The impact of employment advisors in NHS Talking Therapies on benefits receipt, England: April 2014 to April 2024

• Employment Advisor (EA) support is a voluntary service available through NHS Talking Therapies (NHSTT) alongside psychological therapy; the service supports individuals who have employment needs that could be preventing them from returning to, staying in or finding work.

• Among individuals not working at the start of NHSTT treatment, those receiving EA support alongside psychological therapy experienced an average decrease of 2.7 percentage points in the probability of receiving any benefit after three years, compared with the expected change from psychological therapy only.

• Among individuals not working and reporting a long-term sickness or disability at the start of NHSTT treatment, those receiving EA support alongside psychological therapy were 3.2 percentage points less likely to be receiving Universal Credit or employment support benefits after three years, compared with the expected change from psychological therapy only.

• Individuals not working and reporting a long-term sickness or disability at the start of NHSTT treatment who received EA support alongside therapy were 2.4 percentage points more likely to be receiving disability benefits after three years, compared with the expected change from psychological therapy only; this may reflect underlying health conditions that are unlikely to be affected by psychological therapy, particularly given that some disability benefits are not means-tested.

• Among people employed and working at the start of NHSTT treatment, there was an average 4.2 percentage point increase in the probability of receiving any benefit after three years for those receiving EA support alongside psychological therapy, compared with the expected change from psychological therapy only.

• In this analysis, we matched individuals receiving EA support alongside psychological therapy with individuals receiving psychological therapy only in an area where EA support was unavailable; however, unmeasured differences between the groups, such as work-related difficulties among those in employment, may mean some effects are not because of receiving EA support alone.

This project is part of the Support2Work programme, funded by the National Institute for Health and Care Research (NIHR) under its Work and Health Research Initiative (Grant Reference Number NIHR207664). The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.

This study examined the impact of Employment Advisor (EA) support on benefits receipt as part of the Support2Work programme.

We have published the impact of EA support on earnings and employment status in our The impact of employment advisors in NHS Talking Therapies on employee earnings and employment status, England: April 2014 to March 2025 bulletin, which used the same study cohort.

Further work, conducted by other members of the Support2Work team, will include in-depth qualitative interviews with EA staff, service leads, therapists, and service users. The work will also include an examination of the wider effects of EA support; for example, the impacts on population health (including mental health), healthcare costs, and societal welfare.

The impact of EA support was estimated across five benefit categories:

• Any benefit

• Disability benefits

• Employment support benefits

• Universal Credit benefits excluding employment support benefits

• Universal Credit benefits including employment support benefits

We used genetic matching to match individuals who received both EA support and psychological therapy with similar individuals who received psychological therapy only. The individuals who received therapy only were in a geographical area where EA support was unavailable, rather than those who opted not to take up support. We used staggered difference-in-difference modelling to compare changes in benefits receipt over time to estimate the additional impact of EA support.

We used the psychological therapy-only group to estimate the outcome trajectory that EA support recipients would have been expected to follow without EA support. This allowed us to isolate changes associated with the intervention of EA support beyond underlying trends, and the effects of therapy alone. Definitions of the terms in this section are available in Section 4: Glossary.

The main results are shown in Figure 1. Further information is available in our accompanying dataset, including results stratified by age group, sex, presenting symptoms, number of EA sessions, mental health recovery status, and Index of Multiple Deprivation (IMD) quintile group.

Figure 1: The largest reduction in the probability of receiving any benefit was observed among non-working individuals up to three years after first receiving employment support

Changes in the probability of benefits receipt for all employment statuses, with time since first receiving employment advisor (EA) support, aged 18 to 65 years, recipients of EA support alongside therapy in NHS Talking Therapies between 1 April 2018 and 31 March 2023, compared with the expected change from psychological therapy only

Notes

1. Data include individuals who received employment advisor (EA) support alongside psychological therapy in NHS Talking Therapies (NHSTT) between April 2018 and 31 March 2023, attended at least one EA session and at least two psychological therapy sessions, were aged between 18 and 65 years at the time of NHSTT assessment, were resident in England at the time of treatment, and were recorded as usual residents in the 2011 Census or Census 2021. 

2. Outcomes are compared with a matched sample of individuals who received psychological therapy only, in an area where EA support was unavailable.

3. The error bars are 95% confidence limits.
4. There are five benefit groups, including any benefit group (any benefit from the other categories); disability benefits (Disability Living Allowance and Personal Independence Payment); employment support benefits (Employment Support Allowance, Incapacity Allowance, Passported Incapacity Benefit, and Severe Disablement Benefit); Universal Credit benefits excluding employment support benefits (Income Support, Job Seeker's Allowance, Housing Benefit, and Universal Credit); and Universal Credit benefits including employment support benefits (Income Support, Job Seeker's Allowance, Housing Benefit, Universal Credit, Employment Support Allowance, Incapacity Allowance, Passported Incapacity Benefit, and Severe Disablement Benefit).
5. Benefits receipt is defined as receiving a benefit payment greater than £0 in a given quarter.
6. The Benefit dataset follow-up covers the period between 1 April 2014 and 30 April 2024.

NHS Talking Therapies

NHS Talking Therapies (NHSTT) is a free national NHS service in England that provides evidence-based psychological talking treatments for adults experiencing common mental health difficulties, particularly anxiety and depression disorders. The service started in 2008 and offers a range of National Institute for Health and Care Excellence (NICE) recommended therapies such as cognitive behavioural therapy, counselling, mindfulness and guided self-help, delivered face-to-face, remotely, one-to-one, or in group settings. Anyone registered with a general practitioner (GP) in England can access the service.

More information can be found on the NHS Talking Therapies for anxiety and depression webpage.

Employment advisors

The Employment Advisors (EAs) pilot started within NHSTT from 2018, with expanded rollout from 2022. The rollout was implemented in phases. Initially, only a small number of Clinical Commissioning Groups (CCGs) in England had EA access, but with each successive wave, EA access became available in additional CCGs. EA support is a voluntary intervention offered to all individuals accessing NHSTT where appropriate. It is funded by the Joint Work and Health Directorate and jointly sponsored by the Department for Work and Pensions and the Department of Health and Social Care.

EAs do not provide psychological therapy. Instead, they focus on practical, employment-related support for people who are unemployed, on sickness absence, or struggling to remain in work. The intention is for EA support to run alongside therapy, by addressing employment-related issues such as job searching, workplace challenges and return-to-work planning, although support may begin earlier in practice because of waiting times.

Average treatment effect on the treated

Average treatment effect on the treated (ATT) is the primary causal estimand, and measures the average effect of receiving EA support. ATT is defined at the cohort-time level, with cohorts indexed by calendar quarter of first EA contact.

It measures the difference between observed outcomes for treated individuals and the counterfactual outcomes that would have occurred in the absence of EA contact, using never-treated individuals as the comparison group. In this study, never-treated individuals are defined as having no record of receiving EA support within the study period (between 1 April 2018 and 31 March 2023), as well as receiving psychological therapy in a geographical area where EA support was unavailable.

This differs from individuals who were offered EA support but opted not to take up EA support. Identification relies on a conditional parallel trend assumption, which is conditional on pre-treatment covariates.

Staggered difference-in-differences approach

A staggered difference-in-difference (DiD) approach is used to estimate causal effects when treatment occurs at different points in time across individuals or cohorts. In this study, we use it to exploit the staggered geographical rollout of EA support across England to compare pre- and post-treatment outcomes for matched individuals, controlling for time-invariant unobserved heterogeneity and Lower-layer Super Output Area (LSOA)-level unemployment.

Cohort and time-specific effects are estimated using never-treated individuals with matched NHSTT assessment dates as clean controls, and are then aggregated following a modified Callaway and Sant'Anna approach. This allows treatment effects to vary over cohorts and time while avoiding bias from conventional two-way fixed effects models.

Employment status

Employment status is the employment status an individual identifies for themselves, based on information they provide at the start of their NHSTT treatment, reflecting their own understanding of their work status. Categories include:

• Employed

• Not working, long-term sick or disabled

• Not working, not seeking work

• Not working, seeking work

• Missing or not stated

We used information from self-reported sick status (whether an individual is off sick or not) to create further categories, including:

• Employed, working

• Employed, off sick

• Employed, sick status unknown

• Not working, seeking work

• Not working, not seeking work

• Not working, homemaker

• Not working, long-term sick or disabled

• Missing or not stated

Genetic matching 

We used genetic matching, implemented through the matchit function. Genetic matching uses an iterative search algorithm (each iteration is a "generation") to optimise matches by minimising imbalance across observed characteristics. It uses generalised Mahalanobis distance as the primary distance metric.

This helps to reduce systematic differences between groups and reduces potential confounding before estimating treatment effects. For characteristics used in matching, see Section 5: Data sources and quality.

Linked datasets

An extension of the Public Health Data Asset (PHDA) was used for this analysis. The de-identified, linked dataset includes data from:

• NHS Talking Therapies (1 April 2018 to 31 March 2023)

• the 2011 Census and Census 2021 for demographic and socio-economic information

• Office for National Statistics (ONS) death registrations (registered between 1 January 2009 and 31 December 2024)

• the HM Revenue and Customs (HMRC) Pay As You Earn (PAYE) Real Time Information (RTI) records (from 1 April 2014 to 30 April 2024)

• the Department for Work and Pensions (DWP) Benefit dataset (BENs) (1 April 2014 to 30 April 2024)

Datasets were linked using NHS numbers and encrypted National Insurance numbers using the ONS Demographic Index. All data were de-identified prior to analysis. More details of the linkage processes are provided in our The impact of employment advisors in NHS Talking Therapies on employee earnings and employment status, England: April 2014 to March 2025 bulletin.

Benefits

We included benefit types related to health and employment for working-age individuals in our analysis. We grouped these into five categories:

• Any benefit (comprises any benefit included in the following categories)

• Disability benefits (including Disability Living Allowance and Personal Independence Payment)

• Employment support benefits (including Employment Support Allowance, Incapacity Allowance, Passported Incapacity Benefit, and Severe Disablement Benefit)

• Universal Credit benefits excluding employment support benefits (including Income Support, Job Seeker's Allowance, Housing Benefit, and Universal Credit)

• Universal Credit benefits including employment support benefits (including Income Support, Job Seeker's Allowance, Housing Benefit, Universal Credit, Employment Support Allowance, Incapacity Allowance, Passported Incapacity Benefit, and Severe Disablement Benefit)

For some benefits, such as Universal Credit, applications can cover two partners living in the same household. In these instances, both partners were assigned as benefit recipients in our analysis. Attendance Allowance, Bereavement Benefit, Pension Credit, State Pension, and Widow's Benefit were excluded from the analysis.

Inclusion criteria

We included individuals if they could be linked to both NHS and National Insurance records and had a valid Census identifier.

The study population was restricted to individuals who:

• started NHS Talking Therapies (NHSTT) between 1 April 2018 and 31 March 2023

• attended at least two therapy sessions

• were aged 18 to 65 years at the time of their NHSTT assessment

• were resident in England at the time of treatment and recorded as usual residents in the 2011 Census or Census 2021

Students and retired individuals were excluded, so that we could focus our analysis on the working-age population. All included individuals had at least 12 months of follow-up. Where individuals had multiple eligible treatment episodes, the earliest episode was selected. For individuals with multiple eligible NHSTT pathways, the pathway with the earliest therapy start date was selected.

Exposure groups

The exposed group consisted of individuals who received psychological therapy and attended at least one Employment Advisor (EA) session between 1 April 2018 and 31 March 2023 within their NHSTT pathway. The unexposed group comprised individuals who received psychological therapy in an NHSTT service where EA support was not yet available, and where the individual had no recorded EA support during this period.

Follow-up

Individuals were followed from the start of their NHSTT pathway for a minimum of 12 months and up to a maximum of 36 months (pre- and post-therapy). For the exposed group, follow-up spanned periods before and after first contact with an EA.

The BENs follow-up covers the period between 1 April 2014 and 30 April 2024, and ends at the earliest of the following points:

• the end of the study period (30 April 2024)
• reaching age 66 years
• death

Pre-therapy follow-up was censored when participants were under 18 years of age. Because the study cohort included individuals who received EA support in NHSTT between 1 April 2018 and 31 March 2023, not all participants could be observed for the full three-year follow-up period within the BENs data, which were available until 30 April 2024. For example, an individual who first received EA support on 31 March 2023 could not contribute three years of post-EA follow-up. Sample sizes for each follow-up period are available in our accompanying dataset.

Matching

Matching was used to make the exposed and unexposed groups as comparable as possible. Each individual who received EA support was matched, on a one-to-one basis, with an individual who did not receive EA support, using genetic matching. Individuals who did not receive EA support were in a geographical area where EA support was unavailable, rather than those who opted not to take up support.

Further information on the matching methodology is available in our The impact of employment advisors in NHS Talking Therapies on employee earnings and employment status, England: April 2014 to March 2025 bulletin.

Quality

The quality of the PHDA is described in Section 5: Data sources and quality of our The impact of bariatric surgery on monthly employee pay and employee status, England: April 2014 to December 2022 article.

EA support is a voluntary service available to individuals with common mental health conditions who access NHSTT and are either in work but experiencing difficulties in the workplace, off work because of sickness, or looking for work. This is described by the Department for Work and Pensions (DWP) and Department of Health and Social Care (DHSC) in their Employment advisers in improving access to psychological therapies reports.

In areas where EA support was available, only 8.3% of people in the study population received EA support. As a result, any estimated effects are likely to reflect outcomes for a relatively small proportion of the eligible population, rather than the wider group of NHSTT users. To reduce selection bias, the comparison group comprised individuals in services where EA support was unavailable rather than those who opted not to take up support.

Although individuals were matched on a wide range of demographic, economic and clinical characteristics, independent unobserved differences may remain between those who did and did not receive EA support. In particular, we were unable to observe workplace-related factors that may influence both the reason for engagement with EA support and subsequent benefits receipt, such as difficulties at work, workplace conflict, job insecurity, risk of redundancy, or an individual's motivation to seek support, which could particularly affect the results for the employed group.

While the Work and Social Adjustment Scale (WSAS) was included in the matching process, and may partially capture some work-related difficulties, direct information on workplace circumstances was unavailable. This limitation is likely to be relevant among employed individuals, for whom reasons for accessing EA support may be more diverse and complex. As a result, some of the estimated effects may reflect residual differences between the groups rather than the impact of EA support alone.

Disability- and health-related benefits, particularly those that are not means-tested, may be awarded for health conditions other than common mental health conditions. Therefore, it is important to acknowledge that EA support and NHSTT treatment may have a limited impact on disability benefit receipt as an outcome, as these benefits may be linked to conditions that are unlikely to be affected by psychological therapy.

The BENs dataset extract used in this analysis does not allow us to distinguish between health-related and non-health-related Universal Credit (UC) claims. Employment support benefits have also been partially replaced by UC, meaning that the "employment support benefits" and "Universal Credit benefits excluding employment support benefits" categories may not be fully independent. To address this, we created an additional benefit category titled "Universal Credit benefits including employment support benefits", which combines UC and employment support claimants. However, given the substantially larger number of UC claims, any effects associated with employment support benefits are likely to be masked within this combined group.

We observe a statistically significant dip before time 0 in some models. This is common in mental health settings. Individuals often seek help after negative shocks (for example, job loss or worsening mental health), which can affect benefits receipt before treatment. This is worth noting in the context of parallel trends but does not invalidate the results. Pre-trends move in the opposite direction to post-treatment effects, suggesting estimates may be conservative. The pre-treatment decline likely reflects selection into support following adverse shocks (Ashenfelter's dip), rather than anticipation or treatment effects.

Acknowledgements

We would like to thank the Support2Work Service Experts Advisory Panel (SEAP) and Policy Expert Advisory Panel (PEAP) for their ongoing collaboration and feedback.