Introduction

Depression is a common mental disorder typically chronic, disabling and frequently comorbid that affects more than 260 million people every year [1] and causes considerable personal suffering and has great economic costs for Western societies [2]. Depression was expected to be the leading cause of disability in 2030 [3] but, as early as 2021, it was declared the leading cause of disability worldwide and a major contributor to the overall global burden of disease according to the World Health Organization [4].

Although pharmacological treatment of depressive disorders has shown a considerable efficacy, patients do not always take their medication as instructed. When talking about the behaviors of patients in taking medication, adherence and persistence need to be examined.

Medication adherence can be defined as the process to which a patient acts within the prescribed range and dose of a dosage regimen, described by three quantifiable phases: 1) initiation, when patient takes the first dose; 2) implementation, defined as the process to which a patient's actual dosing corresponds to the prescribed dosing regimen; and 3) discontinuation, when the next dose to be taken is omitted and no more doses are taken thereafter [5]. Persistence refers to the duration of time from initiation to discontinuation of therapy [5]. In this sense, non-adherence to appropriately prescribed medicines remains a major challenge in current clinical psychiatric practice that compromises the efficacy of available treatments and interferes with patient recovery [6].

The impact of non-adherence to antidepressants increases the likelihood of relapse and/or recurrence, emergency department visits, and hospitalization rates; increases symptom severity and decreases treatment response and remission rates [7]. Non-adherence subsequently translates to an increase in medical and total healthcare utilization [7]. Available literature shows primary medication adherence (when a patient properly fills the first prescription for a new medication) rates ranging between 74 and 82% [8, 9], but unfortunately, approximately 50% of patients prematurely discontinue therapy [10, 11].

Socio-demographic variables, such as age, positive attitudes to prescribed medication and previous experiences were found to be factors predicting better adherence in patients with depressive disorders. Conversely, experience of side effects, dissatisfaction with treatment and a poor patient–professional relationship were found to be associated with poorer adherence [12].

Several interventions have been designed to improve medication adherence. Some evidence suggests that multifaceted interventions targeting the patient, physician and structural aspects of care are more effective than single-component interventions [13,14,15]. However, it is considered that intervention strategies should be designed to address the specific factors associated with non-adherence to psychotropic medication for each psychiatric disorder [16, 17]. Moreover, interventions rarely target the adherence phase but recruit patients independently of their treatment journey that is, at the beginning (initiation), during implementation or while discontinuing (persistence) [18].

The aims of the present study are to identify, critically assess and synthesize the available scientific evidence on the effectiveness of interventions aimed at improving adherence (initiation and the implementation phase) to medication among adults with depressive disorders.

Material and methods

A systematic review and meta-analysis were performed according to the Cochrane Handbook [19] and reported in accordance to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [20]. The protocol of the present review was registered in Prospero (CRD42017065723).

Information sources and search strategy

The following electronic databases were searched (January 2022): Medline (OVID interface), EMBASE (Elsevier interface), CENTRAL (The Cochrane Library interface), PsycINFO (EBSCO interface), SCI-EXPANDED (Web of Science interface) and SSCI (Web of Science interface). The search strategy was initially developed in Medline, using a combination of controlled vocabulary and free text terms and was then adapted for each of the other databases. Search terms included the following: depressive disorder, medication and adherence. Searches were limited to the English and Spanish languages and no date restriction was imposed. The full search strategy is available in Supplementary Material (see Supplementary Table 1). The reference lists of all included papers were also examined to identify possible additional studies meeting selection criteria.

Selection criteria

Studies were eligible for inclusion if they fulfilled the following criteria: 1) randomized controlled trials (RCTs) or non-randomized controlled trials (nRCTs), with allocation of both individuals and clusters; 2) any type of intervention (whether they were psychotherapeutic, educational interventions or other clinical intervention such as monitoring and adjustment of pharmacological treatment) aimed at increasing adherence (initiation and/or implementation phase) to anti-depressive medication administered to adults (18–65 years) with a diagnosis of depressive disorder. If a study addressed a heterogeneous group of patients, the study was included as long as the results for patients meeting the inclusion criteria were reported separately or they accounted for more than 80% of the target population. If the phase of adherence was not specified according to the taxonomy of Vrijens et al. [5], the reviewers determined the phase in which the evaluation was carried out based on the characteristics described in the study (adherence measurement method and moment); 3) usual care or alternative intervention as comparison group; 4) studies assessing initiation or implementation phase divided into three temporary spaces: short-term (closest to 3 months), medium-term (closest to 6 months) or long-term (closest to 12 months) adherence to prescribed medication; 5) studies published in English or Spanish. Exclusion criteria included: 1) studies examining patients with bipolar depression or schizoaffective disorder, and 2) studies with fewer than 10 study participants.

Study selection process

Two reviewers addressed eligibility independently and in duplicate. Firstly, the title and abstract of references identified in the electronic search were screened. Secondly, the full text of the studies that appeared to fulfil the pre-specified selection criteria was read and evaluated for inclusion. Disagreements between reviewers were resolved through discussion with the research team until consensus was reached.

Data collection process

A data extraction form was prepared by the authors, pilot tested on two studies and refined accordingly. One reviewer extracted the following data from the included studies: identification of the article (author, date of publication, country), study objective and methodology (design, context, duration), details of participants (selection criteria and demographics), interventions (type, modality and number of sessions), comparators and outcome (adherence definition, measurement method and value), and finally results. A second reviewer subsequently verified the extracted data. When any required information was missing or unclear in a paper, an effort was made to contact the corresponding author.

Risk of bias assessment

Two reviewers independently and in duplicate assessed risk of bias of included studies using the Cochrane Risk of Bias tools for RCT (RoB 2.0) [21] with the additional guidance for cluster-RCT [22] and nRCT (ROBINS-I) [23]. Discrepancies of judgments between the reviews were discussed by the research team until consensus was reached.

Assessment of publication bias

According to the recommendations of the Cochrane Collaboration [19], the presence of publication bias was assessed considering the size and sponsorship of the included studies, and by constructing a funnel plot and computing the Egger’s regression test using metafunnel and metabias commands in STATA version 14, respectively.

Analysis and synthesis of results

Meta-analyses and forest plots were performed for the adherence rate using the metan commands in STATA version 14. Effects of interventions were estimated as odd ratios (OR), with 95% confidence intervals (CI). Heterogeneity was assessed using the I2 statistic. When there was heterogeneity (I2 ≥ 25%), meta-analyses were performed using a random-effects model using the method of DerSimonian and Laird and taking the estimate of heterogeneity from the Mantel–Haenszel model. When there was neither clinical nor statistical heterogeneity, a fixed-effect model was used [24].

Several sources of heterogeneity relating to the characteristics of the study population and the interventions were anticipated. Predictive variables included age, gender, diagnoses, type of intervention, providers of the intervention (multidisciplinary vs. non-multidisciplinary team), modality of intervention (face-to-face vs. telephone, mails and/or website) and number of sessions. When reported in most studies, the effect of these study-level variables on the effectiveness closest to six months after intervention using subgroup analyses (diagnoses, type of intervention, providers of intervention and modality of intervention) and meta-regression techniques (age, gender, and number of sessions) were explored using the metareg command.

Sensitivity analyses were conducted to assess the stability of the effects of excluding certain types of studies (n-RCT).

Cumulative meta-analysis was used to evaluate the sufficiency and stability over time of the effects of interventions aimed at increasing adherence to anti-depressive medication. Studies were sequentially added by year of publication to a random- effects model using the metacum user-written command.

Results

Out of a total of 2,839 initially identified references after eliminating duplicates, 40 studies were selected after full-text screening (Fig. 1). The manual search provided six additional studies, thus, 46 studies (published in 51 papers) were finally eligible for inclusion according to the pre-established selection criteria [25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75].

Fig. 1
figure 1

Flow diagram of the selection process of studies

Full size image

Characteristics of included studies

The 46 included trials were published in English between 1976 and 2021 (Table 1). Thirty-four are individual-RCT [25, 29,30,31,32,33,34,35,36, 40, 42,43,44, 46, 48,49,50,51,52, 55,56,57,58,59,60,61, 64,65,66,67, 70, 71, 74, 75], seven are cluster-RCT [26, 38, 41, 52, 53, 63, 72], four are individual-nRCT [28, 39, 45, 47], and one is cluster-nRCT [27]. The duration of reported follow-up ranged from 4 to 76 weeks (median 32 weeks). Seven studies specified incentive payments to patients [27, 29, 38, 39, 46, 55, 61] and 43 of them were carried out in outpatient [25, 26, 28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43, 46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74].

Table 1 Main characteristics of included studies
Full size table

Study size ranged from 19 to 12,919 participants, with a mean average of 526 per study. In the 46 studies, a total of 31,832 participants were recruited and 24,324 were finally assigned to intervention (RCT: 7,608; cluster-RCT: 3,470; nRCT: 13,147; cluster-nRCT: 99). The mean age of participants was 42.40 years (SD: 15.66) and 65.05% of them were female. Approximately 10% were lost in the follow-up, thus 2,404 patients completed the studies.

Most of the studies enrolled patients with depression at different levels of severity. However, five studies required a combination of major depressive disorder with panic disorder, social phobia or generalized anxiety disorder, or anxiety [34, 44, 45, 52, 65, 66].

All the studies assessed individual interventions and used usual care as comparator. In general, the number of sessions or contacts of the interventions ranged from 1 to 20. A total of 10 studies assessed the effects of the Collaborative Care Model (CCM) consisting of the following four elements of collaborative care: 1) a multi-professional approach to patient care; 2) a structured management plan, included either or both pharmacological and non-pharmacological interventions; 3) scheduled patient follow-ups to provide specific interventions, facilitate treatment adherence, or monitor symptoms or adverse effects; and 4) enhanced inter-professional communication. Five studies assessed the effects of interventions with only an educational focus while eight studies evaluated the effects of education and support, three of them used the RHYTHMS programme, a patient education programme which mails information directly to patients being treated with antidepressant medicines in a time-phased manner. Education was also added to Cognitive Behavioural Therapy (CBT), CBT and motivational interview, coaching, monitoring and psychiatric consultation. Psychotherapy was another type of included intervention; in particular, six studies used CBT, one study included short psychodynamic supportive psychotherapy and one study included interpersonal psychotherapy. Other types of interventions were shared decision-making, support, counselling, the use of medication reminder applications for mobile phones, Enhanced Care and Treatment Initiation and Participation, an intervention aimed at modifying factors such as psychological barriers, concerns about treatment, fear of antidepressants and misconceptions of depression treatment.

Intervention modalities included face-to-face meetings alone (18 studies) or in combination with telephone conversations (3 studies), leaflets (2 study), videotapes (2 studies), mails (1 study) or website. Eight studies used telephone-conversations and two studies used the same intervention in combination with mails and one study combined the same intervention with letters. Moreover, leaflets were used in three of the studies, while consultation of websites was included in two studies. Another intervention modality was the use of a smartphone (2 studies).

The intervention providers varied among studies: multidisciplinary teams (16 studies), primary care professionals -general practitioners, clinicians or internal medicine doctors- (8 studies), pharmacists (8 studies); psychiatrists, psychologists or therapists (5 studies), nurses (2 studies), research assistant (1 study), and health worker (1 study). In the remaining studies, the providers were required to deliver intervention (2 studies) or not reported (1 study).

All patients in the included studies were in the implementation phase of the adherence. Twenty-five studies provided short-term (ranged from 4 to 16 weeks), 22 studies provided mid-term (ranged from 20 to 36 weeks), and seven studies provided long-term (ranged from 48 to 76 weeks) outcomes. Both self-report and direct measures were used for assessing adherence. Approaches for subjectively assessed adherence included questionnaires, diaries and interviews, and approaches for objectively assessed adherence included electronic measures, pill count and plasma drug concentration.

Risk of bias in the included studies

Out of the 41 RCTs identified, three were classified as having low risk of bias in all RoB 2.0 domains [34, 57, 70] (Table 2). In the remaining RCTs, the most common methodological concerns involved bias arising from the randomization generation and allocation concealment process (3 RCTs at high RoB) and bias in measurement of the outcome (6 at high RoB).

Table 2 Risk of bias of included RCTs
Full size table

For the five n-RTCs identified, risk of bias was generally low-to-moderate across all of them, all presenting risk of bias in at least three domains (Table 3).

Table 3 Risk of bias of included nRCTs
Full size table

Publication bias

No evidence of publication bias was found according to the funnel plot of the observed effect (Fig. 2) and the Egger’s regression test (P = 0.50).

Fig. 2
figure 2

Funnel plot – Potential publication bias

Full size image

Synthesis of results

Results on adherence of the selected studies are available in the Supplementary Material (see Supplementary Table 2). Results of all meta-analyses and subgroup analysis are also available in the Supplementary Material (see Supplementary Tables 3 and 4).

Interventions aimed at improving the implementation phase of medication adherence in adults with depressive disorders had a positive effect on adherence outcome at 6 months after intervention compared with usual care (Odd ratio [OR] 1.33, 95% confidence interval [95% CI]: 1.09 to 1.62; p < 0.01) (Fig. 3). As anticipated, there was a moderate level of heterogeneity between studies (I2 = 59.30%).

Fig. 3
figure 3

Forest plots for effect of intervention on adherence rate. Note: A at 6 months; B at 6 months; C at 12 months

Full size image

In the patients of these studies, the overall trend for clinical improvement was observed to emerge at 3 months after intervention (OR 1.62, 95% CI: 1.25 to 2.10; p < 0.01) but the effect was attenuated at 12 months after intervention (OR 1.25, 95% CI: 1.02 to 1.53; I2 = 4.10%; p = 0.40) (Fig. 3). Substantial between-study heterogeneity was also found at 3 months (I2 = 66.10%).

Causes of heterogeneity

Sufficient study-level data were available from 35 of the studies for the effect of the predictor variables to be entered into a subgroup or meta-regression analysis. Results of subgroup analysis and meta-regression are available in the Supplementary Material (see Supplementary Tables 3 and 4, respectively).

Diagnosis

Interventions aimed at improving adherence to medication when addressed to adults with depression at different levels of severity were associated with a significantly increased effect size (OR Major depressive disorder or dysthymic disorder and anxiety studies 2.77, 95% CI: 1.74 to 4.42; p < 0.01; OR High risk for recurrent depression 1.69, 95% CI: 1.13 to 2.54; p = 0.01; OR Major depressive disorder or dysthymic disorder 1.32, 95% CI: 1.08 to 1.61; p < 0.01; I2 = 35.80%). However, pooled effect sizes of studies on patients with depressive symptoms (OR, 2.50, 95% CI: 0.86 to 7.31; p = 0.29; I2 = NA%), depressive episode (OR, 0.88, 95% CI: 0.69 to 1.12; p = 0.29; I2 = 0%), and major depressive disorder with or without dysthymic disorder (OR, 0.68, 95% CI: 0.30 to 1.50; p = 0.29; I2 = 70.70%) were not statistically significant.

Type of intervention

In the case of CCM interventions, the pooled result showed a significant increase in adherence (OR 1.88, 95% CI: 1.40 to 2.54; p < 0.27; I2 = 23.00%) compared to the control group. However, statistically significant differences were not found for other specific forms of intervention (see Supplementary Table 3).

Providers of the intervention

A multi-professional approach to patient care involving at least one primary care provider and another health professional (e.g., nurse, psychologist, psychiatrist or pharmacist) was associated with an increased effect size (OR 1.73, 95% CI: 1.21 to 2.46; I2 = 53.70%). A non-multidisciplinary approach was not statistically significant (OR 1.15, 95% CI: 0.94 to 1.40; I2 = 42.90%).

Modality of intervention delivery

Effect sizes did not significantly differ by the modality of intervention delivery used (see Supplementary Table 3).

Other sources of heterogeneity

The number of intervention sessions was related to adherence (β, -0.08; 95% CI: -0.14 to -0.02). However, none of the other sources of heterogeneity investigated (age and gender of participants) had an effect.

Cumulative meta-analysis of outcome at 6 months

When we assess interventions aimed at improving adherence to medication over time (Fig. 4), it is unclear whether earlier trials meeting the inclusion criteria demonstrated a high degree of heterogeneity or a high percentage of negative results. There is a sufficient body of evidence to demonstrate a reliable, consistent and statistically significant benefit of interventions aimed at improving adherence to medication over usual care. In general, the overall effect size has remained relatively stable within an effect size between OR 1.17 and 1.56.

Fig. 4
figure 4

Cumulative meta-analysis of studies ordered by year of publication

Full size image

Discussion

Our findings support and confirm the notion that interventions aimed at improving adherence to medication among adults with depressive disorders are effective in improving outcomes in implementation phase of adherence in the studied patients, when these were analysed at 3 and 6 months after the intervention. The evidence, when given using cumulative meta-analysis, shows that further trials are unlikely to overturn this positive result. However, it is possible to appreciate a small decline in effect size over time.

The evidence shows that collaborative care is effective in improving adherence. In this respect, a multi-professional approach to patient care was more effective than primary or mental healthcare teams. This finding supports the idea that collaborative care might not only be clinically effective for symptom management in adults with depressive disorders [76, 77], but could also have a major effect on improving adherence to treatment [7]. This is in line with previous literature and suggests that multifaceted interventions targeting all dimensions that affect medication adherence problems, i.e., the patient, the healthcare provider and the health care delivery system, are more effective than single-component interventions to improve medication adherence [14, 15]. In fact, this positive effect of multicomponent interventions has also been observed in other psychiatric disorders [16, 17] and non-psychiatric pathologies [78]. Moreover, the number of intervention sessions was negatively related to adherence. A similar result has been observed in other studies of behaviour changes [79, 80]. Although the optimal number of intervention sessions is not clear, this a priori surprising result would support the usefulness of brief interventions or therapies to improve treatment adherence, however, it needs to be confirmed with more research.

Nevertheless, subgroup analyses indicate how other characteristics of the intervention may not help to enhance adherence. The modality of intervention and the provider profile were unrelated to effect size. Effect sizes also did not differ significantly by the modality of intervention delivery used (face-to-face vs. telephone, mails and/or website). Computer support systems, mobile technologies, web-based e-mail or telephone-based assistance can be used for improving adherence to medication [81, 82]. In this regard, these interventions may be available across different geographic areas and in different clinical settings [83].

Generally, it might be expected that patients with severe symptoms would have different treatment and support needs, and thus may profit from this type of interventions compared to patients with moderate or mild symptoms. However, the findings here do not show a clearly relationship between the severity of disease and adherence. Several interventions are effective in improving adherence outcomes among patients diagnosed with depression and anxiety at the same time. Although effectiveness is also demonstrated in the cases of patients at high risk of recurrent depression and in patients with major depressive disorder or dysthymic disorder, the results do not present such high values. Other patient characteristics such as age or gender were unconnected to adherence outcome.

The main limitation of the present review is the methodological differences between studies, mainly the diversity of both intervention procedures and severity and diagnosis of depressive disorder of participants, as well as the absence of an adequate psychopathological evaluation of the patients included in the studies. Interventions aimed at improving medication adherence among adults with emotional disorders have been designed with varying levels of intensity. Consequently, the review here found significant between-study heterogeneity. Subgroup and meta-regression analyses have been used to explore some of the issues related to the diversity of interventions (i.e.: type of intervention and providers) and patients’ characteristics (i.e.: severity of depression) that may influence the adherence result. Although, up to 770 determinants of adherence have been described in previous literature [84], only a few could be explored in this review. Although the prescribed antidepressant treatment has been shown to be a predictor of adherence [85, 86], most included studies did not report the specific antidepressant medicines that patients receive (Table 1). Moreover, there were studies that did not specify the patient´s phase of adherence, some of them because they were published before the publication of Vrijens et al. taxonomy [5]. However, after the evaluation based on the characteristics of the studies, we have determined that all patients in the included studies were in the implementation phase of the adherence. Finally, the exclusive reliance on English-language studies may not represent all the evidence. For this reason, we have also considered studies published in Spanish, however, limiting the systematic review to studies written in English and Spanish, which could introduce a language bias.

Despite all these limitations, our comprehensive systematic review provides an updated assessment of the effectiveness of different types of interventions aimed at improving medication adherence among adults with emotional disorders, supported by meta-analyses, using cumulative meta-analysis, assessing risk of bias of included studies, exploring important sources of heterogeneity and following rigorous and transparent methods compared to the previous systematic review [15].

The systematic review reported here shows that interventions aimed at improving short and medium-term adherence to medication among adults with depressive disorders are effective. Compared to short and medium-term adherence outcome, the available evidence on the effectiveness of long-term adherence is insufficient and supports the need for further research efforts.