2018). These and other findings have raised the possibility of targeting sleep problems as a preventative, early intervention in mental health (Freeman et al., 2017). This proposal is further supported by sleep problems acting as a generic contributor to a range of mental health problems in addition to (or perhaps as part of) their relationship with psychosis (Freeman et al., 2020), including during childhood and adolescence (Gradisar et al., 2020).
This raises the question of how early this relationship between sleep and psychosis exists. It is known that psychotic-like experiences are roughly 10x more common in children than they are in adults (reported by 66% of 7-8 year old children compared to 5.8-7.2% of adults; (Laurens et al., 2012; Linscott & van Os, 2013; McGrath et al., 2015). Clearly not all childhood psychotic-like experiences are pathological (Maijer et al., 2019), nevertheless persistence, distress, and amount of different psychotic-like experiences reported in childhood is associated with higher risk of developing adult psychotic disorder (Dominguez et al., 2011; Fisher et al., 2013) and a predictor of poor physical and mental health more generally later in life (Davies et al., 2018; Kelleher et al., 2012; Trotta et al., 2020). The age at which psychotic-like experiences are reported is also positively related to risk of later psychotic disorder. Reporting psychotic-like experiences at age 10-11 is associated with a 5x increased likelihood of developing a later psychotic disorder (Fisher et al., 2013), whereas reporting these experiences by ages 14-17 is associated with a 10x increased likelihood (Dominguez et al., 2011). This set of findings have been encompassed under a ‘proneness to persistence’ model of psychosis, with these non-clinical psychotic experiences existing on a continuum with psychotic disorders (Linscott & van Os, 2013).
The relationship between sleep problems and psychotic experiences in children is less well-researched than in adults (Reeve et al., 2015). However, many studies have demonstrated that children or adolescents with sleep problems (including nightmares, insomnia, or oversleeping) are more likely to report psychotic experiences (Jeppesen et al., 2014; Koopman-Verhoeff et al., 2018; Y. J. Lee et al., 2012). Furthermore, longitudinal observational studies support that insomnia is specifically linked to paranoia (Bird et al., 2017; Hennig & Lincoln, 2017), with Hennig and colleagues additionally showing that the relationship seems primarily from insomnia to paranoia rather than vice versa. Several large population cohort studies have reported that sleep problems during childhood predict psychotic experiences in teenage years (Fisher et al., 2014; Morales-Muñoz et al., 2020; Thompson et al., 2015). Furthermore, as in adults, sleep problems in children are linked to a wide range of other negative mental and physical health outcomes (Carskadon & Barker, 2020; Dong et al., 2019; Gregory & Sadeh, 2012; Zhang et al., 2017).
Despite this evidence indicating that sleep problems have an important role in the onset of psychotic experiences during childhood and adolescence, it is less clear whether sleep problems contribute to persistence of psychotic experiences in this age group. This is particularly important given that persistence is a particularly strong predictor of poor outcome. The effect of remission of sleep problems on psychotic experiences is also unknown – but if improvements in sleep are associated with a reduction in psychotic experiences, this would provide evidence suggesting sleep interventions to reduce psychotic experiences could be viable.
The current study
We used the Adolescent Brain and Cognitive Development study cohort data (release 2.0.1) to investigate the relationship between sleep and psychotic experiences in children aged 9-11 to address a set of pre-registered hypotheses:
Do sleep disorders and psychotic experiences co-occur in childhood?
Do sleep disorders predict later onset of psychotic experiences?
Do sleep disorders predict persistence of psychotic experiences?
Does remission of sleep disorders predict remission of psychotic experiences?
These were examined use a pre-registered set of linear and logistic regression models applied to the baseline and one-year follow up data using a pre-registered analysis plan with all analysis available as open code: https://osf.io/8ks72/
Method
Recruitment
The Adolescent Brain and Cognitive Development (ABCD; https://abcdstudy.org/) 2.0.1 release study data was used in the current analysis (Feldstein Ewing & Luciana, 2018; Volkow et al., 2018). The ABCD is an ongoing longitudinal cohort study aimed at recruiting a representative sample of US children. At the initial stage, all families of children aged 9-10 in geographic catchment area of study sites across the USA were contacted via schools with information about the study. Families that volunteered were then contacted by the research team and screened for inclusion. Participants were purposively recruited to fulfil specific representativeness requirements to match sociodemographic factors in the US, for example utilising targets for each of five major race/ethnicity classifications (White, African-American, Hispanic, Asian All Other). The 2.0.1 release includes baseline data on 11,873 individuals, and one year follow-up from 4,951 of those participants. Full details of the ABCD study design are available in a special issue of Developmental Cognitive Neuroscience (Feldstein Ewing & Luciana, 2018).
Measures
Psychotic experiences
Children’s psychotic experiences were assessed using the Prodromal Questionnaire – Brief Child version (PQBC). This measure was adapted for and validated within the current ABCD dataset showing high reliability (Cronbach’s alpha = 0.863 for total score and 0.873 for distress subscale; (Karcher et al., 2018). The PQBC is a self-report questionnaire assessing presence and distress associated with psychotic like symptoms in children. For each psychotic-like experience the child is asked if they experience it (yes/no), and if they do, how much it distresses/bothers them on a pictographic 1 to 5 scale showing a human cartoon figure in various levels of distress. This questionnaire yields two outcome variables – the sum of symptoms endorsed (range 0-21) and the sum of distress reported (range 0-126).
For the current study, both continuous and dichotomous outcomes of the sum of symptoms were used. For the dichotomous outcome the symptom score was transformed to a categorical 0-1 variable where 1 indicates presence of at least one psychotic symptom and 0 indicates no psychotic symptoms present. When applied to baseline and follow up this was then used to derive variables indicating ‘new onset’ of psychotic symptoms (i.e. 0 at baseline, 1 at follow up) ‘persistence’ of psychotic symptoms (i.e. 1 at both time points), and ‘remission’ (1 at baseline, 0 at follow up).
Sleep disorders
Presence of sleep disorders was assessed using the Sleep Disorder Scale for Children (SDSC), which is a parent-reported questionnaire assessing the presence of a range of sleep disorder symptoms in children (Bruni et al., 1996). It is composed of 26 Likert items assessing the frequency of various disturbances over the past 6 months on a 1 to 5 Likert scale (1 = never, 5 = always/daily experiencing a particular issue). All items can be summated for a total measure of sleep disturbance, for which a cut-off point at 39 has sensitivity of 0.89 and specificity of 0.74, correctly identifying 73.4% of a control group and 89.1% of sleep disordered participants.
This cut-off was used to categorise participants according to absence or presence of disturbed sleep. The categorical score at baseline and follow up was then used to derive variables ‘persistence’ of sleep disturbance (i.e. present at both time points), and ‘remission’ (present at baseline, absent at follow up).
Potential confounders
Variables were chosen to address a range of possible confounders to the hypothesised relationship between sleep and psychosis. Gender (male/female), ethnicity (white/black/Hispanic/Asian/other) were reported within the basic demographic questionnaires of the study. Ethnicity was re-coded in to white/non-white for the purposes of all analyses. Socioeconomic status was indexed by using the sum score of seven yes/no items in the parent demographic survey questions relating to experiences of family hardship (e.g. ‘in the past 12 months has there been a time when you and your immediate family needed food but couldn’t afford to buy it or couldn’t afford to go out to get it?), with higher scores on this indicating lower socioeconomic status. Neighbourhood deprivation was assessed using the area deprivation index of the home address, which provides a national percentile where higher values indicate higher levels of deprivation. Child IQ was assessed using the WISC-V matrix reasoning subscale score (range 1-19), with higher values indicating higher IQ. Family conflict was measured using the 9-item family conflict subscale of the family experiences, each item reported by parents as true or false (e.g. ‘We fight a lot in our family’), with higher values indicating higher levels of conflict. Lastly prescribed medication fields were searched for any stimulant medications (e.g. Methylphenidate) including their trade names (e.g. Ritalin) with their presence/absence coded accordingly.
Analysis
The pre-registration document outlining the hypotheses and analysis plan, and the analysis code used in this study are available online at the following link: https://osf.io/8ks72/. R Studio version 3.6.2 was used for all analyses, Supplementary Materials 1 contains a list of packages and version numbers.
For all hypotheses, a set of planned regression analyses were pre-specified. The pre-registration was completed before the ABCD 2.0 data release, i.e. before the one-year follow-up data was made available. In each case, the regression model was first estimated with only the key explanatory (sleep) and dependent (psychosis) variables, and if a significant association was found, the process was repeated with potential confounders included in order to test robustness of the hypothesised association between sleep and psychosis.
We tested the following four hypotheses:
1. Do sleep disorders and psychotic experiences co-occur in childhood?
Tested cross-sectionally at baseline with a) a linear regression to test continuous association between sleep symptoms and psychotic experiences b) a logistic regression to test if sleep symptoms (continuous) predicted presence of psychotic experiences (dichotomous) and c) a logistic regression to test if presence of sleep disorder (dichotomous) predicted presence of psychotic experiences (dichotomous).
2. Do sleep disorders predict later onset of psychotic experiences?
We tested whether sleep disorders at baseline predict psychotic experiences at follow-up. This was investigated by logistic regression models testing if the presence of sleep disorder a) at baseline and b) at both baseline and follow-up predicted psychotic experiences at one year follow up.
3. Do sleep disorders predict persistence of psychotic experiences?
We tested whether sleep disorders at baseline predict the persistence of psychotic experiences from baseline to follow-up (i.e. psychotic experiences being present at both time points). This was investigated by logistic regression models testing if the presence or persistence of sleep disorder predicted persistence of psychotic experiences.
4. Does remission of sleep disorders predict remission of psychotic experiences?
Finally, we tested whether the remission of sleep disorders (i.e. present at baseline, absent at follow up) also predicted remission of psychotic experiences using logistic regression.
Alterations to the pre-registered statistical plan
Alterations to our analysis plan were made according to subsequently published guidance on analysis of the ABCD cohort (Heeringa & Berglund, 2020), namely introducing multi-level clustering by site and family. The planned unweighted results are reported here, but population weighted demographic and descriptive variables and weighted analyses (where population weights have been converted to scaling weights to allow their inclusion in the multi-level regression model using the ‘Method A’ and ‘Method B’ algorithm proposed by (Carle, 2009)), are available in supplementary material.
Results
Demographic and descriptive results
Table 1 displays descriptive statistics of the study. Most descriptive variables remained relatively consistent between baseline and follow-up, however the follow-up group has a preponderance of White children (59.5% compared to 52.1%) and fewer Black children (9.3% at follow up compared to 15.0% at baseline).
Sleep disorders were present in the cohort at similar proportions at baseline and follow up (30.4% and 28.4% respectively). However, this masks some turnover, with roughly a third of the sleep disorder cases at follow-up representing ‘new onset’ cases (9.8%) to replace a similar proportion that remitted over the year (10.0%). For psychotic experiences there is a clear decline (61.5% of children reporting at least one at baseline, versus 46.8% at follow-up), a result of the high remittance rate (24.6%) compared to new onset (9.6%). This is also reflected in the reduced average PQ-BC score at follow-up versus baseline.
Nevertheless, where sleep disorders and psychotic experiences were present the majority were persistent across the study period: 65.4% of sleep disorder cases and 78.7% of psychotic experience cases at follow up had been present at baseline. This is supported by cross-tabulation of sleep disorders and psychotic experiences which shows relatively low numbers of cases where both sleep disorders and psychotic experiences are new onset (1.1%) and where they have both remitted (2.5%). Out of cases where both sleep disorder and psychotic experiences were present at follow up, the vast majority were ‘persistent’ cases, where both difficulties had been present at baseline (joint persisting group represents 91.5% of joint onset and persistent cases).
RQ1: Do sleep disorders and psychotic experiences co-occur in childhood?
The regression results associated with this research question can be found in Table 2. They illustrate that sleep disorders and psychotic experiences are significantly and strongly associated within the baseline time point. Sleep disorder symptoms above cut off is associated with an OR of 1.50 (95% CI 1.30-1.74) of at least one psychotic experience being present. The OR remains significant but reduces slightly (to 1.40, 95% CI 1.20-1.63) once the control variables are added (RQ1f). In this final model, sleep disorder cut off is the largest predictor of current psychotic experiences out of all the parameters tested.
RQ2: Do sleep disorders predict onset of psychotic experiences?
The results of the logistic regression analyses on this research question are shown in Table 3. Sleep disorders were found to be strongly associated with reporting psychotic experiences in all models. Sleep disorder symptoms at baseline were found to predict psychotic experiences one year later (OR=1.62, 95% CI 1.41-1.86), even once sleep disorder symptoms at the later time point were controlled for (OR=1.41 95% CI 1.20-1.65). This association remained after control variables were added (OR=1.32, 95% CI 1.11-1.57). The only stronger predictor than sleep disorder in the final model (RQ2d) was stimulant medication (OR=1.61, 95% CI 1.22-2.10).
RQ3: Do sleep disorders predict persistence of psychotic experiences?
The results of the logistic regression analyses relating to persistence of psychotic experiences are presented in Table 4. Presence of sleep disorder symptoms at baseline (OR=1.71, 95% CI 1.48-1.97) and persistence of sleep disorder symptoms (OR=1.89, 95% CI 1.60-2.23) were found to highly significantly predict persistence of psychotic experiences at 12 months. These effects remained strong even once control variables were added (ORs of 1.56 and 1.72 for baseline and persisting sleep disorder symptoms respectively). Persistence of sleep disorder symptoms was the strongest overall predictor of persistence of psychotic experiences in the corrected model (RQ3d), with the next largest effect being stimulant medication (OR=1.46, 95% CI 1.12-1.91).
RQ4: Do sleep disorders predict remission of psychotic experiences?
Remission of sleep disorder symptoms was not a significant predictor of remission of psychotic experiences (OR=1.041, p=0.766, 95% CI 0.80, 1.35) in the uncontrolled model, therefore the analysis with control variables was not carried out.
Differences arising from weighted analysis
The results using the population weighted values are reported in the supplementary material. The pattern of all results remained substantially the same except for non-white ethnicity which became a significant predictor of onset of psychotic experiences at 12 months (alongside all other factors). See supplementary material for full details.
Discussion
We report that sleep disorder symptoms are strong and significant predictors of both the co-occurrence and persistence of psychotic experiences in 9-11 year old children, even after controlling for a range of other variables including socioeconomic status, ethnicity, IQ, and stimulant medication. However, the remission of sleep disorders did not predict remission of psychotic experiences, and we discuss possible reasons for this negative association below.
Psychotic experiences were common in our study – reported by over two-thirds of children at baseline. This is similar to previous studies in this age group (Laurens et al., 2012), and the relatively high rate of remission versus new onset by the one year follow up is also in keeping with existing developmental models of psychotic experiences. Repeating these analyses through adolescence – where persistence of psychotic experiences over time increasingly predicts risk of later psychotic disorder – will be relevant to further specify what factors are most associated with onset, persistence, and remission of psychotic experiences.
Sleep disorder symptoms were also relatively common in the study group, with just under a third of children having significant sleep issues. Guidelines for parents and clinicians on child sleep difficulties tend to advise that many difficulties ameliorate naturally over time, which is contrary to our findings that around two-thirds do not (at least over the one-year observation period reported here). Previous studies have also found a similar rate of persistence (Fricke-Oerkermann et al., 2007). There is evidence of increasing rates of child and adolescent sleep problems over time (Matricciani et al., 2012) suggesting the need for additional support for parents and health services to address childhood sleep difficulties.
Besides our primary focus on the influence of sleep on psychotic experiences, our findings support the association of a wide range of epidemiological factors (male gender, socioeconomic status, neighbourhood deprivation) with psychotic experiences in this age group. Lower socioeconomic status of parents and increased neighbourhood deprivation were consistent predictors in every model as supported by previous research (Solmi, Lewis, Zammit, & Kirkbride, 2020). Non-white ethnicity was not found to be a significant predictor in the unweighted models but was significant in some analyses once the models were weighted (see Supplementary materials), supporting the longstanding robust association reported elsewhere e.g. (Jongsma et al., 2020). The results therefore support the importance of broader scale social interventions as preventative of mental health issues, and further investigation of the mechanisms between, for example, experiences of deprivation and the development of psychotic experiences.
Lastly, the strong relationship between stimulant medication and psychotic experiences is in keeping with previous literature e.g. (Shyu et al., 2015). However, the full range of reasons for the extent of this association are not discernible from the data reported here. Given the well-established association between dopamine agonist stimulant medication and psychotic experiences (Tost, Alam, & Meyer-Lindenberg, 2010), it is likely that a proportion of this association can be accounted for as an adverse effect of the medication. However, confounding by indication is likely also to play a role and children who may fulfil criteria for prescribing may also have shared risk factors for psychotic experiences that manifest through other causal pathways (Hollis et al., 2019).
We also note the lack of a relationship between the remission of sleep problems and the remission of psychotic-experiences in this study. As this study was observational, it is not possible to confirm a causal relationship although we suggest several hypotheses. The first is that sleep problems may cause an increase and persistence in psychotic symptoms but through altering a mediating mechanism that remains when sleep problems resolve. In adult studies, the role of affective dysregulation has been highlighted as a potential mediating factor (Akram et al., 2020) with some evidence that this may also be apparent in children (Jeppesen et al., 2014) and we suggest this is a plausible candidate for such a mechanism. Alternatively, sleep problems may cause sensitisation of the mechanisms that generate psychotic symptoms but more rapidly than desensitisation occurs when sleep problems resolve, and although recovery may occur, it may need more than the year interval measured with this data. We note here that ‘mechanism’ may be neurocognitive (for example, arousal-based increases in aberrant salience) or social (for example, sleep disorder causing behaviour difficulties, exclusion or victimisation, subsequently impacting on psychotic experiences). Nevertheless, this does raise a question about the potential efficacy of sleep improvement intervention in reducing psychotic-like symptoms in preadolescent children, given than an association between improved sleep and reduced psychotic experiences would be expected if there were a simple causal association. However, we note that this study only tested remission of sleep problems as observed in the study, and it may be that a sleep intervention would address some of the shared risk factors for both poor sleep and psychotic-like experiences leading to an improvement in psychotic-like experiences not apparent in naturalistic remission.
Limitations
We note that the observational nature of this study limits the causal conclusions we can draw as we cannot rule out that any changes or associations observed are due to unobserved factors. Another limitation of this being an observational study is that the application of planned analyses is reliant on the data that emerges. This may have limited our ability to reliably test some aspects of the relationship. For example, the particularly low proportion of cases (2.71%) where psychotic experiences and sleep disorders remitted from baseline possibly limited the ability of the remission models to identify any effect (as indicated by the wide confidence intervals for these models). However, given that prior research has suggested a large proportion of psychotic experiences and sleep problems in this age group naturally remit, the lack of such an effect observed in this cohort deserves further investigation. The slight demographic shift between baseline and follow-up group is also worth considering as this may have reduced the accuracy of our estimates (even post-weighting) due to the follow-up group being less representative of the study sample. Confirmation of these findings once the full one-year dataset is released – and with later phases of ABCD data - would be worthwhile.
We have been limited in variables available by the measures used in the ABCD study, which have their own benefits and limitations. For example, while self-reported scales for psychotic experiences in adolescence are recommended, false positives are known to occur (Kelleher et al., 2011; K.-W. Lee et al., 2016) and therefore these associations need to be additionally explored with alternative measures. The sleep measure was parent-reported and parents may vary in their detection of sleep disorders in their children, and some sleep disorders may be more detectable by parents than others (e.g. sleep-walking versus insomnia).
In conclusion, this study indicates that sleep problems in late childhood/early adolescence are common, and are associated with increased likelihood of reporting psychotic experiences both at the time and at one year follow up. Sleep problems were also strongly associated with persistence of psychotic experiences from baseline to one year – this is of particular importance given that persistence of psychotic experiences in this period is associated with raised risk of later psychotic disorder. These associations remained strong even while controlling for a broad range of confounding variables, including ethnicity, neighbourhood deprivation, family conflict, and prescription of stimulant medication. In conclusion, this study suggests further attention towards sleep problems in this period both as a plausible contributing factor to later mental health problems, but also as a possible preventative treatment target.
Data Availability
The data are available from application the NIH ABCD study committee. Our pre-registered analysis plan and analysis script are available at https://osf.io/8ks72/
Acknowledgements
The authors would like to thank the investigators and participants of the ABCD study for the provision of the data used in the analysis.