We conducted an evaluation of the IDEA-RS using a new sample intentionally recruited for prospective assessment. Leveraging a predictive framework based entirely on sociodemographic variables, the model effectively stratified the risk of developing depression in adolescence. Notably, over a 3-year period, we observed that one out of three HR adolescents, as identified by the IDEA-RS, developed depression compared with one out of eight among those stratified as LR. This meant a likelihood more than 2.5-fold greater of depression onset in HR individuals, supporting the predictive validity of the IDEA-RS, highlighting its critical role in the early identification of at-risk adolescents.

Since the IDEA-RS is anchored on sociodemographic variables that can be obtained directly from adolescents, it provides a practical and highly applicable approach to risk stratification across a variety of contexts, bypassing the need for specialised training. Particularly significant in low-resource settings—where most of the global youth population lives20—the IDEA-RS emerges as a promising tool for preventive initiatives targeting the individuals who are most susceptible to the development of depressive disorders in adolescence.4

In our study, we used sociodemographic information to stratify adolescents for the risk of depression, diverging from conventional methods that frequently rely on subthreshold symptomatology to predict full-blown syndromes in psychiatry.21 Focusing on subthreshold symptoms often requires training and extensive assessments; also, implementation can be challenging, especially in population-wide contexts given the difficulty in drawing a clear line between high levels of subsyndromal symptoms and full-blown presentations.22 Moreover, considering that depressive symptoms may fluctuate over time, assessing them at a single point may merely capture transient manifestations of an already established condition, conceptually challenging the prediction of an existent condition.23

Our approach differs from studies using familial depression to assign HR status. While a positive family history has been consistently replicated as a risk factor and widely used as a variable for defining elevated risk—a recent meta-analysis documented a 2.3-fold increased risk for depression among offspring of a previously diagnosed parent24—our score simplifies data collection with items that can be directly informed by adolescents. This approach mitigates the need for caregiver involvement, often seen as an encumbrance. Furthermore, research deploying depression polygenic risk scores has indicated a 2.5-fold elevation between the highest and lowest deciles.25 Although genetic and family history-based approaches provide valuable information, they are not consistently available across settings due to accessibility barriers and the costs involved in laboratory analyses, as well as the need for family members to contact services to receive a diagnosis; these limitations may be felt particularly in low-resource settings.26 In contrast, the IDEA-RS showed a comparable predictive ability in estimating the risk for developing depression among youth using information that can be easily collected directly from the adolescent. The extent to which the IDEA-RS encapsulates latent risks beyond positive family history or genetic predisposition remains a question for future research.

In addition to a standalone risk stratification tool, the IDEA-RS represents a means to refine research methodologies: assigning individuals to a control group based on the absence of a current diagnosis can be particularly problematic for younger participants, as these approaches often fail to account for the various risk factors that could predispose those classified as non-cases for the future onset of the disorder.4 Adopting a nuanced approach can enhance our understanding of the mechanisms involved in depression pathophysiology, enriching clinical practice and research. The use of risk stratification tools such as the IDEA-RS can be extremely valuable in research for parsing the heterogeneous group of individuals without depression into those with a low or a high probability of developing the disorder, underscoring the importance of not lumping them together as a homogeneous group of non-cases.

A paramount strength of this study was the replication of the original IDEA-RS from the Pelotas cohort using all 11 variables from the initial development sample. Despite this precise replication, the IDEA-RS estimated a higher baseline probability for adolescents in Porto Alegre (5.3%) to develop depression within 3 years compared with Pelotas (3.4%).13 This variation likely reflects the higher prevalence of most IDEA-RS variables in Porto Alegre, except for school failure, which was more common in Pelotas. Such differences were to some extent expected, given the population-based sample in Pelotas and the school-based sample in Porto Alegre. Despite these contrasts, a network analysis revealed a similar pattern of associations among the IDEA-RS variables in both cohorts, suggesting no evidence of major differences in terms of connectivity or structure.13 This reinforces the stability of the IDEA-RS and the generalisability of our results and potential applicability across diverse populations, further bolstering its validity as a reliable predictive measure of depression risk.

Our results indicated that the conversion rate to depressive disorders exceeded the initial IDEA-RS projections for both risk groups (table 1), a phenomenon that warrants further investigation. Several factors may explain this; the IDEA-RiSCo study was conducted in Porto Alegre, a city over three times larger than Pelotas, where the original cohort was based. This discrepancy in urbanisation levels, previously associated with higher depression prevalence, suggests a potential influence on these findings.3 26 Additionally, the decade-long gap between data collection for the Pelotas cohort (2008) and the IDEA-RiSCo cohort (2018/19) suggests cohort effects and differential exposure to risk factors. While the Pelotas cohort belonged to the Millennial generation, the Porto Alegre cohort included Generation Z adolescents, a group facing rising depression prevalence.27 The temporal gap between the cohorts also likely exposed IDEA-RiSCo participants to contemporary risk factors such as increased digital technology use28 and the impact of the COVID-19 pandemic.29 These factors highlight the importance of considering environmental and generational dynamics when assessing depression risk in contemporary adolescent populations.

This study boasts several key strengths. Confirming the predictive performance of a predictive score such as the IDEA-RS in a prospective cohort through significant IRRs is a rare achievement in psychiatric research and highlights the generalisability of the model.5 7 Low, evenly distributed attrition rates ensured balanced longitudinal data. Bias was minimised by using gold-standard clinical interviews at both baseline and endpoint, conducted by clinicians blinded to participants’ risk group. Adjustments for baseline depressive symptoms further controlled for confounding factors, enhancing result robustness. Focusing on a narrow age range during a critical developmental period for the onset of depression allowed consistent examination of clinical outcomes.

Some limitations should also be noted. While rigorous participant selection and detailed clinical profiling are strengths of this study, the consequent limited sample size is a constraint. The sample was designed to investigate associations between risk status, depression symptoms and neurobiological features, which may reduce the statistical power to compare less common outcomes or parameters with higher variability between risk groups, potentially masking significant differences.13 Additionally, our stratified sample, focusing on extreme percentiles, limits the use of discriminative measures like the C-statistic, calibration and decision curve analysis. These analyses typically require a broader range of risk scores to accurately assess model performance and clinical utility across different decision thresholds.30 Consequently, applying these methods in our stratified sample, which excludes individuals with intermediate risk, would not provide meaningful insights and may lead to biased conclusions.30

Conversely, several factors may have influenced the model’s predictive performance using traditional discrimination metrics. First, the relatively small sample size likely limited our ability to fully explore the variability between risk groups. The fact that IDEA-RS predicted only conversion to depression in this sample contrasts with the associations with other diagnostic categories seen in the original IDEA-RS study.8 Second, while the IDEA-RS model relies on 11 sociodemographic variables validated in the Pelotas cohort, these variables may not capture all local factors impacting depression risk in Porto Alegre. These contextual differences might contribute to a higher-than-expected baseline risk, potentially affecting the model’s calibration and accuracy. Despite previous beyond-chance discrimination in samples from five continents,8–12 further prospective replications in other regions are warranted.

An empirical sociodemographic risk assessment tool for future depression demonstrated robust predictive ability in a new, dedicated prospective cohort study. The likelihood of developing depression over 3 years was more than 2.5 times greater among HR youth compared with LR youth. The IDEA-RS, easily implemented and proven effective in diverse settings, holds promise as a valuable tool for targeting preventive interventions to ultimately reduce the burden of depression globally.