An investigation of treatment return after psychological therapy for depression and anxiety

Abstract Background: Some patients return for further psychological treatment in routine services, although it is unclear how common this is, as scarce research is available on this topic. Aims: To estimate the treatment return rate and describe the clinical characteristics of patients who return for anxiety and depression treatment. Method: A large dataset (N=21,029) of routinely collected clinical data (2010–2015) from an English psychological therapy service was analysed using descriptive statistics. Results: The return rate for at least one additional treatment episode within 1–5 years was 13.7%. Furthermore, 14.5% of the total sessions provided by the service were delivered to treatment-returning patients. Of those who returned, 58.0% continued to show clinically significant depression and/or anxiety symptoms at the end of their first treatment, while 32.0% had experienced a demonstrable relapse before their second treatment. Conclusions: This study estimates that approximately one in seven patients return to the same service for additional psychological treatment within 1–5 years. Multiple factors may influence the need for additional treatment, and this may have a major impact on service activity. Future research needs to further explore and better determine the characteristics of treatment returners, prioritise enhancement of first treatment recovery, and evaluate relapse prevention interventions.


Introduction
Mental health problems are widespread globally, with more than 300 million suffering from depression, and a similar number suffering from anxiety (World Health Organisation, 2017).Healthcare systems were estimated to spend a global total of US$0.8 trillion on the treatment of mental health problems in 2010, with this amount projected to more than double to US$2 trillion by 2030 (Trautmann et al., 2016).In England, the National Health Service (NHS) spent £14.3 billion on mental health services in 2020-2021 alone (Baker, 2021).Despite this expenditure, many barriers preventing access to treatment remain.For example, the supply of appropriately trained professionals often does not meet population demand (Andrilla et al., 2018;Health Resources and Services Administration, 2016), consequently creating long wait-times for treatment during which problems can become chronic (Goldner et al., 2011;Triggle, 2019).
Another factor that may impact treatment costs and waiting times is the phenomenon of patients returning to services for additional treatment, previously labelled as the 'revolving door' phenomenon (Roscoe, 2019).Treatment return may indicate persistent distress, whilst also further lengthening wait-times for first-time treatment seekers (Kazdin, 2018).This potential issue may have been particularly problematic in recent times, considering that referrals to mental health services have also been increasing gradually since the first stages of the COVID-19 pandemic (Bauer-Staeb et al., 2021;NHS Digital, 2022), and that multiple initiatives to improve access to psychological treatment for new and under-represented populations have recently been introduced (e.g.frontline healthcare staff, people from ethnic minority backgrounds; Beck et al., 2019;Cole et al., 2020).It is therefore important to understand the extent of treatment return and its impact on service efficiency, so that services can be supported to effectively manage the increased uptake of psychological treatment in healthcare systems where it is available.
However, there are relatively few studies that have examined treatment return.Available evidence indicates that 14% of adult patients in routine practice (Boerema et al., 2016), 28% of youth patients (Reeder et al., 2020), and 30% of university student patients (Kilcullen et al., 2021) seek additional mental healthcare support within 2-4 years of an initial treatment episode.Further research is necessary to better understand the extent of treatment return across different service contexts, care systems, countries, disorders, and interventions.
This study aimed to help address this research gap by providing an initial exploration of treatment return rates in a psychological therapy service that is part of the NHS Talking Therapies for Anxiety and Depression programme (previously known as the Improving Access to Psychological Therapies programme; IAPT) in England (Clark, 2011).Given the focus of this programme on maximising the availability of therapy at a national level, the rate of treatment return is of central interest to the adequate functioning of this programme.Two objectives guided this study: (1) to determine recovery, premature discontinuation, and treatment-return rates for initial and subsequent treatment episodes; and (2) to compare the clinical characteristics of the treatment return group with those of the non-returning group.

Design, setting and ethics
This study was a retrospective, observational cohort study, based on the analysis of routinely collected clinical data at a psychological therapy service in the north of England (January 2010 to June 2015), which is part of the NHS Talking Therapies programme (IAPT).Archival clinical records were collected, which included unique NHS identification numbers for each patient, and this enabled accurate identification of repeated treatment episodes for specific patients.To ensure that patients had at least one year following treatment discharge to be able to return for further treatment, only data for those patients who were first discharged between January 2010 and July 2014 were included.Within this period, 1.4% of treatment episodes had missing NHS episodes, and these cases were excluded.
NHS Talking Therapies services follow National Institute for Health and Clinical Excellence (NICE) guidelines (National Institute for Health and Clinical Excellence, 2011; National Institute for Health and Clinical Excellence, 2022) for the treatment of depression and anxiety (Clark, 2011).This is achieved using a stepped-care approach where most patients are initially offered low-intensity psychological treatments, which consist of brief, highly structured, protocol-driven, guided self-help interventions facilitated by qualified Psychological Wellbeing Practitioners (PWPs).If patients do not respond to this initial treatment, or they present with more severe symptoms or complex problems, they are subsequently 'stepped-up' and offered high-intensity interventions, which primarily involve lengthier psychotherapeutic interventions delivered by qualified and accredited therapists.Following this stepped-care approach, a single treatment episode may involve access to more than one intervention (e.g.low-intensity followed by high-intensity).All low-intensity interventions delivered in NHS Talking Therapies services are based on principles of cognitive behavioural therapy (CBT), while formal CBT is also the most commonly available high-intensity treatment in this setting (Baker, 2018;Clark, 2018).Nevertheless, other high-intensity therapies such as counselling for depression and interpersonal psychotherapy are also offered.Following clinical guidelines (National Institute for Health and Clinical Excellence, 2011), patients with some conditions (e.g.post-traumatic stress disorder, social anxiety disorder) are only offered high-intensity treatments.

Participants
During the study period, N = 54,339 patients were referred to the service at least once, of which n = 21,086 accessed at least one episode of treatment.A single treatment episode was determined to have occurred when a patient was recorded as having at least two appointments with the service, 1 consistent with national guidelines (National Collaborating Centre for Mental Health, 2020).Due to data-inputting errors in clinical records, n = 29 patients were excluded from further analyses.In addition, n = 28 patients were excluded as they were referred for a second episode of treatment within 7 days of being discharged from their first treatment.Due to the short time frame between episodes, it was hypothesised that the first treatment was probably ended for administrative reasons.This exclusion criterion was also adopted to enable comparison with the study by Boerema et al. (2016), which also excluded patients who were referred for a second episode within 7 days of initial treatment discharge.Following the exclusion of these n = 57 patients, a final sample of n = 21,029 was analysed to address the study objectives.An average of 1079.9 days (3.0 years) elapsed between patients' discharges following their initial treatment episode and the end of the study period (SD = 439.7 days; range = 365-1986 days; IQR = 677-1451 days).Demographic and clinical information on the study sample can be found in Table 1.

Measures
Patient Health Questionnaire (PHQ-9; Kroenke et al., 2001) The PHQ-9 is screening tool for major depression containing nine items assessing how often a specific symptom is experienced over a 2-week period (measured on a 0-3 scale).Summed responses produce an overall severity score (range = 0-27), a cut-off of ≥10 detects clinically significant depression (Kroenke et al., 2001), and a difference of ≥6 points indicates statistically reliable change (National Collaborating Centre for Mental Health, 2020).
Generalised Anxiety Disorder Assessment (GAD-7; Spitzer et al., 2006) The GAD-7 is a screening tool for anxiety disorders containing seven items assessing how often a symptom is experienced over a 2-week period (measured on a 0-3 scale).Summed responses produce an overall severity score (range = 0-21), a cut-off of ≥8 detects an anxiety disorder (Kroenke et al., 2007), and a difference of ≥4 points is indicative of statistically reliable change (National Collaborating Centre for Mental Health, 2020).
Work and Social Adjustment Scale (WSAS; Mundt et al., 2002) The WSAS is a questionnaire that assesses the impact of a mental health problem on five life domains (work; home management; social life; leisure activities; family and relationships).The impact on each domain is measured on 0-8 scales (i.e.0 is 'no impairment', 8 is 'severe 1 At least two appointments were required as the first recorded appointment involves an initial assessment to determine suitability for psychological treatment.Patients with two or more recorded appointments, therefore, accessed at least one therapy session after the initial assessment.impairment'), and the five scores are summed to derive an overall score of functional impairment (range = 0-40).
Pre-and post-treatment severity scores on the above three measures were available.Missing data from final treatment sessions were imputed using the last-observation-carried-forward (LOCF) method, to analyse data in a way that is consistent with other NHS Talking Therapies studies (Clark et al., 2009).Given that only pre-and post-treatment data were available for the study, it was not possible to distinguish whether patients received only low-intensity, only highintensity, or combined treatments.As such, data were analysed for 'treatment episodes', defined as access to psychological treatment in a stepped-care service within a fixed period of time defined by referral and discharge dates.

Outcome definitions Treatment return
This was defined as when a patient started another new episode of treatment after being discharged from their first treatment within the study period for any reason (e.g. agreed end of care, premature discontinuation, referral to another service).'Non-returners' received one episode of treatment, 'single returners' received two treatments, and 'frequent returners' received three or more treatments.Therefore, both 'single returners' and 'frequent returners' received multiple episodes of treatment and are thus both classed as 'treatment returners'.As an observational study based on routine clinical data, the dataset was limited by the data collection time frame, which makes it impossible to determine whether patients had received treatment before the study period.

Clinical outcomes
These were defined following the IAPT Manual (National Collaborating Centre for Mental Health, 2020) and definitions proposed by Clark (2018).A patient was considered to have shown 'reliable improvement' after therapy when their PHQ-9 or GAD-7 score reliably decreased (i.e. a decrease of ≥6 points for the PHQ-9, or ≥4 points for the GAD-7) and their score for the other scale did not reliably deteriorate, compared with pre-treatment measures.'Reliable deterioration' occurred when a patient's PHQ-9 or GAD-7 score reliably increased over treatment and their score on the other scale did not reliably improve.'Stasis' refers to a lack of either reliable improvement or reliable deterioration.'Recovery' occurred when patients scored above the clinical cut-off on either PHQ-9 and/or GAD-7 at an initial (pre-treatment) assessment (i.e.met 'caseness'), and both their PHQ-9 and GAD-7 scores dropped below the clinical cut-offs post-treatment.'Reliable recovery' occurred when a patient met criteria for both recovery and reliable improvement.'Premature discontinuation' was determined to have occurred when clinical records indicated that the patient either declined further therapy sessions after they started the treatment, or if there had been an unscheduled discontinuation of treatment.Finally, a patient was classified as 'relapsed' when: (1) they had met criteria for recovery at the end of their initial treatment; (2) at least one of their PHQ-9 or GAD-7 scores was above the cut-off at the beginning of the second treatment; and (3) the outcome score above the cut-off displayed reliable deterioration.
It should be noted that the reliable change indices for the PHQ-9 and GAD-7 adopted for the measure of relapse (≥6 and ≥4, respectively) differ to indices applied by other studies previously conducted.For example, Ali et al. (2017) adopted an index of ≥5 for both measures when investigating relapse following low-intensity CBT.The indices adopted in this study were applied to ensure consistency with national guidelines (National Collaborating Centre for Mental Health, 2020).

Analyses
Descriptive statistics were used to summarise service utilisation and outcomes, and this included calculating premature discontinuation, recovery, and treatment-return rates, alongside average assessment-wait time and average treatment durations.Service utilisation and clinical outcomes were compared for non-returners and treatment returners.The potential relevance of the amount of time that patients had available within the study period to be able to return for treatment (i.e.time between initial treatment discharge and the end of the study period) on treatment return status was also examined.Between-group comparisons applied independent samples t-tests and Pearson's chi-squared tests.

Treatment pathways and outcomes
Within this section, reported statistics for each variable were calculated using only those cases with a recorded response for the relevant variable (i.e.excluding missing values).Figure 1 illustrates the STROBE summary (Vandenbroucke et al., 2007) of treatment pathways.Overall, N = 21,029 were discharged following their first treatment within the study period, and these patients had waited an average of 26.2 days (SD = 22.8; range = 0-377) between referral and assessment.At assessment, 89.4% (n = 18,686) met criteria for caseness (i.e.scored above clinical cut-off on either PHQ-9 and/or GAD-7).Treatment involved an average of 7.3 sessions (SD = 5.9; range = 2-52), spread over an average of 161.4 days (SD = 115.6;range = 1-1343), with 34.5% (n = 6985)

Treatment return
A total of n = 2878 patients (13.7%) returned for additional treatment episodes within 1-5 years.Specifically, n = 2567 (12.2%) returned for a single additional treatment episode (i.e. the single return group), whilst n = 311 (1.5%) returned multiple times (i.e. the frequent return group).
Figure 2 illustrates the proportions of non-returners, single returners and frequent returners, with associated recovery rates.Whilst n = 2759 patients (13.1%) were referred for additional therapy after their initial treatment, they did not attend further sessions.Therefore, a total of 26.8% (n = 5637) patients were re-referred following first treatment and approximately half of these actually went on to receive another treatment episode.Across the study period, the investigated n = 21,029 patients received a combined total of 232,663 therapy sessions.The subgroup of 13.7% treatment returners received a total of 62,105 sessions across both initial treatment episodes and subsequent episodes (i.e.26.7% of all sessions provided to the clinical population).Specifically, 14.5% (33,663) of the total sessions provided by the service were delivered during additional treatment episodes.
For those patients who returned to treatment at least once, the average time between first discharge and second referral was 483.4 days (SD = 384.3;range = 8-1822; IQR = 169-727).It was identified that 6.7% (n = 1417) of the total sample returned to treatment within the first year immediately following initial treatment discharge (49.2% of all treatment returners).Meanwhile, when excluding n = 5945 patients who were not followed up for at least two years within the study period, it was found that 11.1% (n = 1678) of the remaining sample returned within two years.Lastly, when excluding n = 10,825 patients who were not followed up for at least three years, a total of 14.6% (n = 1490) were identified as having returned for treatment within three years of initial treatment discharge.
Of all n = 2878 returning patients, n = 1659 (58.0%) were above the clinical threshold for either PHQ-9 and/or GAD-7 at the first discharge, indicating that they did not fully recover from their symptoms at the end of their initial treatment episode.Meanwhile, n = 915 (32.0%2 ) patients who were classed as recovered at the end of their prior episode had experienced a demonstrable relapse by the start of the second treatment episode.By the end of the study period, n = 2576 returning patients had been discharged following a second treatment episode (with the remaining still undergoing treatment).Of these, n = 2359 (92.3%) met caseness criteria at the beginning of this episode, with 36.4% (n = 859) recovering following the second treatment, and 34.8% (n = 821) reliably recovering.A total of n = 2194 treatment returners who had been discharged from two treatment episodes met criteria for caseness at the start of both treatments.Nearly half of this subsample did not recover following either treatment (n = 1042; 47.5%).Meanwhile, 18.9% (n = 415) recovered following both treatments; 17.4% (n = 381) recovered following the first but not the second treatment and 16.2% (n = 356) recovered following the second treatment after not recovering following the first.For the second treatment, pre-to-post treatment changes in PHQ-9, GAD-7 and WSAS scores reflected effect sizes of d = 0.81, d = 0.88 and d = 0.56, respectively.
A total of n = 311 returned for a third treatment, with n = 264 of these being discharged from the service following this treatment before the end of the study period.Of these n = 264 patients, 94.3% (n = 246) met caseness criteria at the beginning of this third episode, with 34.6% (n = 85) recovering by the end of treatment, and 34.1% (n = 84) reliably recovering.For the third treatment, pre-to-post treatment changes in PHQ-9, GAD-7 and WSAS scores reflected effect sizes of d = 0.81, d = 0.93 and d = 0.67, respectively.A total of n = 47 returned for a fourth treatment, with n = 30 concluding their episode before the end of the study period.Of these, 90% (n = 27) met caseness criteria, with six (22.2%) recovering following treatment, and all six having a reliable recovery.For the fourth treatment, pre-to-post treatment changes in PHQ-9, GAD-7 and WSAS scores reflected effect sizes of d = 0.60, d = 0.65 and d = 0.24, respectively.Three patients returned for a fifth treatment episode.As can be seen in Fig. 2 and the reported statistics above, recovery rates and symptom changes either decreased or remained static over treatment episodes.

Comparisons between treatment return type
Few differences were found between non-returners (NR) and treatment returners (TR) in terms of service utilisation and clinical outcomes related to the first treatment episode.Each group waited similar times for assessment (NR mean = 25.9 days, TR mean = 28.2days), received similar amounts of treatment (NR mean = 7.2 sessions, TR mean = 7.6 sessions), and had similar amounts of time between assessment and discharge (NR mean = 161.0days, TR mean = 164.0days).Although there was no significant difference in terms of time between assessment and discharge (p = .195),significant differences in terms of assessment wait times (t 3664 = 4.7, p<0.001, d = .10)and sessions received (t 3571 = 3.1, p = 0.002, d = .06)were identified.However, effect sizes suggest little, if any, relationship between treatment return and assessment wait time, or treatment return and contacts received.
Approximately one-third of patients in each group prematurely discontinued the first treatment (NR = 34.3%,TR = 36.4%).There was a significant difference in terms of premature discontinuation (χ 2 (1) = 4.54, p = .033),but the effect size suggests little, if any, association between premature discontinuation and treatment return (Cramer's V = .01).Finally, there appeared to be a small difference in terms of recovery following treatment for those patients who initially met criteria for caseness, with non-returners having both a higher recovery rate (NR = 43.5%,TR = 38.6%)and higher reliable recovery rate (NR = 41.5%,TR = 36.7%)than treatment returners.Again, despite significant differences being found in terms of recovery (χ 2 (1) = 22.42, p<.001) and reliable recovery (χ 2 (1) = 21.73,p<.001), effect sizes suggest little, if any, associations between recovery and treatment return (Cramer's V = .03,for both findings).
One difference that was identified between the two groups was related to the study method; specifically, the average length of time that each group had available within the study period to be able to return for treatment (i.e.time between first treatment episode discharge date, and end date of study period).Indeed, treatment returners had a longer average length of time available (mean = 1203.1 days) than non-returners (mean = 1060.4days).These differences were significant, with the relationship representing a small effect (t 3996 = 17.1, p<0.001, d = .33).All comparisons were repeated exploring three groups of participants (non-returners, single returners, and frequent returners) and similar results were observed (see Supplementary material).

Discussion
This study investigated treatment return within routine psychological care across a five-year observation period.A large sample of naturalistic, routinely collected data from an English steppedcare mental health service was examined.We found that 26.8% of patients were re-referred after initial treatment, with approximately half of these patients accessing additional treatment.Overall, a treatment return rate of 13.7% within 1-5 years was observed.This return rate is comparable to the 14% return rate within 3 years reported by Boerema et al. (2016), in their investigation of N = 85,754 patients seeking treatment for depression in the Netherlands.Whilst the current study investigated treatment return following predominantly CBT-based treatments, Boerema et al. (2016) investigated return from a range of interventions (predominantly supportive psychotherapy and/or pharmacotherapy).This suggests that the rate of treatment return in routine care for common mental health problems may be similar regardless of interventions and/or service settings.
Combined, these studies indicate that approximately one in seven patients who receive psychological treatment in routine care return for additional treatment within 1-5 years.Delivering additional treatment to these patients appears to impact service efficiency, as the group of returning patients accounted for 26.7% of all the therapy sessions provided in the service.Consequently, this may limit or delay access to those seeking initial treatment (Kazdin, 2018), in a context of broader increases in referral rates and treatment uptake (NHS Digital, 2022).This is particularly concerning considering that longer wait-times within NHS Talking Therapies services are associated with worse treatment outcomes (Clark et al., 2018).Furthermore, the treatment return rate of 13.7% may also be an under-estimation as it does not take account of those patients who returned to treatment after the observation period, received treatment that occurred before the observation period, or received additional treatment at different services (e.g.accessed pharmacotherapy or private therapy, etc.).
One potential explanation as to why some patients seek additional treatment is that the initial treatment episode was not effective at addressing the initial problems experienced by patients.In this study, 58.0% of returning patients had enduring symptoms of depression and/or anxiety above clinical thresholds at the end of the first treatment, indicating that initial treatment did not lead to recovery.However, many patients also did achieve recovery during the first treatment and yet still returned.Many of these patients experienced a deterioration following discharge, as demonstrated by 32.0% of treatment returners experiencing a demonstrable relapse by the beginning of the second treatment.Depression and anxiety have been demonstrated to be highly recurrent disorders (Bruce et al., 2005;Hardeveld et al., 2010), and relapse is common following psychological interventions (Delgadillo et al., 2018;Vittengl et al., 2007).It may be that some patients seek additional treatment due to a lack of initial treatment effectiveness, while others return due to relapse or recurrence.This reflects the finding of Siddall et al. (1988), that some treatmentreturning patients perceive their previous therapy as being helpful, while others perceive it as being not helpful, and some patients return for a new problem, while others return for the same problem.Overall, it is plausible that there are multiple reasons why patients return to treatment.
An alternative perspective on the 'revolving door' phenomenon is that treatment return represents a successful outcome in which patients are socialised during the initial treatment to implement healthy and adaptive help-seeking behaviours.Indeed, treatment return is perhaps an inevitable clinical reality of routine practice.In typical controlled trials, patients cannot return to a clinical trial after the trial has ended.Clearly, patients who feel they have relapsed, or are suffering due to an ongoing problem, should feel encouraged to seek ongoing support.Consequently, it is important that returning to psychological services should not be actively discouraged.However, it is also important to consider the economic costs for services and the opportunity costs for clinical populations of frequent treatment return.The present findings indicate that psychological interventions may be less effective with each additional treatment episode (i.e.lower recovery rates and pre-post symptom changes).The findings from this study indicate that, relative to patients who have a single episode of treatment (recovery = 43.5%), the recovery rates for treatment returners are lower (first episode recovery = 38.6%),and these decrease further after the second (36.4%), third (34.6%) and fourth (22.2%) treatment episodes.Therefore, it should be a priority for future research and clinical practice to consider ways to minimise the need for treatment return and reduce its occurrence, without actively discouraging it on an individual basis.

Limitations and directions for future research
One significant limitation of this study was its retrospective, observational design, which involved the analysis of routinely collected data from a single NHS Talking Therapies service.As patients were not routinely followed up by the service after discharge, it is possible that some patients classed as non-returners may have received additional treatment at other mental health services or private providers.The lack of prospective monitoring of symptom changes also precluded a more detailed investigation of relapse occurrence between initial treatment discharge and treatment return.For example, some seemingly-recovered patients may have experienced unmonitored relapses following initial treatment discharge, and then returned for treatment while not meeting full relapse criteria.In addition, the routinely collected data had limitations, such as the omission of whether patients had received previous treatment.Due to this, some patients labelled as nonreturners may have been returning to the service after a previous treatment episode that occurred before the study period.Consequently, the limited information related to clinical history and posttreatment experiences limits our ability to assess the true extent of the rates of treatment return and relapse, and the potential associations between baselines characteristics and treatment return.
Another related limitation was that patients did not have similar amounts of time to be able to return to treatment following their treatment discharge.Those patients with the largest amount of time available would have had greater opportunity to return for additional treatment compared with those with the smallest.Indeed, treatment returners had significantly longer post-initialtreatment-discharge periods than non-returners.Furthermore, when examining a subsample of patients who were followed up for at least three years, a three-year treatment return rate of 14.6% was estimated (i.e. a higher rate identified than when looking at the full sample with varying lengths of follow-up time).Consequently, considering that some patients may have received additional treatment from different, unmonitored mental health services, and that some patients had less opportunity than others to return to the same service within the study period, it is likely that this study under-estimated the rate of treatment return.
This study was also limited by only analysing data collected by a single service, using an archival and historical dataset.Considering that national access and recovery rates have increased in recent years (NHS Digital, 2022), this may limit the generalisability of the findings to services operating today.Indeed, the estimated recovery rate (42.8%) was smaller than the national recovery rate for NHS Talking Therapies observed in 2021-2022 (50.2%;NHS Digital, 2022).However, the estimated recovery rate was identical to the national recovery rate observed at the time of data collection (2012)(2013), and the recovery rate observed for the service in 2021-2022 (41%) was in fact smaller than its recovery rate estimated in this study.Furthermore, baseline data for patients who met caseness criteria (PHQ-9 mean = 16.22,SD = 5.65; GAD-7 mean = 14.38,SD = 4.43) were highly similar to recent publicly available data for the same service (PHQ-9 mean = 15.9,SD = 5.6; GAD-7 mean = 14.4,SD = 4.4; NHS Digital, 2022) and for the national programme (PHQ-9 mean = 15.6,SD = 5.5; GAD-7 mean = 14.3,SD = 4.4; NHS Digital, 2021).Consequently, although it is possible that there is variability in the extent of treatment return between services and over time, it is likely that these findings still apply to current practice.Furthermore, we did not have access to more granular data to identify whether patients received low-intensity and/or high-intensity interventions during treatment episodes.This prevented the comparison of treatment return rates between service steps and prevented an estimation of the health economic costs of delivering additional treatment.
Due to these limitations, further studies are required in which patients are observed for consistent periods of time.Such studies also need to incorporate a system of monitoring patients in which both their clinical history and their access of additional treatment through different treatment providers can be considered.Future research should also compare treatment return rates between different services and different interventions, thus enabling variability between services in terms of treatment return to be examined.Qualitative methods could also help to explore the reasons as to why patients seek additional treatment to provide a greater understanding of the different pathways behind treatment return.

Clinical implications
This study indicates that treatment return may hinder service efficiency and that treatment effectiveness may be reduced or remain static with each subsequent treatment episode delivered.This highlights the importance of the effectiveness of the first treatment delivered, maintenance of treatment gains, and the patient learning of skills to manage their relapse risks.To help increase recovery rates, services should ensure the delivery of the appropriate number of treatment sessions across all treatment episodes (i.e.provide the appropriate 'dose'), as per relevant clinical guidelines (e.g.National Institute for Health and Clinical Excellence, 2011; National Institute for Health and Clinical Excellence, 2022).To prevent relapse, services could deliver booster sessions after achieving recovery, as these have been demonstrated previously to help to prevent relapse (Gearing et al., 2013;Whisman, 1990).NICE guidelines state that such sessions should be offered when patients are at high risk of relapse (e.g.due to having residual depressive symptoms; National Institute for Health and Clinical Excellence, 2022).However, between April and September 2021, only 1.9% of patients who finished treatment in NHS Talking Therapies services received a follow-up appointment (NHS Digital, 2022).This demonstrates that follow-up is rarely provided, and services need to consider how this provision can be implemented whilst also meeting ongoing demand for treatment.Additionally, further research is required to better understand moderators and predictors of relapse and to develop and evaluate low-intensity relapse prevention interventions.

Conclusion
This study acts as an initial exploration of treatment return within stepped-care psychological services.Further explorations are required in which patients are prospectively observed for consistent periods of time, and more clinical information is available.This study estimates that approximately one in seven patients receiving treatment for common mental health problems in routine practice return to the same service for additional treatment within 1-5 years.This phenomenon may impair service efficiency by increasing waiting lists and limiting access.Treatment return should not be stigmatised or actively discouraged, but the identification of methods that help reduce the need for its occurrence should be prioritised.Indeed, there may be different explanations for treatment return, with some patients returning to address an unresolved initial problem and others returning due to relapse or recurrence.Future research should therefore prioritise improving recovery and decreasing relapse to minimise the personal, societal and economic impact of the 'revolving door' phenomenon.
Supplementary material.To view supplementary material for this article, please visit https://doi.org/10.1017/S1352465823000322 Data availability statement.Requests for access to data should be addressed in writing to the data custodian: j.delgadillo@ sheffield.ac.uk.

NFigure 1 .
Figure 1.Flow chart of different pathways taken by patients through the psychological service.End of study period = June 2015.CSP, met criteria for clinically significant problem; RR, recovery rate; PD, premature discontinuation.

Figure 2 .
Figure 2. Proportions of sample who did not return for treatment, returned for treatment at least once, or returned multiple times.RR, recovery rate.Some patients returned for a fourth (N = 47) and fifth (N = 3) treatment episode during the study period.These were not included in this figure due to limited sample sizes.
Finally, services need to consider what interventions have been delivered previously and what interventions are currently available when providing additional treatment.

Table 1 .
Demographic and clinical information related to first treatment received by patients (N = 21,029) a Percentages calculated using only those cases with full data (i.e.excluding missing).Abbreviations: GAD-7, Generalized Anxiety Disorder scale; PHQ-9, Patient Health Questionnaire; WSAS, Work and Social Adjustment Scale.