Outcomes of formal diagnostic labelling in young people with depressive symptoms: a longitudinal investigation

The role of diagnostic labels in mental healthcare has long been contested, with clinicians and service-users expressing concerns over both underdiagnosis and overdiagnosis (Callard et al., 2013; Frances, 2016). The evidence base regarding the consequences of receiving a psychiatric diagnosis demonstrates mixed effects (O’Connor et al., 2018; Perkins et al., 2018; Sims et al., 2021). Routes through which formal diagnosis can improve wellbeing include access to specialist treatment, communities of similar others, self-understanding and social explanation of atypical behaviour. Crucially, for conditions such as depression, a formal diagnosis is often a prerequisite for accessing evidence-based interventions. However, labelling theory suggests diagnosis can also exacerbate mental ill-health by establishing self-fulfilling expectations and social disadvantage (Link et al., 1989). In the context of youth depression, empirical data confirm diagnostic labelling poses risks of lowered self-esteem, fatalism about future prospects, stigmatisation and eliciting cognitive or behavioural patterns that exacerbate mental ill-health (Ahuvia et al., 2024; O’Connor et al., 2018). The implications of diagnostic labelling are context-dependent, varying according to the precise diagnosis and its social and personal connotations. For example, while disclosure of Autistic Spectrum Disorder diagnoses tends to reduce stigmatising responses, disclosure of depression diagnosis may exacerbate the stigma already elicited by depressive-typical behaviours (O’Connor et al., 2021). These dynamics are further complicated by broader societal shifts and temporal fluctuations. In particular, how the specific context of the COVID-19 pandemic intersected with the processes and consequences of diagnostic labelling remains uncertain.

Amidst these shifting contexts, there has been a documented increase in depressive symptoms among young people over recent years, alongside a concurrent rise in diagnostic rates (Dykxhoorn et al., 2025; Newlove-Delgado et al., 2023; Pitchforth et al., 2019). Notably, however, depression diagnoses in youth appear to be rising at a rate that outpaces the actual deterioration in emotional wellbeing (Dykxhoorn et al., 2025; Pitchforth et al., 2019). However, this trend is met with debate; while some express concern regarding over-medicalisation, others argue that significant numbers of youth with depressive disorders may still be underdiagnosed (Callard et al., 2013; Frances, 2016).

It is crucial to understand the broader implications of depression labelling on youth trajectories. Minimal research examines the long-term consequences of receiving a depression diagnosis as a young person. The current analysis leverages data from a large-scale household survey to gain insight into the longitudinal implications of diagnostic labelling of depression on youth wellbeing. The guiding question is: Do young people in the UK with a diagnosis of depression show differing trajectories of wellbeing over time, compared with young people with depressive symptoms who do not receive a diagnosis?

Data came from Understanding Society, the UK Household Longitudinal Survey (UKHLS) (University of Essex, Institute for Social and Economic Research, 2024). UKHLS data is collected using a mixed-mode approach (face-to-face, telephone, web and paper) from individuals of all ages from over 40,000 UK households. The study is sampled through a representative stratified sample of the UK population, with oversampling of underrepresented groups. The current analysis used Waves 10–14 (collected yearly between 2018–2020 and 2022–2024), due to the absence of relevant variables (e.g. psychotherapy usage) from earlier waves.

Participants were excluded from the analysis if they were not aged between 16 and 25 years at Wave 10 (baseline), and had not either ever received a formal diagnosis of depression or met the threshold for depressive symptoms (described in detail below) at Wave 10. Participants were also excluded if they had ever received another psychiatric diagnosis other than depression (n = 199). Sixty of these self-reported having previously received a formal diagnosis of depression before Wave 10, while 1,759 had not received a formal diagnosis but had depressive symptoms.

Baseline mental health: The SF-12 Mental Component Summary (MCS; Ware et al., 1996, 2002) at Wave 10 was used to establish baseline symptom severity. The MCS represents overall mental health functioning (e.g. “how much of the time during the past four weeks have you felt downhearted and blue?”) with potential total scores ranging from 0 = low functioning to 100 = high functioning.

The MCS was also used to establish whether participants reported substantial depressive symptoms to be considered for inclusion in the analysis. A total score of ≤ 45.6 on the MCS indicates participants meet probable diagnostic criteria for depression (Vilagut et al., 2013). However, as this is not a formal clinical assessment, we define participants meeting this threshold as having severity of depressive symptoms in line with a potential diagnosis of depression.

Diagnosis status (0 = no diagnosis, 1 = diagnosis): Participants in the UKHLS are asked at each wave to self-report whether a healthcare professional ever told them they had depression. Participants who responded yes at or before wave 10 were coded as having a formal diagnosis of depression. Participants who did not respond yes were coded as not having a formal diagnosis of depression.

Psychological distress: wellbeing outcomes were assessed at each of the five waves using the 12-item General Health Questionnaire (GHQ; Goldberg, 1988; Goldberg et al., 1997), (e.g. “Have you recently felt constantly under strain?”), with potential scores ranging from 0 = the least distressed to 36 = the most distressed.

Several characteristics were considered for use as covariates in the initial model. These were chosen based on existing literature, including previous analyses using UKHLS that indicated associations with GHQ and MCS (Parra-Mujica et al., 2023). These included: age, sex, ethnicity (White, Asian/Asian British, Mixed/Other, Black/Black British), equivalised household income, living situation (lives as couple/with parents/alone/other), level of education (lower secondary/upper secondary/bachelor/masters/none), baseline mental health functioning (MCS) and psychotherapy engagement (yes/no). In wave 10, participants indicated whether they had made use of psychotherapist services (including psychiatrist or analyst) in the past 12 months. As this information was collected only at waves 10 and 12, we were unable to account for prior engagement in psychotherapy in the analysis. No information pertaining to pharmacological intervention was available in these waves.

To establish whether there were significant differences in the likelihood of receiving a formal diagnosis of depression across sociodemographic and other relevant characteristics, a series of chi-square and t-test analyses were conducted.

To address the primary research question, latent growth curve modelling was used to examine changes in GHQ scores over five waves of data, and to test whether these trajectories differed by depression diagnosis status. GHQ was modelled with three latent factors:

1. an intercept (baseline GHQ);

2. a linear slope (rate of change of GHQ over time); and

3. a quadratic term (curvature of change of GHQ).

Depression diagnosis status was included as a predictor variable. The initial model also included a range of covariates measured at baseline: psychotherapy engagement, baseline mental health functioning (MCS), age, sex, ethnicity, equivalised household income and level of education. Due to only two individuals receiving a depression diagnosis in later waves (waves 11–14), this factor was not accounted for in the analysis.

To improve model parsimony and maximise statistical power given the small proportion of participants with a diagnosis, nonsignificant covariates with no strong theoretical justification for retention were removed in a stepwise manner. Psychotherapy engagement was considered due to theoretical relevance, but excluded due to non-significance in the initial model and the low number of participants who both received a diagnosis and were engaged in psychotherapy (n = 11).

The final model retained sex, baseline mental health functioning and depression diagnosis status. Fit indices for the full and final models were compared to ensure that trimming covariates did not meaningfully reduce model fit. The main effects of diagnosis status were consistent across both models.

The estimator used was maximum likelihood with robust standard errors. Fitting adequacy was established by the following: Tucker–Lewis Index (TLI) and Comparative Fit Index (CFI) ≥ 0.95 indicating excellent fit; root mean square error of approximation (RMSEA) and standardised root mean square residual (SRMR) ≤ 0.05 indicating excellent fit. The analysis was conducted using the lavaan package in RStudio v2023.6.1.524 (Posit team, 2025; Reeve et al., 2018).

Participant characteristics are summarised in Table 1.

There were no significant differences in diagnosis status across sex $χ21, N=2,018=0.2, p=0.904$⁠, age, $t2,016=0.5, p=0.619$⁠, education level, $χ21, N=2,018=7.8, p=0.101$⁠, living situation $χ23, N=2,018=0.94, p=0.815,$ equivalised household income $2,016=2.25, p=0.134$or baseline mental health functioning, $t2,016=1.8, p=0.074.$ However, participants with a diagnosis were more likely to engage in psychotherapy (n = 11, 18.3%) than participants without a diagnosis (n = 74, 4.2%), $χ21=22.586, p<0.001.$ There were significant differences across ethnicity, $χ23=11.285,p=0.020$⁠. Small sample sizes limit interpretability: however, results indicate that White participants were more likely to receive a diagnosis, while those of Mixed/Other ethnicity were less likely.

Eleven participants were removed from the analysis due to missing exogenous variables, leaving a final sample of n = 1808, of which 58 had a depression diagnosis. For GHQ, rates of missingness ranged from 2% at Wave 10 (baseline) to 45% at Wave 14 due to participant attrition over time, with a marginally non-significant difference in attrition across diagnosis status $χ21=3.64, p=0.057$⁠. Full information maximum likelihood estimation was used to handle missing data.

The final latent growth curve model included sex, baseline mental health functioning and diagnosis status, and demonstrated excellent fit to the data: CFI = 0.997, TLI = 0.995, RMSEA = 0.015, SRMR = 0.014. Growth parameters for this model (intercept = 28.65; slope = –7.67; quadratic = 1.29) indicated that there was an overall decrease in psychological distress over time, with the rate of improvement slowing in later waves.

Sex significantly predicted baseline psychological distress $(β=0.50, SE=0.23, p=0.028)$⁠, indicating that females demonstrated higher initial psychological distress than males. However, sex did not significantly predict the linear slope $(β=0.33, SE=0.33, p=0.313)$ or quadratic term $(β=-0.08, SE=0.08, p=0.376)$⁠, indicating that the rate of change or curvature of psychological distress over time did not significantly differ by sex.

While controlling for baseline mental health functioning and sex, diagnosis status significantly predicted the intercept $(β=1.28, SE=0.63, p=0.040)$ and slope $β=2.64, SE=0.83, p=0.001$⁠, indicating that participants without a diagnosis demonstrated lower psychological distress at baseline, and experienced slower decreases in psychological distress over time than participants with a diagnosis. In addition, diagnosis status significantly negatively predicted the quadratic term $β=-0.61, SE=0.21, p=0.003$⁠, indicating that this trajectory changed in later waves. Inspection of the adjusted differential wellbeing trajectories across diagnosis status highlighted that while individuals who previously received a diagnosis experienced an initial increase in psychological distress, this trajectory did not remain consistent throughout each wave, with distress decreasing in later waves. Figure 1 presents the adjusted differential wellbeing trajectories across diagnosis status, and highlights that while individuals with a diagnosis experienced an initial increase in psychological distress, this trajectory did not remain consistent throughout each wave, with distress decreasing in later waves. Table 2 presents the parameter estimates for the final model.

The analysis examined whether young people with a formal diagnosis of depression show distinct mental health trajectories over time, compared with those without a diagnosis who had depressive symptoms.

Significant differences were found in the trajectory of psychological distress between groups, while accounting for initial psychological distress. Individuals without a diagnosis experienced less rapid changes in psychological distress over time, demonstrating a gradual improvement which slowed in later waves. However, individuals with a diagnosis experienced an initial increase in psychological distress, with this trajectory evolving in later waves to a decrease in psychological distress.

Existing literature has identified mixed effects associated with diagnostic labelling (O’Connor et al., 2018; Perkins et al., 2018; Sims et al., 2021). The current findings expand this complexity to suggest that the impact of experiencing depression labelling on general wellbeing may change over time. For instance, individuals with a diagnosis may experience initial adverse effects, such as those related to self-concept and stigma (Ahuvia et al., 2024; O’Connor et al., 2018), which fall off over time.

Notably, the findings should be interpreted within the context of the COVID-19 pandemic. The Wave 10 data collection period ended in May 2020; Waves 11 and 12 were collected during the pandemic (2019–2022); and Waves 13 and 14 were collected after (2021–2024). This could indicate that individuals with a diagnosis were more negatively impacted by the pandemic, whereas those without a diagnosis demonstrated gradual improvements during the pandemic, with continual later improvements. A potential explanation is that the pandemic may have limited the usual advantages of receiving a diagnosis, such as enhanced treatment access and increased leniency from others (O’Connor et al., 2018), while the adverse effects, such as self-stigmatisation (Sims et al., 2021), remained salient. Further research is required to clarify.

The use of latent growth curve modelling allowed for the capture of non-linear trajectories of wellbeing across five waves while accounting for baseline symptom severity, ensuring observed trajectory differences were not driven by initial differences in distress. While the analysis met adequate statistical power for a latent growth model with diagnosis status as a predictor (>50 per group), the low available sample size of young people with a diagnosis (n = 58) excluded the possibility of conducting multiple-group growth models (>100 per group) (Curran et al., 2010). This limited the ability to compare the shape of the trajectories, and meant that the current analysis assumed equal residual variances, factor loadings and measurement error across both groups. Another limitation is the use of self-report measures for establishing whether participants had depressive symptoms, or were told by a healthcare professional that they had depression. While established MCS diagnostic cut-offs were used to establish whether participants had depressive symptoms (Vilagut et al., 2013), we cannot determine with full certainty whether participants would have met diagnostic criteria had they undergone a formal assessment. As mental health medication information was unavailable, we were unable to account for receipt of pharmacological intervention in the analysis. Finally, because the study is observational, causality cannot be inferred and further research is needed to determine whether diagnosis status causally affects wellbeing over time.

The findings warrant clinicians to be mindful of the potential adverse implications associated with receiving a depression diagnosis. Mental health services should be prepared to provide immediate psychosocial support during the diagnostic period to mitigate these initial risks.

Future research is required to pinpoint the mechanisms underlying the differences in wellbeing across those with and without a depression diagnosis. Additional longitudinal analyses are also required to establish whether the observed trajectories remain consistent across various populations, cultures or diagnostic labels.

The study highlights a complex relationship between diagnostic labelling and wellbeing trajectories in young people. While a diagnosis of depression is associated with an initial increase in psychological distress, these effects appear to shift over time, potentially influenced by broader societal disruptions like the COVID-19 pandemic. As diagnosis rates continue to rise, understanding the temporal nuances of how a depression label shapes a young person’s wellbeing is essential for effective intervention.

Leigh Huggard is based at School of Psychology, University College Dublin, Dublin, Ireland, and the School of Public Health, University College Cork, Cork, Ireland.

Cara Booker is based at Institute for Social and Economic Research, University of Essex, Colchester, UK.

Cliodhna O’Connor is based at School of Psychology, University College Dublin, Dublin, Ireland.