This long-term follow-up of a staging process applied to a large group of help-seeking youths confirmed our hypothesis of an incrementally increased risk of later psychiatric morbidity for each of the three stages compared to the background population. Help-seeking youths were more likely to later use psychiatric services compared to the background population. There was a highly significant trend of increased risk of later psychiatric morbidity with increased stage level. The results were remarkably consistent across a broad range of separate, clinically significant primary and secondary outcomes, based on objective measures drawn from national administrative registries.

A major strength of this study is a thorough and comprehensive questionnaire and interview-based assessment of a large population of help-seeking youths who were stratified into three stages of developmental psychopathology based on this data collection and followed for more than three years by the linkage to individual-level data drawn from several Danish administrative registries, enabling monitoring a wide range of different aspects of social functioning and mental illness. Furthermore, close to complete follow-up of participants was possible due to the use of register-based outcomes, which limited bias due to attrition [27]. Another strength was participation by self-referral, meaning that the study population reasonably represents the help-seeking population in the Danish municipalities. The setup, which involved data originating from four different municipalities in three regions of Denmark increased the generalizability of our findings to the help-seeking population in Denmark and possibly similar Western societies. The ecological validity was high, since the study was conducted in ‘real-world’, primary care-settings. Furthermore, we were able to rule out the impact of participating in the MMM intervention among participants in Stage 2 in our analyses.

Several limitations should be considered when interpreting our findings. First, some relevant information was unavailable in the registries [27]. Youths who were diagnosed in the private sector but without prescribed medications would not be registered. We assume that these cases would be mostly reflected among milder cases, since they were not registered elsewhere. Although this information would have been relevant, it would likely not have diminished our findings. Second, the quality of data in some registers, e.g. regarding school absence, was not optimal for this purpose. School absence was registered as absence instead of presence, making the interpretation of missing values difficult. We examined the differences in missing values between groups and found no relevant differences. Although a more complete and consistent registration of school absence would have been preferable, the available data appeared to have similar limitations across the staged groups. Third, due to the sample size and the rarity of some mental disorders, grouping of specific diagnoses was required. Fourth, the outcomes regarding diagnoses and treatment only identified the youths who sought and received service provision, missing youths with diagnosable mental health problems who did not seek help. Ultimately, although self-referral can be regarded as a strength, visitation relies on the help-seeking behavior of the families and e.g., youths from immigrant families were less likely to seek help compared to Danish families. Fifth, regarding representativeness of the cohort, the matched background population from the four municipalities had a lower proportion of immigrants (10%) compared to the entire country (around 15%) [28], indicating that the municipalities were not fully representative of immigrants.

Finally, one should also consider that the occurrence of diagnoses and the use of psychotropic medicine is not necessarily an adverse outcome. Youths suffering from mental disorders will likely benefit from being diagnosed in CAMHS, since this often facilitates the possibility of getting help, either in CAMHS, in school, or psychosocial support for the family. These primary outcomes were dependent on the caregivers’ help-seeking behavior and the availability of resources to navigate a complex service system. Hence, we wanted to investigate outcomes that were not dependent on help-seeking behavior or family resources to the same extent. Therefore, in addition to the data from the health registers, we retrieved data on school absence (reported from the schools) and notifications of concern (reported from the municipalities) and the reduction in parents’ labor market attachment (from the Ministry of Employment’s database), to get a more nuanced depiction by including additional objective outcomes.

While we are not aware of studies utilizing a similar approach to stratification, some studies have examined other stratified care models [29, 30]. The first study [29] demonstrated that stratified care, where patients were allocated to either low- or high-intensity treatments at the initial assessment, corresponding to our stage-based approach, was cost-beneficial compared to stepped-care (in this case, treatment as usual) in a cluster-randomized trial among adults with depression. The second study [30] found that stage 1b level symptoms (corresponding to our stage level 2, in need for indicated prevention) of depression, anxiety and psychosis were interrelated and associated with greater emotional and behavioral difficulties in early adolescence, as well as with life events in late adolescence.

We found that all three stratified groups had increased adverse outcomes compared with the background population. Although youths in Stage 1 presented with only mild symptoms, their use of psychiatric services was markedly higher than that of the background population. It could be speculated that proxies for healthcare use alone may partly reflect help-seeking behavior among caregivers - sometimes referred to as “the worried well” [31]. However, the youths in Stage 1 were also more than twice as likely to have been the subject of a municipal notification of concern as those in the background population during the three-year follow-up. Since this outcome is often independent of parental help-seeking, the “worried well” interpretation is challenged. This suggests an alternative interpretation of this group, e.g. “the prodromal” [31], where the social network has identified a problem not yet detected by the visitation algorithm. This may indicate that our threshold for indicated prevention was set too high.

As for Stage 2, comprising youths with moderate symptoms, we aimed to identify those in need of early, moderate-intensity intervention as indicated prevention. However, as reported previously, 80% of youths at Stage 2 met research-based criteria for mental disorders at baseline [8], and one fourth received a clinical psychiatric diagnosis in CAMHS within three years, based on our results. The most ill individuals in Stage 2, who appeared to present with mental disorders already at baseline, received the full treatment benefits of the MMM intervention [32, 33], however, with no maintenance of effects after three years [34], perhaps calling for a combination of psychotherapy and long-term monitoring with easy access (fast-track) to CAMHS when needed.

For Stage 3, which comprised youths with severe symptoms and a need for specialized services, only half received treatment in CAMHS. There may be multiple reasons for this gap between needs and actual help, e.g. (a) the families had not reached the point where they sought professional help to be referred to CAMHS, (b) did not have resources to comply with appointments, or (c) bureaucratic processes or referral rejection from CAMHS.

We did not observe any parental reduction in labor-market attachment during the follow-up period; however, baseline results showed that only 60% of mothers and 77% of fathers of children in Stage 3 were employed. This may suggest a possible floor effect, where already burdened families had limited labor-market attachment at baseline, making further net reductions during follow-up difficult to capture.

Mental health service providers are experiencing a growing number of help-seeking families. Consequently, the system is becoming increasingly overwhelmed, to the extent that youths in need of effective therapeutic interventions often must wait until their condition deteriorates before receiving the treatment that might have prevented such deterioration. This situation leaves the ‘missing middle’ [9], corresponding to youths at Stage 2, at risk of growing in number while the condition deteriorates.

The stage-based and measurement-based visitation model developed for the MMM trial and evaluated in the current study could, alongside an easily accessible and effective therapeutic intervention implemented within the municipalities, serve as a useful tool in addressing this problem. This approach could be further strengthened by integrating knowledge about which subgroups of help-seekers benefit most from specific forms of indicated prevention.

Our hypothesis that an incrementally increased risk of later psychiatric morbidity for each of the three stages was confirmed: the stage steps at baseline cast shadows into the future. The knowledge that we can stratify the help-seeking population according to their subsequent need for service use supports the measurement- and stage-based, stepped care model. However, the relatively high risk of adverse outcome for the help-seeking youths at all levels and the finding that one third of the total group of help-seekers eventually received psychiatric care within three years highlights the need for qualified care at the first levels and easy access to care at the next level – without delay. Thus, early visitation might open an essential window of opportunity to change adverse trajectories for youths with mental health problems.