Observed (unadjusted) differences in outcomes between White-British individuals and people from different minoritized ethnic groups was reduced and, in some cases, disappeared entirely when accounting for socio-demographic and clinical factors. In whole-sample analyses in this study, people in the Black group were more likely to drop out, but not to experience worse treatment outcomes than White-British individuals when controlling for other factors. People in the Asian, ‘Other’ and White-other groups continued to experience poorer outcomes indicating that further efforts are required to enable people from these communities to benefit equally from talking therapies. Asian, Mixed and ‘Other’ groups also showed higher odds of disengagement. Existing research has identified poorer outcomes for people from Black and Asian communities [37], yet recent NHS reports suggest a decreasing disparity on some outcomes between White and other ethnic groups [8].

Access and outcome discrepancies experienced by Asian people are well-researched [38, 39]. The results from the current study might be understood in terms of identified challenges, such as awareness, cultural differences, stigma and social isolation [40] which might persist. Steps services might take to support at-risk groups could include consistent adoption of recommendations to ensure treatment suitability [41]. The reasons behind observations of poorer outcomes could be explored further using qualitative methods, both with patients belonging to at-risk groups and with clinicians delivering treatment.

A higher number of treatment sessions was associated with better outcomes for Black and Mixed individuals. Similarly, a higher number of sessions was associated with reduced drop-out for Asian and White-other groups. This is reinforced by previous research which has suggested that time waited to start treatment can lead to negative outcomes [11] and the current study found that primary waiting time was associated with increased risk of deterioration for people in the Mixed group. Reducing waiting times and increasing the number of sessions might support improved outcomes for these groups. While both of these actions are likely to be challenging for services to employ given the increasing demand for talking therapies and workforce shortages, innovative use of digital technologies to offer support remotely [42, 43] or to keep patients informed about their wait for treatment [44] are ways that services could address these challenges. Additionally, making organization-level adaptations such as supporting access to treatment in more accessible spaces (such as community, religious and non-healthcare settings) [12] can support people to access care more quickly, which is associated with improved outcomes and with lower likelihood of requiring more intensive and longer treatments [11]. As such, these adaptations might also be used to increase access within existing limited resources.

Despite observing some consistencies in certain outcomes for some ethnic groups across genders (such as for White-other groups and deterioration), there was variation when both gender groups were analysed independently. Previous research has identified that factors such as experiences of discrimination, cultural insensitivity and power imbalances impact access to mental health services for people from minoritized ethnic groups [10]. Research into differences in mental health service use across genders has suggested factors such as gendered societal expectations, mental health literacy and methods of communicating with health professionals can impact engagement [45,46,47]. This paper highlights the importance of services understanding how gender, including varying cultural perceptions of gender, may influence outcomes for different ethnic groups. Factors impacting treatment are likely to vary significantly between cultures and age groups and as such, future research involving more focussed exploration of specific factors that lead people to disengage from treatment could help services to understand what could reduce drop-out.

Factors associated with gender, may interact with organization-level factors to influence outcomes differently between ethnic groups. Existing research suggests gender inequalities in engagement are associated with intersecting factors of ethnicity, religion and socio-economic status [20]. Organization-level variables appeared to interact with ethnicity to influence outcomes in some cases, but the presence of a significant interaction did not necessarily result in significant outcome disparities when comparing people from minoritized groups to White-British people. These findings might be understood in terms of the intersecting gender and ethnicity characteristics that may be differentially impacted by societal or cultural factors contributing to differences in outcomes [20, 48].

The study included a binary categorisation of ethnic groups, and analyses using ONS ethnicity categories. Amalgamating people from minoritized ethnic groups into one group was useful to explore comparisons in terms of outcomes achieved by people who are racialised or perceived as ‘non-White’ versus those who are ‘White-British’; this approach reflects how people from minoritized communities can be socially assigned as ‘minorities’ and report differential treatment within services and healthcare discrimination [3, 5]. Across gender sub-group analyses, the White-other group experienced higher odds of deterioration compared to White-British individuals. This is interesting when considering the impact of ethnicity and discrimination on outcomes, especially in light of the discussion around social assignment of minority status. The White-other ethnic category includes people of European descent, including immigrants. A speculative explanation might be that these groups are racialised as ‘White’ by society, yet factors such as recency of immigration or not having English as their first language might counteract the beneficial effects of treatment. Despite being socialised as ‘minorities’, people from Black, Asian or other groups may experience a protective factor in comparison to people in the White-other category, if they were born in the UK or have been residents for a longer period of time. There is evidence to suggest that recency of immigration can impact psychological treatment outcomes [49]. This might be considered in further studies exploring differences in outcomes between ethnic groups, especially when making decisions about grouping by ethnicity. The study highlights the risk of ‘hiding’ potential differences between discrete groups which could lead to missed opportunities to improve outcomes for certain at-risk populations.

Using ONS ethnicity categories allowed for more granular exploration of differences between groups, yet categories remain high-level. The way in which variables such as ethnicity are structured when they are collected routinely for the purposes of clinical practice limited the number categories available for analysis as they were based on census data, and there was measurement error introduced by the cross-over of constructs related to race, ethnicity and nationality, all captured in the single variable. Amalgamating discrete ethnic groups for analysis risks erasure of important differences and nuances between groups and is an extant limitation of inequalities research; the risk of ‘hiding’ inequalities by grouping people together may result in misleading conclusions. Additionally, the data did not allow for analysis of impact of immigration or refugee/asylum seeker status on outcomes.

Categories of ‘male’ and ‘female’ are the only options patients had to self-identify their gender at the time data used in this study were collected. These are terms which more accurately refer to biological sex rather than gender, and their use limits the exploration of potential intersecting outcomes for people who identify their gender outside of these two categories.

Which factors should be adjusted for as confounders is contentious given the lack of consensus on causal pathways and the impact of decisions about adjustments on the interpretation of findings. Further research, using an updated sample of data would be of use to confirm the associations found in this study.

There was a large number of tests conducted for this study which might have increased the chance of making Type 1 errors [50]. However, in line with recommendations by Rothman (1990) [51] and Perneger (1998) [52] no adjustments were made to mitigate for this. All analyses have been presented irrespective of statistical significance and thresholds for statistical significance were not used to inform interpretations of the findings. This does not remove the possibility of some Type 1 errors but does prevent many of the issues related to data-mining or ‘p-hacking’ [50, 53].

The results show ethnic and gender differences in outcomes. Controlling for other factors did not reduce the likelihood of treatment drop-out for Black, Asian or ‘Other’ individuals, suggesting challenges in treatment retention and engagement remain. People in the Asian, ‘Other’ and White-other groups experienced worse outcomes than White-British people across all outcomes, suggesting that additional changes to treatment may be necessary to improve outcomes. The results provide insight into the different organization-level factors that might be adapted as part of IAPT care to improve outcomes for people with different characteristics. Interactions observed regarding drop-out for the Asian group (for factors including number of sessions received and primary waiting time) suggest these organization-level factors may play an important role in treatment retention. Increasing the number of sessions offered and reducing waiting times are actions that services could adopt and monitor the impact on outcomes. Further research is needed to better understand which other factors may interact with ethnicity to influence outcomes in Asian, ‘Other’ and White- other groups. Ethnic group might not be the driving factor for these differences which may be attributed to interactions between organization-level factors and other variables (perhaps linked to ethnicity) not explored in this study. Existing research shows that making adaptations to organization-level factors can improve outcomes for people from minoritized ethnic groups, but the results of this study indicate that it is likely that a variety of factors contribute to the success or failure of treatment to lead to better outcomes for different people. Finally, consideration should be given to ethnic group categorisation, due to the potential for issues impacting discrete groups to be missed. Future research should avoid amalgamating all ‘White’ groups as this may lead to failure to identify hidden inequalities which may be linked to immigration and other factors.