Main findings
Early adolescent psychotic experiences and depression were both associated with mid adolescent suicidal behaviour. However, once mutually adjusted for the other psychopathology and other confounders, the association between psychotic experiences and suicidal behaviour was attenuated by approximately 36 %, whereas the association between depression and suicidal behaviour was attenuated by approximately 15 %. A similar pattern was observed for cross-sectional associations between depression and psychotic experiences and suicidal behaviour in mid-adolescence, although evidence for an association between psychotic experiences and suicidal behaviour was stronger than in the longitudinal analysis. The cross-sectional association between psychotic experiences and suicidal behaviour that we observed was weaker than that found in previous cross-sectional studies with clinical participants [
9,
10], though similar to that reported in two other general population studies [
4,
6]. This is not unexpected given that psychotic experiences in individuals from a general population sample and psychotic disorder in individuals from a clinical sample are likely to lie at different locations on the psychosis continuum. Alternatively, the association detected in studies with clinical participants could be an example of Berkson’s bias. Attendance at a clinic may be related both to psychotic experiences and suicidal behaviour. A non-causal association may therefore have been erroneously introduced [
21]. Our effect size was substantially smaller than those reported in one cross-sectional population-based study, and in a previous longitudinal study, both of which reported approximately 10-fold increases in odds of suicidal behaviour in those with psychotic experiences compared to those without [
8,
11]. Even at the upper-most limit of our confidence intervals, our results were not compatible with a more than four-fold increase in odds of suicidal behaviour in those with psychotic experiences. In comparison, the effect sizes we observed between depression and suicidal behaviour were large, with an approximately eight-fold increase in suicidal behaviour in those with depression compared to those without depression at age 16, and a four-fold increase in the longitudinal analysis.
There are several possible reasons for the large difference in the strength of the association detected between psychotic experiences and suicidal behaviour in our study and in the studies above, including the use of different instruments for measuring psychotic experiences and suicidal behaviour, different methods of data collection (i.e. self-report and semi-structured interviews), different informants (parent and adolescent), confounding and a shorter follow-up period in the case of the longitudinal study [
11]. For instance one of the studies [
8] used extremely detailed interviews to measure psychotic experiences, suicidal behaviour and depressive symptoms, whereas in our study self-report measures were used to collect information on all of these variables at 16 years. Additionally depression was measured as both lifetime and over the previous month, whereas in our study questions about depressive symptoms just covered the previous 2 weeks. It is possible that these more detailed measures increased the sensitivity in this study [
8] and therefore stronger associations were found. Also in the longitudinal study Kelleher study [
11] suicidal behaviour was assessed using much more detailed questions than in our study which also may have increased sensitivity and therefore effect sizes. A further possibility is that the effect size is over-estimated in the two studies which reported the strongest associations; it is recognised that studies with small samples and low statistical power are more likely to suffer from effect inflation [
22].
Furthermore, whilst the effect of psychotic experiences on suicidal behaviour was described as being strongest within a subgroup of those with depression and with suicidal thoughts in one study [
8], we found no evidence in the prospective analysis that the association between psychotic experiences and suicidal behaviour differed in those who were also depressed compared with those who were not or in those who had suicidal thoughts compared to those who did not.
The PLRs and the ROC tests also indicated that depressive symptoms were a substantially better predictor of future suicidal behaviour than psychotic experiences. Furthermore, whilst depressive symptoms at 12 had a probability of correctly predicting 64 % of future suicidal behaviour, adding information on psychotic experiences hardly altered this, whereas whilst psychotic experiences at 12 had a probability of correctly predicting 56 % of future suicidal behaviour, adding depressive symptoms substantially increased the probability of prediction.
Implications
If we had replicated the very strong associations found in previous studies [
23,
24] this would have suggested that psychotic experiences in early adolescence are a very strong indicator of suicidal behaviour both at the time and in later adolescence. However our results call into question the utility of psychotic experiences as a predictor of later suicidal behaviour. Our findings suggest that screening for psychotic experiences at 12 years of age tell us little about suicide risk at age 16; information on gender was more predictive of suicidal behaviour, whilst a measure of depressive symptoms at age 12 was the most useful for prediction of suicidal risk. Whilst clinicians should investigate whether those with psychotic experiences also have suicidal thoughts, assessing depressive symptoms is of more importance when assessing suicidal risk. The area under the curve (ROC) analysis shows that adding information about psychotic experiences to that of depressive symptoms infers little added benefit when addressing future suicidal behaviour risk whereas adding information on depressive symptoms is highly informative when added to information on psychotic experiences.
The association observed between psychotic experiences and suicidal behaviour in other studies is likely to be, at least partly, due to co-morbid depressive symptoms since there is convincing evidence of the strong overlap between psychotic experiences and depressive symptoms [
25,
26] and our results showed the association between psychotic experiences and suicidal behaviour attenuated substantially after adjusting for depressive symptoms. Suicidal behaviour may also result from particular psychotic experiences such as command hallucinations or delusional thought processes that drive forward this behaviour, although this explanation seems less likely in non-clinical populations.
Whilst psychotic experiences in early adolescence seem to have limited value for predicting suicidal behaviour in mid-adolescence suicidal behaviour does index likelihood of having psychotic experiences to some extent. In our dataset 27 % of those who had attempted suicide also had psychotic-like experiences at that time compared to 6 % of those without suicidal behaviour.
Strengths and limitations
The prospective design of our study is a strength which has enabled us to determine the utility of psychotic experiences at predicting suicidal behaviour over time in a large general population sample of adolescents that is broadly representative of the UK adolescent population. Furthermore this is one of the largest samples available to address this question with longitudinal data in such a well characterised cohort.
Selection bias due to participant attrition is a limitation of our study since data were missing for a substantial proportion of the cohort. Those not included in the analysis were more likely to have had psychotic experiences and suicidal behaviour indicating that the effect sizes that we detected might have been larger in the absence of participant attrition.
A previous study has shown that attrition within ALSPAC tends to have little effect on estimates of association [
27] and it seems unlikely that attrition has led to substantial distortion of the associations between psychotic experiences, depression and suicidal behaviour in this cohort.
Since the questions about suicidal behaviour asked about lifetime experience it is possible that in the cross-sectional analysis suicidal behaviour and psychotic experiences may not have been concurrent, although others [
8] have found that positive responses to questions about psychotic experiences in adolescence nearly always refer to very recent experiences.
Only psychotic experiences which were not attributed to the effects of alcohol or drug use, falling asleep/waking up or the effects of fever are used in this analysis and therefore the findings are not likely to be confounded by these factors.