Interventions to improve children's access to mental health care: a systematic review and meta-analysis

Aims Mental disorders in children are a significant and growing cause of morbidity worldwide. Although interventions to help overcome barriers along the pathway to accessing health care for children with mental disorders exist, there is no overview of randomised controlled trials (RCTs) on these interventions as yet. This study aimed to systematically identify RCTs of interventions to improve access to mental health care for children and synthesise them using a conceptual framework of access to health care. Methods This systematic review was performed following a predefined protocol registered with PROSPERO (ID: CRD42018081714). We searched the databases MEDLINE, EMBASE, PsycINFO and CENTRAL for RCTs up to 15 May 2019 using terms related to the concepts ‘young people,’ ‘mental disorders’ and ‘help-seeking interventions’ and scanned reference lists from relevant studies. Two reviewers independently screened all identified articles in a two-stage process, extracted results on outcomes of interest (knowledge, attitudes, intentions, help-seeking, accessing care, mental health outcomes and satisfaction), assessed the risk of bias and conducted meta-analyses where deemed appropriate. Results After screening 5641 identified articles, 34 RCTs were eligible for inclusion. Eighty per cent of universal school-based interventions measuring knowledge (n = 5) and 67% measuring attitudes (n = 6) reported significantly better results compared with controls on those outcomes, whereas 20% measuring access to care (n = 5) and none measuring mental health outcomes (n = 7) did. In contrast, 71% of interventions targeting at-risk individuals (n = 21) reported better access to care compared with controls, while just 33% (n = 6) did for mental health outcomes. For satisfaction with care, this proportion was 80% (n = 5). Meta-analyses of interventions measuring initial appointment attendance yielded combined odds ratios of 3.11 (2.07–4.67) for appointment reminder interventions and 3.51 (2.02–6.11) for treatment engagement interventions. The outcomes for universal school-based interventions were heterogeneous and could not be summarised quantitatively through meta-analysis. Conclusions To have a population-level effect on improving children's access to mental health care, two-stage interventions that identify those in need and then engage them in the health-care system may be necessary. We need more evidence on interventions to target contextual factors such as affordability and infrastructural barriers.


Introduction
Mental disorders are one of the most significant causes of disability-adjusted life-years worldwide, and they continue to grow in importance as a major contributor to the global burden of disease (GBD 2015 DALYs and HALE Collaborators, 2016). Because mental disorders usually first occur early in life  and are characterised by recurrent episodes and symptoms that strongly affect work capacity (Simon et al., 2001), they have a significant impact on public health and society.
Childhood and adolescence are particularly critical periods for the identification and treatment of mental disorders. At 45% of the overall burden of disease in [15][16][17][18][19] year-olds, mental health issues are the leading cause of disability in adolescents (The Lancet, 2017). In addition, young patients with a mental disorder have a lower probability of receiving treatment and a longer delay between disease onset and first treatment compared with adults (Christiana et al., 2000;Wang et al., 2005;Iza et al., 2013). Despite the magnitude and importance of mental health problems in childhood and adolescence, international studies have consistently revealed a treatment gap: estimates of the gap between those in need of mental health care and those who access it exceed 50% (Saxena et al., 2007). Levesque et al. (2013) define access to health care as 'the opportunity to reach and obtain appropriate health-care services in situations of perceived need for care' (Levesque et al., 2013).
They have proposed a comprehensive conceptual framework describing accessing health care as a series of steps beginning with the opportunity to perceive health-care needs that can lead to opportunities to seek health care, reach health-care services, utilise health-care services and ultimately have health-care needs fulfilled (Levesque et al., 2013). At each stage, supply-side dimensions of accessibility of services (e.g., approachability, availability or affordability) interact with demand-side abilities of persons (e.g., abilities to perceive, pay or engage) to determine access to health care (see Appendix 1) (Levesque et al., 2013). In other words, the care that is obtained depends on the interplay of characteristics of individuals, such as their socio-economic status or where they live, and those of services and the environment, such as how much services cost and where they are located. Potential barriers that could explain the treatment gap can be found at each transition from step to step along this pathway to accessing care (Levesque et al., 2013). Barriers to mental health help-seeking in young people include lack of knowledge about services and stigma about mental health problems (Gulliver et al., 2010). As an example of barriers on the supply side, paediatricians perceive a wide variety of organizational hindrances, including inadequate reimbursement and lack of time and space, and many feel they lack the training and confidence to treat mental disorders (Horwitz et al., 2007).
To close the treatment gap, interventions targeting one or more dimensions of accessibility of services and/or abilities of persons have been designed to address the barriers along the pathway to accessing care (e.g., screenings, health literacy promotion); however, there is little high-quality evidence on these interventions (National Institute for Health and Clinical Excellence, 2011). Moreover, systematic reviews conducted in the past on interventions to improve access to mental health care for children and adolescents have limited searches to specific types of interventions and disorders (Ingoldsby, Richardson et al., 2017). This study thus aimed to systematically identify randomised controlled trials (RCTs) of all interventions designed to improve access to mental health care for children along the entire pathway to accessing care, describe them using Levesque et al. (2013)'s conceptual framework of access to health care (Levesque et al., 2013) and conduct meta-analyses for intervention types with comparable outcomes.

Methods
The methods used for this systematic review are based on the Centre for Reviews and Dissemination's guidance for undertaking reviews in health care (Centre for Reviews and Dissemination, 2009), and our reporting follows the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) (Moher et al., 2009). A PRISMA checklist can be found in Appendix 2. We registered our systematic review protocol with the International Prospective Register of Systematic Reviews (PROSPERO, ID: CRD42018081714).

Types of participants
We included interventions designed for children and adolescents <19 years old, both from the general population and vulnerable groups. If the age range exceeded 18 years old, the intervention was only included if more than 50% of the ages considered were under 19. Interventions that addressed the following disorders from the International Classification of Diseases, 10th Revision (ICD-10) (World Health Organization, 1992) as well as suicidal ideation were considered: F10-F59 and F90-99 (all mental disorders except for mental disorders due to known physiological conditions, disorders of adult personality and behaviour, intellectual disabilities and pervasive and specific developmental disorders including autism spectrum disorders). We also included studies that targeted children with emotional or behavioural problems since children are not always given a specific diagnosis.

Types of interventions
Any intervention designed to improve access to mental health was included; thus, the intervention could target any one of the five supply-side dimensions or five demand-side abilities included within the conceptual framework. Examples of specific intervention targets are listed next to each dimension or ability in Appendix 1. For example, an intervention could change where services are offered or deliver services via the Internet (National Institute for Health and Clinical Excellence, 2011). The interventions could target the child or adolescent directly or others, including parents/caregivers, teachers, friends or health-care professionals (potential helpers).

Types of outcome measures
We defined outcomes using the conceptual framework and expanded upon these using conceptualisations from previous systematic reviews on help-seeking and treatment engagement interventions (Gulliver et al., 2012;Lindsey et al., 2014). Outcomes at all steps in the process of accessing health care were included in the review: knowledge about accessing mental health care, changed attitudes or beliefs about seeking care, intentions to seek care, help-seeking attempts to access health-care services (successful or not) or action taken by a potential helper, mental health outcomes and satisfaction with health-care services. For a study with outcomes on health measures and satisfaction with care to be included in the analysis, the study also had to measure access to care as an outcome. We excluded studies for which it was not possible to calculate any effect sizes.

Search methods for identification of studies
We performed the literature search on 15 May 2019 in the following electronic databases: MEDLINE, EMBASE, PsycINFO and the Cochrane Central Register of Controlled Trials (CENTRAL). The search strategy included terms relating to the concepts 'young people,' 'mental disorders' and 'help-seeking interventions.' The full search strategy can be found in Appendix 3. Publications not originally published in English were excluded from the search. We enhanced our search by scanning the reference lists of papers (both primary studies and reviews) that were identified by the database search. Duplicates were removed during the title and abstract screening.

Selection of studies and data extraction
Two reviewers (LW, DG) independently assessed the title and abstract of all identified papers, recorded their decision about whether the paper should be included for full-text assessment and discussed discrepant decisions until a consensus was reached. All papers deemed potentially eligible by the reviewers were included in the full-text assessment, in which the two reviewers decided on study inclusion based on the inclusion criteria and discussed any discrepant decisions until they reached a consensus.
The two reviewers independently extracted data on the following study characteristics: title, first author, year, country, study design, age range, intervention setting, condition in focus, sample size, response rate, intervention condition, control condition, length of intervention, evaluation time points, method of outcome assessment and results.

Assessment of risk of bias in included studies
Two reviewers (LW, DG) assessed the risk of bias of each article using the Cochrane Collaboration's tool for assessing risk of bias in randomised trials (Higgins et al., 2011) and discussed discrepant evaluations until they reached a consensus. Because our review included a large variety of interventions and outcomes, we could rarely assess the heterogeneity, imprecision and indirectness beyond a single or a few studies and therefore decided against using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach (Guyatt et al., 2011) to judge the overall quality of evidence. Instead, we used the risk of bias assessment as an indicator of the quality of evidence.

Data synthesis and measures of effect
We mapped the study results using the conceptual framework by Levesque  Articles included * For a study with outcomes on health measures and satisfaction with care to be included in the analysis, the study also had to include a measure about improved access to care. Epidemiology and Psychiatric Sciences 3 outcome (Fig. 3). For dichotomous outcomes, we extracted or calculated the odds ratio and 95% confidence interval, whereas, for continuous outcomes, we calculated the standardised mean difference and 95% confidence interval using Cohen's d with the package 'esc': Effect Size Computation for Meta Analysis in R (Lüdecke, 2017). For intervention types with comparable outcomes, we conducted meta-analyses using the inverse variance method. We calculated a fixed-effects model if I 2 was <30% and both fixed and random effects models if I 2 was >30% using the package 'meta': General Package for Meta-Analysis in R (Schwarzer, 2007).

Results of the search and excluded studies
The electronic search yielded 5688 articles, and an additional 43 records were identified through hand searching. Of these 5731 records, 5641 unique studies remained after duplicates were removed. A total of 71 articles were considered eligible for fulltext screening following the title and abstract screening. During the full-text screening process, 37 articles were excluded; the full list of articles excluded along with reason for exclusion can be found in Appendix 4. The remaining 34 articles were included in the systematic review. For an overview of the search and screening process, please see the study flow diagram (Fig. 1).

Included studies
A summary of the characteristics of the 34 RCTs identified through the two-stage screening process can be found in Appendix 5. These studies fell into two main categories: (1) universal school-based interventions targeting the general population (13 studies) and (2) interventions to engage at-risk individuals who had already been identified by the health-care system (21 studies). The vast majority of these studies were conducted in the USA (22 studies); the rest were conducted in Australia (five studies), UK (three studies), Canada (two studies), Portugal (one study) and Israel (one study).
The interventions in the first study category were designed to improve outcomes for general mental health problems, mental distress, suicide, depression and attention-deficit hyperactivity disorder. All studies in the second category took place in healthcare settings (e.g., primary care, emergency department, mental health agency) and targeted general mental health problems, behavioural health problems, suicide, depression, substance abuse and conduct disorder. Interventions designed to help younger children access care tended to be addressed towards caretakers, whereas interventions targeting older age groups tended to address the adolescent directly. Figure 2 provides an overview of the step or steps along the pathway to accessing care that each intervention targeted. Interventions within the first category exclusively targeted service providers' approachability (i.e., service providers making their existence known to individuals) and the abilities of individuals to perceive a need for and to seek care. These interventions included educational curricula, live or virtual contact with a mentally ill person, screenings and helper training programs. The vast majority of engagement interventions from the second category mostly targeted service providers' appropriateness or individuals' ability to engage. Forty-eight per cent (10/21) of these interventions consisted of programs to engage and motivate patients or to improve service providers' communication skills (henceforth called treatment engagement interventions), while 24% (5/21) involved using a telephone or letter reminder mechanism to improve first appointment attendance (henceforth called appointment reminder interventions). Just one intervention was infrastructural in nature and involved providing onsite mental health services for primary care patients. None of the identified studies targeted the acceptability (cultural and social factors that make it possible for individuals to accept services) of service providers, affordability of care, or individuals' personal ability to reach (e.g., their personal mobility or support from their social network) or pay for care.

Risk of bias in included studies
The results of the risk of bias assessment can be found in Table 1.
The randomisation procedure was not described in half of the studies, and two studies described a non-random sequence generation procedure. Details on allocation concealment were only provided in 21% of the studies. In all studies except for one, it was unclear whether a lack of blinding of participants and personnel would influence the outcome. However, we judged that a lack of blinding of outcome assessment would not have an impact on the outcome since most outcomes were evaluated either by questionnaire or service use records. All but one study had a low risk of bias for incomplete outcome data. Three studies did not report all outcomes and three studies had other sources of potential bias. Although we did not formally grade the quality of evidence using the GRADE approach (Guyatt et al., 2011), we considered the criteria heterogeneity, risk of bias and precision where appropriate when reporting the effects of interventions below. Figure 3 provides a graphical overview of which outcomes were measured by which studies and whether or not the interventions had a significant effect on the outcome measures. The full report of intervention effects can be found in Table 2, and details on how outcomes were defined in each study can be found in Appendix 6.

Effects of interventions
Among the studies on universal school-based interventions targeting the general population, 80% (4/5) of those that assessed knowledge about accessing mental health care, 67% (4/6) of those that assessed attitudes or beliefs about seeking care, 22% (2/9) of those that assessed help-seeking or intentions, 20% (1/5) of those that assessed accessing care or taking action and none (0/7) of those that assessed mental health outcomes had a significant impact on the respective outcome. Thus, universal school-based interventions targeting individuals from the general population tended to have a significant impact on steps earlier on the pathway to accessing care, especially knowledge and attitudes, but not on later steps, such as actually accessing care or mental health outcomes. The risk of bias for studies on these interventions  ranged from low to high (see Table 1). The effect sizes ranged from −0.06 to 0.96 for knowledge about seeking health care, −0.02 to 2.56 for attitudes about seeking health care and −0.15 to 0.30 for intensions to seek health care or help others seek health care. Both odds ratios and effect sizes were calculated for the outcomes help-seeking, action taken and health outcomes. The odds ratios ranged from 0.96 to 21.64 for help-seeking, 0.99 to 11.34 for accessing care and 0.62 to 1.12 for health outcomes. The pattern of significant outcomes found for universal interventions differed from that observed for interventions targeting at-risk individuals who had already been identified by the healthcare system. Among studies on these interventions, all assessed accessing care (e.g., the proportion of study subjects who attended the first appointment or number of appointments attended) as an outcome, and 71% (15/21) of these interventions had a significant impact on that outcome. Eighty per cent (8/10) of studies on treatment engagement interventions (e.g., a family-based session to increase motivation during an emergency room or motivational telephone calls with trained staff) and 80% (4/5) of studies on appointment reminder interventions had a significant effect on accessing care. Just three interventions targeting at-risk individuals assessed outcomes that preceded accessing care. The effects on knowledge about accessing mental health care, attitudes or beliefs about seeking care and help-seeking were thus unclear due to the limited number of studies measuring these outcomes. Of the seven interventions that assessed the remaining two outcomes along the pathway to accessing care, 33% (2/6) of those that assessed mental health outcomes and 80% (4/5) of those that assessed satisfaction with care were significantly better as compared with controls on the respective outcome. Interventions targeting at-risk children who had already been identified by the health-care system therefore generally yielded more access to care and satisfaction with care as compared with controls, but not necessarily improved mental health outcomes. The risk of bias found for appointment reminder and treatment engagement interventions ranged from low to high (see Table 1). The most important outcome comparisons for these types of interventions are summarised in the meta-analyses below.

Meta-analyses
We conducted meta-analyses for two types of interventions that measured the same outcome (accessing care) using the binary measure first appointment attendance (yes/no): (1) appointment reminder interventions (five studies) and (2) treatment engagement interventions (10 studies). For the appointment reminder interventions, we only calculated a fixed-effects model since heterogeneity was low (I 2 = 0%). For the treatment engagement interventions, heterogeneity was substantial (I 2 = 70%), so we calculated fixed effects and random-effects models. Forest plots for each of these two types of interventions can be found in Fig. 4. The combined odds ratio of the appointment reminder interventions was 3.11 (2.07-4.67), and the combined odds ratio calculated using the random-effects model for the treatment engagement interventions was 3.51 (2.02-6.11). In other words, the odds of attending an initial appointment were 3.11 times higher for those who received an appointment reminder as compared with controls and 3.51 higher for those who participated in a treatment engagement intervention as compared with controls, indicating that overall, both types of interventions yielded significantly higher first appointment attendance in the target population as compared with controls.

Summary of main results
This systematic review identified 34 RCTs of interventions that fell into two main categories: universal school-based interventions targeting the general population and interventions to engage at-risk individuals who had already been identified by the healthcare system. Interventions in the first category generally yielded significantly better knowledge and attitudes about accessing care as compared with controls, but did not have an impact on actually accessing care or on mental health outcomes. Most interventions targeting at-risk children who had already been identified by the health-care system yielded significantly better access to care and satisfaction with care as compared with controls, but did not seem to have a significant impact on mental health outcomes. Meta-analyses of appointment reminder interventions and treatment engagement interventions measuring the outcome accessing care using the binary measure first appointment attendance found that both types of interventions yielded significantly more access to care as compared with controls. We did not identify studies that targeted the domains of acceptability and affordability or individuals' ability to reach or pay for care.

Comparison with other reviews
In our study, we used Levesque et al.'s conceptual framework of access to health care to design a systematic review of RCTs of interventions to improve access to mental health care for children (Levesque et al., 2013). This approach enabled us to provide a broad overview of RCTs of interventions that was not limited to particular types of interventions or disorders by structuring our findings along the entire pathway to accessing care. Our results on the effects of universal interventions targeting individuals  Epidemiology and Psychiatric Sciences from the general population were similar to those from a previous systematic review on help-seeking interventions for depression, anxiety and general psychological distress in adults that found that mental health literacy content significantly improved helpseeking attitudes, but did not have an effect on help-seeking behaviour (Gulliver et al., 2012). Another systematic review on interventions to promote help-seeking for mental health problems found that interventions increased formal help-seeking behaviours when targeting affected or at-risk people with mental disorders, but not the general population (Xu et al., 2018). This is the same pattern that we found for the outcome of accessing care.

Strengths and limitations
This systematic review provided an overview of interventions to improve access to mental health care along the entire pathway to accessing care using a conceptual framework, which allowed us to assess where evidence for effective interventions lies and where evidence is missing. However, taking a broad approach to the search necessitated restricting the search to the title field and English language only. We attempted to address this by hand searching reference lists and key articles. In addition, we did not include retention in treatment as an outcome in this review since we were interested in gaining access to treatment.

Implications for practice and research
Both the results of this systematic review and previous research have shown that interventions can improve knowledge and attitudes about mental disorders and their treatment (Lo et al., 2017); however, there is evidence that such interventions do not necessarily have an impact on health behaviours, such as helpseeking (Kelly and Barker, 2016; Laverack, 2017). Interventions to engage and motivate an at-risk population, on the other hand, have been shown to significantly change health behaviours (Ingoldsby, 2010). From a public health perspective, the problem with this finding is that existing interventions do not improve access to care for people in need from the general population, leaving a large treatment gap. In order to have a population-level effect on improving access to care, it may be necessary to introduce two-stage interventions, i.e., ones that first identify those in need from the general population and then engage them in the health-care system. In this systematic review, the only study in that targeted the general population, yet had a significant impact on accessing care was Husky et al. (2011), which tested systematic referral to mental health services using a brief mental health screening in a school setting . Additionally, five of the six studies targeting at-risk children that took place in a primary care setting used a screening procedure to identify these at-risk  Painter et al.
Three-hour educational curriculum on stigma, mental illness and barriers to help-seeking via presentation, class discussion and video and presentation and discussion with a mentally ill college student   , 2017); however, we identified just one infrastructural intervention that involved providing onsite mental health services for primary care patients . In light of this and the fact that our systematic review revealed gaps in the research on interventions to improve acceptability, affordability and individuals' ability to reach and pay, it seems that more research on interventions that address contextual factors such as these is warranted, although it may be difficult to test some of these interventions via RCT. In addition, targeting individual barriers in isolation, such as cost or insurance coverage, without addressing other barriers like accessibility, acceptability and availability may not improve service utilisation (So et al., 2019). It is possible that interventions that address multiple barriers simultaneously are more likely to have a population-level effect on improving children's access to mental health care, but this must be tested in future research.
Future studies on interventions to improve access to mental health care for young people should attempt to coordinate and standardise the outcomes assessed so that more quantitative comparison among studies via meta-analysis is possible. We especially need more studies testing the effects on mental health care outcomes since this is the ultimate purpose of improving access to care. In addition, longer follow-up periods are required to gain information about the longer-term effects of interventions to improve access to care (Salerno, 2016;Anderson et al., 2017). Finally, none of the studies identified in this systematic review took place in low-or middle-income countries. Due to a shortage of mental health professionals, the fact that detection rates of mental disorders are much lower in many of these countries, less developed infrastructure and potentially more stigma surrounding mental health disorders, different interventions than those that are effective in high-income countries may be required (Patel et al., 2013). More research is therefore needed to draw conclusions about improving access to care in these settings.

Conclusion
In order to bridge the existing treatment gap in mental health care for children, interventions that aim to improve knowledge and attitudes about mental health care in the general population are not sufficient. Instead, a two-stage approach that first identifies young people in need of care from the general population and then engages them in the health-care system should be tested in high quality studies. In addition, we need high quality research on the impact of interventions addressing contextual factors such as affordability and individuals' ability to reach care.
Data. The protocol for this systematic review has been registered with the International Prospective Register of Systematic Reviews (PROSPERO, ID: CRD42018081714). The publication details for the studies included in this systematic review have been included in the reference list.
Acknowledgements. Thank you to Sabine Klein for her help in designing and executing the systematic search, to Diego Morosoli for obtaining every difficult-to-locate article, to Julia Braun and Sarah Haile for their statistical support, and to Anja Frei for her input on the figures.
Financial support. This research received no specific grant from any funding agency, commercial or not-for-profit sectors.

Conflict of interest. None.
Ethical standards. This research did not involve human and/or animal experimentation. State the process for selecting studies (i.e., screening, eligibility, included in systematic review and, if applicable, included in the meta-analysis).

2-3
Data collection process 10 Describe the method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

Data items 11
List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made. 2

12
Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis.

3-4
Synthesis of results 14 Describe the methods of handling data and combining results of studies, if done, including measures of consistency (e.g., I 2 ) for each meta-analysis.

3-4
Risk of bias across studies 15 Specify any assessment of risk of bias that may affect the cumulative evidence (e.g., publication bias, selective reporting within studies).

Additional analyses 16
Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, meta-regression), if done, indicating which were pre-specified.

Study selection 17
Give numbers of studies screened, assessed for eligibility and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram. Fig. 1 Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations. Table 2, Appendix 5

4-5,
Risk of bias within studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12). Table 1 Results of individual studies 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.

Summary of evidence 24
Summarise the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users and policymakers).

7-12
Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias). 8

Conclusions 26
Provide a general interpretation of the results in the context of other evidence and implications for future research.

Funding 27
Describe sources of funding for the systematic review and other support (e.g., supply of data); role of funders for the systematic review.