Our aim was to investigate whether a number of sociodemographic and clinical client factors could predict dropout from a primary care setting, based on indications from previous literature. This had not been studied in the PMHC service context until now. Our results partly support previous findings from the literature that specific sociodemographic factors can predict dropout. These were younger age, being unemployed, lower levels of education, and poor social support. Other sociodemographic factors identified with mixed results in the literature were not significant predictors in this context, such as sex and immigrant background. Contrary to our expectations, clinical factors such as symptom severity, duration of problems, and daily function were not significant predictors of dropout. The overall dropout rate of 25.3% was in accordance with previous rates reported in the literature, notably at the lower end.

Our results showed that clients under the age of 30 had a higher risk of dropout, which is in accordance with former research [12, 22, 39, 48, 52]. Fenger et al. [12] explain the link between younger age and dropout by more profound adherence problems and challenges with engagement. Less developed cognitive abilities might reduce the capacity for self-reflection and psychological mindedness, which are beneficial in therapy [1, 34]. Young adulthood is also characterized by less stable social and personal situations [12]. An unpredictable schedule might increase the chance of not showing up. Furthermore, group affiliation becomes more important for self-evaluation. Therefore, feeling different and experiencing stigma can become a barrier to completing therapy. On the contrary, knowledge and access to mental health treatment is more available today compared to previous generations. This might lower the threshold for seeking treatment, while simultaneously lowering the threshold for dropping out when experiencing that treatment does not work. Finally, the described characteristics of younger clients might make it more difficult to establish a good therapeutic alliance, which in itself is a predictor of dropout [22].

In accordance with previous literature, we found that level of education influenced the likelihood of dropout [12, 52]. Lower levels of education might be linked to dropout on the basis of cognitive abilities, difficulties structuring life, and a low feeling of mastery [7, 12]. Thereby, it might not be education itself that is decisive, but rather the ability to acquire new knowledge. Sharf et al. [45] found in their meta-analysis that the association between therapeutic alliance and dropout was stronger under the condition of lower levels of education. This can be because educated clients are more similar to their therapists, potentially facilitating a good therapeutic alliance [45]. Furthermore, lower levels of education might have secondary consequences such as lower income and poorer working conditions, which can increase perceived life stress. When struggling to meet basic needs, it can be difficult to find time for or remember appointments. Several instances of not showing up in a row, regardless of the cause, might result in treatment rejection, and defining the client as a dropout.

We found in our model that the strength of the relationship between lower levels of education and dropout was somewhat reduced when adding age to the model. The relationship between lower levels of education and dropout might to some extent be explained by age, as more people of younger age are yet to have an education degree.

Poor social support was found to predict dropout, in line with former research [18, 44]. Social support has been identified as an enabling factor for a person's use of healthcare services [1]. Conversely, poor social support can give rise to feeling alone with one's problems and make it more challenging to maintain motivation throughout treatment. These findings underline that the client's ability to show up to treatment is influenced by factors outside the therapist's office.

Another hypothesis is that poor social support can be maintained by the client’s relational patterns. These patterns might be transferred to the therapeutic alliance. Personality traits such as avoidance, hostility, aggressiveness, and low psychological mindedness have been found to negatively influence the therapeutic alliance [1, 22]. A poor therapeutic alliance can subsequently be linked to dropout.

We found that being a student was a statistically significant variable for work status in our first logistic regression analysis. However, when including this variable in the multivariate regression, the significance attenuated. Exploring this further, we found that the relationship between being a student and dropout was reduced when adding age to the model. This is probably due to the fact that students tend to be younger. Based on these results, the possible explanations of the association between dropout and age also applies to the association between being a student and dropout.

We did find an association between dropout and unemployment in this study, which was in line with findings from previous studies [12, 14, 42]. Unemployed clients tend to have lower income levels, which may explain part of the association with drop-out when the therapy is not free of charge. This is not the case for PMHC though and other explanations are therefore warranted. Unemployed clients may on average be less resourceful and may therefore find it more challenging to put in sufficient effort to gain benefit from therapy. It may also be the case that these clients have lower expectations that their own efforts will yield results in therapy. For therapists, it would be worthwhile to be aware of these issues and address them early in therapy.

We did not find an association between dropout and the remaining sociodemographic factors such as immigrant background and sex. Previous literature has provided somewhat mixed results on these predictors. Furthermore, we did not find any effects for the clinical client variables, contrary to previous research. The lack of association between dropout and high symptom severity might be because our sample is drawn from a primary care service. This entails that the target group was clients with mild to moderate depression and/or anxiety. People with more complex and severe problems were referred to specialized health care. Therefore, the clients in our sample generally had a lower and homogenous symptom severity.

The lack of association between lower symptom severity and dropout might be explained by the nature of the service and the definition of dropout in PMHC. Unlike some other services, PMHC does not follow a given protocol including a set minimum or maximum of sessions for the client. The number of sessions are rather determined by the clients´ needs. Furthermore, the definition of dropout was in our study based on the therapist's evaluation of the treatment goal.

Our study has several strengths. When collecting the data, questionnaires and measurements were used to cover a wide range of baseline information regarding the clients. With limited missing data (< 3%) and relatively large sample size (N = 526), we were able to make thorough analyses with relevant baseline factors identified through the literature.

Our instruments were standardized and validated with acknowledged cut-offs for the central measures of anxiety (GAD-7) and depression (PHQ-9). The only exception was the OSSS-3 with a Cronbach’s alpha of 0.58. This might imply that the instrument lacked some consistency across questions in this sample, and potentially underestimated the association between social support and dropout. The various instruments used in this study are applied within the PMHC service, which makes it possible to compare results from PMHC within and across countries to other similar services such as IAPT. This contributes to a strengthened external validity and generalizability of our results.

When performing the analysis, we included therapists and municipalities as fixed effects. This way, we excluded variations that could be attributed to these factors and thereby reduced the potential for Type I error.

The results from this study should be considered in the context of some limitations. Firstly, our study only investigates one group of factors, namely client factors. This was due to the nature of our dataset. Client factors alone can not explain dropout, which is rather a complex interplay between the client, therapist, therapeutic alliance, and service [52]. Our results should therefore be supplemented by findings from other groups of factors.

Secondly, our study had limited data on dropout from guided self-help, only used by 1% of our sample. This is a limitation, as guided self-help is an important component of the mixed care model [28]. Thus, this study can not provide solid information about dropout from this treatment modality.

A weakness concerning our understanding of dropout is that we only had the perspective of the therapist at hand. The clients´ experience might have differed from what the therapists reported, thereby weakening the reliability [52].

Therapists should know that there is an increased risk of dropout among clients of younger age, being unemployed, having lower levels of education or low social support. When therapists identify these predictors, it should encourage them to be more flexible and adaptive to the client. This especially as limited flexibility and individual adjustment from the therapist represents a main reason for dropout [30]. However, therapists often struggle to identify when their interventions are not working, constituting a barrier to being sensitive and flexible [50]. Further on, dropout rates vary substantially between therapists [3, 42].

A way to prevent dropout due to such variations is to seek the client's feedback through formal outcome monitoring systems [17, 50, 54], such as Feedback Informed Treatment (FIT). FIT consists of rating scales monitoring both the client’s improvement and the therapeutic alliance. The tool has been shown to be cost-effective in the context of IAPT [9]. However, FIT only achieves its purpose if it is used correctly [20]. Therefore, implementation of FIT in PMHC must include thorough training on how to utilize results to adapt to the client. To maintain such an implementation over time, it is crucial to establish a feedback culture which should be a leader and service responsibility.

Providing a time perspective of therapy has been found to reduce the risk of dropout [2, 34, 36, 54]. However, giving an absolute time perspective is difficult as sessions in PMHC are based on continuous evaluations of the clients' needs. Therapists can, however, provide an estimate of how many sessions the client can expect, or agree on an “evaluation session” after three appointments.

Forgetting is often mentioned as a reason for not showing up [3, 17]. In our sample 36.1% of the dropout group were terminated because they were unreachable. Pennington and Hudson [37] found lower dropout rates among clients invited by telephone and with a text message reminder, compared to clients invited only by letter. Adapting communication channels to remind clients might engage young people at risk of dropout. However, service routines should also address those who are about to drop out or recently dropped out. Routines for when no-show will lead to discharge are often vague and practices vary between therapists and services [3].

A recent process evaluation of PMHC stresses that the focus on clients’ socioeconomic challenges often has been neglected [28]. It could be argued that socioeconomic challenges should be emphasized more in therapy as a measure to prevent dropout. This aligns well with PMHC’s secondary goal of enhancing work participation. Some clients might even be more in need of work training and social interventions than psychological interventions and should be guided to another service [12].

It is important to remember that dropout is not exclusively negative [29]. Some people leave treatment because they experience improvement already in the first couple of sessions [4, 17]. Others might have low symptom severity to begin with and so they are more ambivalent about treatment [54]. Young people might be overrepresented in this group, as they have a lower threshold for talking about mental health and approaching therapy. Dropout due to early improvement might be especially relevant for primary care services, which aim to be easily accessible and reach people at an early stage. A natural side effect of this strategy is that dropout also becomes an accessible option. Dropout due to early improvement does, however, not guarantee a long-time improvement [4, 34, 54]. Therefore, we need to differentiate problematic cases of dropout from non-problematic cases. It is not realistic to expect dropout-free services. We should rather discuss what kind of dropout is tolerable.