Having psychotic experiences or depression at 12 years of age was associated with adverse life outcomes in early adulthood. Both were associated with poorer educational performance at age 16, and affected individuals were about twice as likely to be NEET at age 20. Social functioning was also impaired, but the association with psychotic experiences was primarily driven by comorbid depressive symptoms, which were either confounding the association or lying on the causal pathway between psychotic experiences and social dysfunction. These competing explanations of psychotic experience-associated social impairment have been raised before [29], but it was not possible to tease them out here. In contrast, the relationship with poor GCSE performance was relatively independent of comorbid symptoms, suggesting that its association with psychotic experiences and depression was mediated differently to that of social functioning. For example, both psychotic experiences and depression could have independent effects on concentration that impair academic performance, whereas the key impact on social functioning might be through a characteristic that is more central to depression, such as low self-esteem.

The evidence of associations at age 12 was generally stronger for the depression score than the binary depression threshold measure. This was as expected given the former offers a more powerful approach, and is consistent with effects being present across the spectrum of depressive symptomatology. In comparison, this was much less evident for the psychotic experience score, perhaps suggesting that the presence of any psychotic experience indicates psychopathology that is sufficiently severe to explain associations with adverse outcomes. Thus, the psychotic experience score might not be assessing the spectrum of sub-threshold effects as the SMFQ does for depression.

The relationship between depressive symptoms at age 12 and educational performance was not straightforward. Unlike the binary variable, the depression score was not associated with worse educational attainment at age 16. Indeed, there was evidence of a weakly positive effect on educational attainment at age 18. This might have been a chance finding, but it could be that the SMFQ was measuring something more than depressive symptoms. Perhaps, it also detected personality traits like neuroticism and introversion that, in moderation, could be associated with better educational performance. However, a positive association was not replicated for depression score at age 18, which potentially undermines this explanation. In fact, depression score at age 18 was associated with worse education outcomes at this age. It could be that the SMFQ measures depression to a greater extent than neuroticism at age 18, but to a lesser extent at age 12; the higher SMFQ mean score and the proportion meeting criteria for depression at this age would be consistent with this account.

By age 18, psychotic experiences were less prevalent than depression, a reversal of the picture seen at age 12. There was strong evidence of associations between psychotic experiences and depression with poorer educational outcomes, harmful drinking, problem cannabis use, smoking, and use of other drugs. Psychotic experiences and depression were additionally associated with an increased likelihood of illegal behaviour, although the association with depression seems to have been largely attributable to comorbid psychotic experiences. The reverse was true for friend dissatisfaction, which was underpinned by comorbid depressive symptoms rather than psychotic experiences.

Aside from friend dissatisfaction, there was little to suggest that comorbid psychotic experiences and depression at age 12 were associated with poorer outcomes than with having either condition alone. In contrast, comorbidity had significantly more negative impact at age 18, particularly when compared with the educational and substance use outcomes of only having depression, supporting the thesis that psychotic experiences are an index of mental disorder severity [32].

Taken collectively, these findings suggest that adolescent psychotic experiences are not as benign as those limited to childhood [27, 28]. Even discounting the minority who go on to develop mental health problems [7, 12–14], the lives of many individuals may be negatively affected in other ways for a considerable period of time. The association of psychotic experiences, both at ages 12 and 18, with poorer educational and occupational outcomes is consistent with findings in other longitudinal studies [12, 26]. This is also true of the association with social impairment, although this seems to have been driven by comorbid depressive symptoms, as has been previously found in the ALSPAC cohort [29]. Other adverse outcomes, particularly those involving antisocial behaviour (as indexed by our measure of illegal and offending behaviour), drug use and alcohol abuse, were only associated with psychotic experiences at age 18. These are cross-sectional analyses, so it is not possible to comment on the direction of causality, but they are consistent with associations described elsewhere [23–26]. These additional adverse associations may also be a function, at least in part, of the time period being described: questions about psychotic experiences at age 12 covered the preceding 6 months compared with 6 years at age 18. Thus, the latter may have captured more chronic presentations.

The results for adolescent depression were broadly consistent with associations previously reported with impaired social functioning [9, 16] and negative educational and employment outcomes [10, 11, 16, 17]. Alcohol abuse, which has been one of the most consistent associations in the past research [11, 18, 19], was the substance-related issue most strongly associated with depression at age 12. Earlier findings that adolescent depression is associated with smoking, illicit drug use, and delinquent behaviour [11, 16, 19] were much more evident in the associations with depression at age 18. Adjusting for confounders did not unduly diminish the associations found here, unlike in the Christchurch cohort where associations between adolescent depression and adverse education and employment outcomes, nicotine dependence and alcohol abuse were eliminated following adjustment [10]. While we adjusted for many of the same covariates (e.g., maternal education, social class, and IQ), we were not able to adjust for others such as childhood sexual abuse and affiliating with substance-abusing peers. Thus, residual confounding could exist within the present study.

The current study has a number of strengths including: a longitudinal design with reasonably long follow-up, access to a range of psychosocial measures, including some that are interviewer-rated (e.g., the PLIKSi), and use of objective national exam data sets. It also has its limitations. Despite adjusting for a reasonably comprehensive set of confounders, residual confounding remains possible. Furthermore, as most of the associations with psychotic experiences and depression at age 18 are cross section, it is possible that psychopathology resulted as a consequence of the poorer psychosocial outcomes. Despite the large sample size, the relatively rare nature of the exposures and outcomes often meant that the analyses were likely to be underpowered, and the confidence intervals were often wide. There was also considerable attrition in the ALSPAC cohort that could have introduced selection bias. This has been explored elsewhere [34], but simulation studies in ALSPAC have shown that while prevalence estimates are under-estimated, associations are only marginally affected by selective attrition [45]. Nevertheless, it is possible that estimates we report are influenced by attrition bias. A further limitation was that it was not possible to infer whether an individual lacking GCSE or A-level data were the result of him or her failing these or not having sat the exam, because data from private schools were not included in the data sets. This meant that students not progressing onto these exams, perhaps because of their psychotic experiences or depression, would not have been included in our analyses, potentially underestimating associations with adverse educational outcomes.

In light of these findings, it is important that adolescent psychotic experiences and depressive symptoms are identified as early as possible because they represent a risk marker for a number of adverse outcomes in later life, most consistently with education and employment. Current UK guidelines for managing adolescent psychosis recommend additional educational support when performance has been affected and advocate the provision of supported employment programmes and work-related activities [46]. The depression guidelines are less explicit, but do identify the need to address educational problems [47]. Meanwhile, the strong links with alcohol, cannabis, and other drug use for older teenagers reporting psychotic experiences and depression, while not necessarily causal, highlight the need to provide psychoeducation and support. Future research needs to build on the limited evidence, we currently have about long-term functional outcomes. One way would be to construct more sophisticated models to test whether variables like drug and alcohol misuse could be mediating outcomes like unemployment.