What are the Social Determinants That May Affect Access to Clinical Psychology Training?

Technical Summary: Social Determinants Affecting Access to Clinical Psychology Training

Problem Statement
Social inequalities in the United Kingdom, aligned with Social Gradient Theory, create barriers to higher education and social mobility for individuals from marginalized communities. Despite the introduction of "contextual admission processes" by universities to evaluate applicants holistically—considering personal life experiences, challenges, and academic attainment alongside grades—there is a lack of empirical research evaluating the effectiveness of these processes in the specific context of Doctorate in Clinical Psychology (DClinPsy) training. Historically, the clinical psychology profession has lacked diversity, with evidence suggesting that racially minoritised and socially disadvantaged applicants face implicit and structural biases. While previous studies (e.g., Phillips et al., 2004) found insufficient evidence of selection bias, their analyses often excluded protected characteristics from regression models. This study addresses the gap in literature by investigating whether protected characteristics (age, gender, disability, ethnicity, religion, sexual orientation) and proxies for childhood disadvantage (first-generation university student, free school meals, Looked After Child status, Index of Multiple Deprivation) predict selection outcomes at the University of Hertfordshire (UH) DClinPsy course.

Methodology
The study employed a retrospective observational design using a dataset of 2,044 applicants to the UH DClinPsy course between 2021 and 2023. Data was sourced from the Leeds Clearing House, which collects standardized contextual admission questions.

• Variables: The study analyzed categorical variables (age, gender, disability, ethnicity, religion, sexuality, first-generation status, free school meals, Looked After Child status) and an interval variable (Index of Multiple Deprivation [IMD] at age 17).
• Data Handling: Missing data (ranging from 7.4% to 59.7% across variables) was treated as a distinct category ("Not Disclosed" or ND) rather than excluded, based on literature suggesting non-disclosure carries meaning. Age groups were consolidated, and rare religious categories were grouped.
• Statistical Analysis: Six binary logistic regression models (LRM) were run using SPSS v29 to predict selection outcomes. The outcomes were treated as ordered hierarchical categories, resulting in six binary dependent variables (e.g., Rejected vs. Waitlisted, Rejected vs. Accepted, Waitlisted vs. Accepted).
• Model Validation: Goodness-of-fit was assessed using the Hosmer & Lemeshow test. Collinearity diagnostics confirmed no significant multicollinearity (VIF < 5). A power analysis indicated the sample size was sufficient. Sensitivity analyses were conducted by excluding the "Not Disclosed" categories to test robustness.

Key Results
The analysis revealed that the majority of variables did not significantly predict selection outcomes:

• Non-Significant Findings: 74.07% (40 out of 54) of the entered variables showed no statistically significant influence on selection outcomes.
• Significant Predictors: Only 14 factors were statistically significant across the six models.
  • Ethnicity: Applicants of "Mixed" ethnicity were significantly more likely to be offered a place or waitlisted compared to White applicants (Odds Ratios: 0.27 and 0.21, respectively). "Other" ethnicities were more likely to be offered an interview than White applicants.
  • Religion: Affiliation with Judaism and Buddhism predicted a higher likelihood of being offered an interview compared to the "Not Disclosed" reference. Conversely, Sikhism predicted a higher likelihood of being waitlisted. Hinduism predicted a higher likelihood of being offered a reserve seat.
  • Age: Applicants aged 20–24 were significantly more likely to be rejected than those aged 25–29. Applicants aged 30–34 were more likely to be offered an interview than the 25–29 reference group.
  • Childhood Disadvantage: Higher IMD scores (indicating greater deprivation) at age 17 were associated with a slightly increased likelihood of being offered an interview.
  • Non-Disclosure: Not disclosing first-generation status or Looked After Child (LAC) status yielded mixed and unstable results with large confidence intervals, suggesting these variables are not robust predictors.

Key Contributions

• Empirical Evaluation of Contextual Admissions: This is the first quantitative study to empirically evaluate the selection outcomes of a DClinPsy course following the implementation of contextual admissions.
• Refutation of Hypothesized Barriers: The study challenges the hypothesis that protected characteristics and childhood disadvantages inherently predict unfavorable selection outcomes in this specific context. The finding that 74% of variables were non-predictive suggests that contextual admissions may be mitigating traditional barriers.
• Methodological Approach to Missing Data: The study contributes a methodological perspective by treating "Not Disclosed" responses as a distinct analytical category, arguing that non-disclosure may reflect stigma or systemic bias rather than random missingness.

Significance and Claims
The authors claim that the results support the value of contextual admission processes as a form of "positive action" that may mitigate barriers for minoritised applicants. The findings reject the study's a-priori hypotheses that marginalized identities predict unfavorable outcomes; instead, some minoritised groups (e.g., Mixed ethnicity, specific religious groups) showed a higher likelihood of favorable outcomes compared to reference groups.

However, the paper maintains a modest tone regarding these claims:

• Causality and Generalizability: The authors explicitly state that the findings are from a single-site study and should be interpreted as hypothesis-generating rather than definitive proof of causality. They caution that some significant findings (particularly regarding religion and ethnicity) may be coincidental due to the relatively small sub-samples and the lack of correction for multiple testing.
• Limitations: The study acknowledges limitations inherent in retrospective design, the use of secondary data, and the heterogeneity of the "Not Disclosed" category. The sensitivity analysis showed that several significant variables (including IMD and specific religious groups) lost significance when the sample was reduced, indicating instability in some findings.
• Conclusion: The paper concludes that while there is no compelling evidence of predictive factors preventing access at UH, further multi-site replication is required to confirm these findings and fully understand the efficacy of contextual admissions in the broader clinical psychology training landscape.