Parental engagement in preventive parenting programs for child mental health: a systematic review of predictors and strategies to increase engagement

Background

Mental health problems are a leading cause of disability in children and young people worldwide (World Health Organization, 2016a). Mental health problems can be defined as a dysregulation of mood, thought and/or behaviour, and categorized more broadly into internalizing or externalizing problems for children (American Psychiatric Association, 2013). In the current review, ‘mental health problems’ is used to encapsulate both internalizing and externalizing problems (American Psychiatric Association, 2013). Mental health problems are typically followed by subsequent adverse outcomes for children including psychological distress, functional impairment, exposure to stigma and increased risk of premature death (Patel et al., 2007).

The World Health Organization (WHO) has called for greater attention to be given to the prevention and promotion of mental health at all levels of society (WHO, 2016b). It is important here to differentiate between mental health promotion and prevention, as these terms are often confused or conflated. Health promotion is the development, progress or establishment of practices that increase overall health and wellbeing (Czeresnia, 1999), and is not addressed in the current review. Instead, the focus is on prevention, defined as “interventions directed to averting the emergence of specific diseases, reducing their incidence and prevalence in populations” (Czeresnia, 1999, p. 705). Despite varying definitions for different stages of prevention, here we follow Haggerty & Mrazek’s (1994) widely adopted stages of universal, selective and indicated. The goal of universal prevention is to target the public and deliver an intervention that can minimize potential risk and increase protective factors for mental health. Selective prevention is designed to deliver interventions to individuals whose risk of developing a mental health problem is higher than others in the population, while indicated prevention specifically targets persons at high risk.

Parenting programs to prevent child mental health problems

Recent systematic reviews demonstrate that parenting behaviours (i.e., less warmth, and more inter-parental conflict) are associated with children’s and adolescents’ risk of mental health problems (Rothbaum & Weisz, 1994; Yap & Jorm, 2015; Yap et al., 2014). Hence, parenting programs that aim to modify parenting behaviours have the potential to prevent mental health disorders in children and adolescents. These programs, whether face-to-face or online, have shown promise in preventing both internalising disorders (Yap et al., 2016) and behaviour problems, as well as increasing other child competencies (Sandler et al., 2015; Sandler et al., 2011). Parenting programs can be defined as any intervention delivered to parents with the main objective of increasing parental knowledge, skills and confidence, whilst reducing the prevalence of mental health, emotional and behavioural problems in children and adolescents (Sanders et al., 2008). These programs assume that changing parenting behaviours will in turn alter a child’s risk of developing mental health problems. This assumption stems from theoretical underpinnings suggested by Sandler and colleagues (2011), that a parenting program improves parenting skills and parental self-efficacy, causing a reduction in barriers to effective parenting, and in turn facilitating long-term benefits for the child. Despite the potential benefits of preventive parenting programs, many studies examining the effectiveness of such programs have reported difficulties in engaging parents (Gross, Julion & Fogg, 2001; Ingoldsby, 2010; Morawska & Sanders, 2006; Orrell-Valente et al., 1999; Panter-Brick et al., 2014). Poor parental engagement, including parental uptake and ongoing engagement (attendance), could lead to both the effectiveness of these programs being under-reported and parents not adequately developing the key skills required to prevent mental health problems (Morawska & Sanders, 2006).

Parental engagement in preventive parenting programs has been defined inconsistently across studies (Gross, Julion & Fogg, 2001; Orrell-Valente et al., 1999). However, these definitions can generally be broken down into three discrete components. The first component, initial engagement, includes two phases: (1) parental intent to enrol, measured either through initial expression of interest rates, occurring prior to or separate from signing consent forms, or through a direct question (e.g., ‘Do you intend to enrol?’); and (2) actual enrolment, as described by the study, as the number of parents who enrolled in the program (e.g., number of parents who signed a consent form) (Dumas, Nissley-Tsiopinis & Moreland, 2007; McCurdy & Daro, 2001; Spoth et al., 1996). The second component, ongoing engagement, is measured by the proportion of parents attending at least one session or completing at least one module of a self-administered or online intervention, the total number of sessions attended by parents, or the number of parents who completed the program (Dumas, Nissley-Tsiopinis & Moreland, 2007; McCurdy & Daro, 2001; Spoth et al., 1996). The third component of engagement, quality of engagement, measures both what parents invest in and receive from the program (e.g., taking part in group discussions or completing homework tasks; Chacko et al., 2016; Orrell-Valente et al., 1999). This component (1) is determined by the type of activities parents are asked to take part in during sessions, specific to each program; and (2) is suggested to form part of the key mechanism for positive parenting change (Kazantzis, Deane & Ronan, 2000; Piotrowska et al., 2017), thus is more closely related to program outcomes than the other two components.

Engaging parents in prevention

In two recent reviews, Ingoldsby (2010) and Chacko et al. (2016) attempted to collate research on parental engagement. Ingoldsby (2010) reviewed ongoing engagement and retention of families attending both intervention and indicated prevention programs designed to improve child mental health (child age range not specified). The main findings from Ingoldsby’s (2010) review included: (1) brief strategies implemented at the beginning of the program addressing families’ practical and psychological barriers effectively increased engagement in early sessions; and (2) strategies that were ongoing throughout the intervention, and focused on motivational interviewing, family systems and family stress, demonstrated longer-term increased engagement. More recently, Chacko and colleagues (2016) reviewed and discussed predictors of parental engagement, including the domains of attendance and attrition (ongoing engagement) and treatment adherence (quality of engagement). Chacko and colleagues (2016) concluded that at least 51% of parents drop out at some stage of the intervention, with this high level of attrition found to be somewhat influenced by lower socio-economic status (SES). Both reviews adopted inclusion and exclusion criteria that may have resulted in the exclusion of studies focusing on universal and selective prevention programs (i.e., Heinrichs et al., 2005; Spoth & Redmond, 2000), as well as those using open access recruitment methods.

Engaging parents across children’s lifespan

During the developmental transition into adolescence, the corresponding changes in the parent–child relationship include increasing autonomy and time spent apart from the parent, and an increased importance of peer relations (Collins, 1990). In this context, parents may perceive that their role in their child’s mental health and well-being is no longer as important as when their child was younger. This could account for the low rates of engagement in preventive programs for parents of adolescents (Burke, Brennan & Roney, 2010). However, there is a substantial body of evidence to suggest that even when a child moves into adolescence, parents still play an important role in their child’s risk for both internalising (Yap et al., 2014) and externalising disorders (Sandler et al., 2011). Chacko and colleagues’ (2016) review only assessed parental engagement in programs for parents of children aged 2–12 years, finding no effect of child age. Hence, it remains unclear if parental engagement in programs differs depending on the age of the child, when considering the whole developmental period from birth through to late adolescence (0–18 years).

Engagement enhancement strategies

In addition to studies that have examined predictors of parents engaging in a preventive parenting program, there is emerging research on the effectiveness of engagement enhancement strategies for parental engagement. Ingoldsby (2010) found that additional strategies implemented at the time of enrolment, including brief intensive engagement interventions that are both practical (e.g., providing transportation) and psychological (e.g., addressing beliefs about the treatment process) in nature, could increase parental attendance during early stages of the intervention. In another review of strategies to recruit any type of participant into a RCT, Caldwell et al. (2010) found that any strategy that increased a participant’s awareness of the health problem being studied increased recruitment, a finding that is in line with the Health Belief Model (Rosenstock, 1974). The Health Belief Model was proposed to explain and predict health-related behaviours, such as attending health care appointments. This model focuses on the attitudes and beliefs of individuals; for instance, if a parent has increased awareness of their child’s susceptibility to developing a mental health problem, they may be more likely to engage in a preventive parenting program. In addition to the Health Belief Model, the Theory of Planned Behaviour and Reasoned Action (Ajzen & Driver, 1991) may also be used to inform engagement enhancement strategies (McCurdy & Daro, 2001). The Theory of Planned Behaviour and Reasoned Action links an individual’s beliefs and attitudes about subjective norms (e.g., the perceived social pressure to perform or not perform a behaviour; Ajzen & Driver, 1991) and perceived behavioural control (e.g., a parent’s perceived ease or difficulty of performing a certain behaviour; Ajzen & Driver, 1991). For example, McCurdy & Daro (2001) posit that subjective norms, as determined by communications with others, can influence a parent’s decision to engage in a parenting program. Therefore, it appears there are additional strategies researchers can utilise during the initial engagement stage of a study to increase parental engagement across subsequent stages. Engagement enhancement strategies are defined in the current review as any methodology that looks to use evidence- or theoretically-based strategies to increase parental engagement.

The current systematic review

The current review will extend on previous findings by reviewing both potential predictors of engagement as well as engagement enhancement strategies used. Specifically, the current review aims to delineate factors/strategies that can be applied across different types of parenting programs. Hence it focuses on the initial engagement and ongoing engagement components, but not quality of parental engagement, because the latter is related to program-specific components. Such a synthesis can inform researchers regarding the predictors of initial and ongoing parental engagement, and suggest some possible theoretical models that could be used in the development of engagement strategies to increase the uptake of evidence-based preventive parenting programs. To identify factors predicting parental engagement in programs where parents are the main target of intervention, the current review follows Yap and colleagues’ (2016) definition of a ‘preventive parenting program’: a program aimed at preventing child mental health problems through education and subsequent skill development of parent and primary caregivers, that specifically involves parents in more than 50% of the program. Additionally, the current review will include all programs specifically focused on the prevention of child mental health problems, across both childhood and adolescence (0–18 years), to explain the association between child age and parent engagement. This review aims to shed light on whether there is a more optimal time to promote parenting programs, and inform future parenting program design and implementation, with the ultimate goal of maximising parental uptake of preventive programs.

Specifically, this review aims to: (1) investigate the predictors of parental engagement in preventive parenting programs, across the initial engagement (intent to enrol and enrolment) and ongoing engagement (attendance) components. Of particular interest, we aim to examine whether parental engagement differs depending on the age of the child at the time of parent participation; and (2) explore whether any strategies used by researchers to increase parental engagement have been successful.

Search strategy

This review was conducted following the Cochrane Collaboration guidelines (Higgins & Green, 2008). The following electronic databases were searched: Cochrane Library, Informit online, Ovid MEDLINE, ProQuest, PsycINFO, PubMed and Scopus. The search was limited to studies written in English and articles published between 2004–2014. This publication date range was chosen to increase the likelihood that findings from this review will be more recent and relevant to current and future parenting programs. The initial search was conducted by the first author (SF) on the 12th of January 2015. To ensure the latest data was included in the review, an update search was conducted (also by SF) on the 21st of July 2016 to include articles published between January 2015–July 2016. Search terms included multiple terms for engagement, parents, programs, prevention, child and mental health (for a full list, refer to Supplemental Information 2).

Unpublished reports, dissertations and grey literature were sourced through Google Scholar and dissertation databases. Manual searching of reference lists of included studies was conducted to locate further relevant articles and dissertations of interest.

Inclusion and exclusion criteria

Studies were included if they were controlled trials (randomised and non-randomised), cross-sectional, case-control, and longitudinal studies. Excluded studies included; therapy/treatment interventions (note: all prevention interventions were eligible for inclusion), reviews or meta-analyses, qualitative studies, discussion papers, and papers published in languages other than English.

Study selection and data extraction

Titles and abstracts of identified studies were reviewed to determine if they met inclusion criteria. Full texts of articles that met inclusion criteria were assessed by the first author (SF). Thirty-five percent of these titles and abstracts were independently assessed by a second author (BS) to confirm inter-rater reliability of the inclusion criteria. Inter-rater reliability of inclusion criteria was 99.2%, with one additional article being included in the review. All reasons for exclusion are documented in the PRISMA diagram (Fig. 1). Disagreements were resolved through discussion and review of the extraction sheet by SF and BS, with involvement of other authors (NP and MY) when necessary.

Coding of predictors

Engagement factors were identified as factors that could influence a parent’s engagement in preventive parenting programs. Categories were specified when two or more of the included studies examined the same engagement variable. The categories identified included: parent age, gender of parent, parent education status, parent employment status, parent race/ethnicity, parental mental health status, child age, child gender, child mental health symptoms, family structure and one- or two- parent households. Definitions for these eleven categories were based on consensus in the relevant literature (see Table 1 for definitions and examples of measures used). In addition, several other factors could not be coded into categories, including parenting behaviour measures and individual and neighbourhood socioeconomic status. Of the six studies which assessed parenting behaviours as a predictor, the types of parenting behaviours were too diverse to combine into a single meaningful category (Yap et al., 2014; Yap & Jorm, 2015). The eight different parenting behaviour categories included; discipline (n = 3), parent self-efficacy (n = 2), parent warmth (n = 1), positive parenting style (n = 2), negative attribution/conflict (n = 2), knowledge of school performance (n = 1), restrictive attitude to alcohol (n = 1) and parenting problems (n = 1). Those parenting behaviour categories with n > 1 were incomparable due to the use of different indicators, and/or they were used to predict different stages of engagement. Additionally, as recommended by Braveman and colleagues (2005), measurement of specific socioeconomic factors were assessed separately rather than combined as an overall socioeconomic position (SEP). This resulted in the following SEP-related categories: parent education status, parent employment status, family structure and one- or two- parent households.

Data analysis and Stouffer’s p

A meta-analysis to assess effect sizes was not possible due to differences in interventions, settings, predictor variables, and analytic methods. To compensate for this limitation, the Stouffer’s method (Stouffer et al., 1949) of combining p-values was used to synthesize the findings of many of the included studies, since it can be applied in cases where studies analyse data in a variety of ways. Stouffer’s z was calculated by dividing the sum of the z(p_i) values by the square root of k (where k refers to the number of associations). Stouffer’s z s were calculated to determine the overall p-value of the associations reviewed for each combination of predictor category and engagement factor. If the resulting Stouffer’s z corresponded to a probability level less than 0.01, the null hypothesis of no effect was rejected. This methodology has been used in other systematic reviews, including by Yap and colleagues (2014; 2015), to assess whether associations between variables are reliable. For information on how p-values were extracted and selected for analysis, see Supplemental Information 2.

Assessing risk of bias

Critical appraisal of quantitative studies was conducted using the Cochrane Risk of Bias Tool (Higgins & Green, 2008), which involved assessing for adequate sequence generation, allocation concealment, blinding of assessors to treatment conditions, the inclusion of intention to treat analyses and assessment of potential confounders. Risk of bias for all included studies was assessed by two authors (SF and BS) using a standardised, pilot-tested extraction sheet. Disagreements were resolved through discussion between SF and BS.

Study characteristics

Synthesis of results: predictors of parental engagement across stages of engagement

Seventeen studies measured the factors that predict parent engagement (Baker, Arnold & Meagher, 2011; Brody et al., 2006; Byrnes et al., 2012; Calam et al., 2008; Carpentier et al., 2007; Eisner & Meidert, 2011; Fleming et al., 2015; Garvey et al., 2006; Heinrichs et al., 2005; Hellenthal, 2009; Mauricio et al., 2014; Mian, Eisenhower & Carter, 2015; Nordstrom, Dumas & Gitter, 2008; Plueck et al., 2010; Reedtz et al., 2011; Skärstrand et al., 2009; Winslow et al., 2009). Most studies focused on predictors that could be coded into categories in this review, but isolated studies also looked at how parent cognitions (i.e., parents’ thoughts and beliefs about themselves and the program; Nordstrom, Dumas & Gitter, 2008) and parent-recorded obstacles to engagement (i.e., need for child care, transportation costs) measured at pre-intervention, can predict parent engagement across a preventive parenting program trial (Mian, Eisenhower & Carter, 2015; Nordstrom, Dumas & Gitter, 2008). Although some of these findings appear promising, there is little consistent evidence across studies to support any predictors of parent engagement, at the time of this review. This inconsistency is demonstrated in Table 5, where only one of the 11 categories assessed yielded a significant Stouffer’s p, and only for the enrolment stage of engagement.

Synthesis of results: effects of engagement enhancement methods

Nine studies attempted to increase engagement through ‘engagement enhancement methods’ (Aalborg et al., 2012; Bjørknes, Jakobsen & Nærde, 2011; Byrnes et al., 2012; Calam et al., 2008; Carpentier et al., 2007; Eisner & Meidert, 2011; Heinrichs, 2006; Helfenbaum-Kun & Ortiz, 2007; Mian, Eisenhower & Carter, 2015). Five of the nine studies tested engagement enhancement methods using randomised controlled trials (Aalborg et al., 2012; Byrnes et al., 2012; Calam et al., 2008; Heinrichs, 2006; Mian, Eisenhower & Carter, 2015), while a further four used engagement enhancement methods to recruit all participants (Bjørknes, Jakobsen & Nærde, 2011; Bjørknes & Manger, 2013; Carpentier et al., 2007; Eisner & Meidert, 2011; Helfenbaum-Kun & Ortiz, 2007).

The studies that randomised participants into different engagement methods had varied results. Studies randomised parents into either paid versus unpaid conditions (Heinrichs, 2006), web-enhanced versus standard video viewing conditions (Calam et al., 2008), or enhanced recruitment (using personalised letters and phone calls) versus recruitment as usual (Mian, Eisenhower & Carter, 2015). Heinrichs (2006) randomised ‘parents’ and ‘child care centres’ into paid and non-paid conditions. Heinrichs found that parents’ intent to enrol was significantly increased when offered payment for attending sessions, however actual enrolment and attendance did not differ between paid and unpaid conditions. Additionally, although Heinrichs hypothesised that offering payment would increase engagement from migrant parents, the research demonstrated that the opposite was true: native-born parents were more likely to engage in the program when offered payment than those who were first-generation migrants (Heinrichs, 2006). In Mian and colleagues’ (2015) enhanced recruitment study for a once-off seminar, parents in the enhanced condition (including personalised letters and follow-up phone calls) were significantly more likely to intend to enrol, and this intent was found to be related to ongoing engagement. Furthermore, Calam and colleagues (2008) investigated whether an internet-enhanced version of a video-based program “Driving Mum and Dad Mad’ would affect parents’ ongoing engagement (recorded as number of videos watched). The internet-enhanced version included parents having access to a website with further information and activities related to each weekly video. Parents watched on average the same number of videos regardless of the condition they were assigned to.

Two RCT studies (reported in three articles) assessed the likelihood that ongoing engagement would increase if parents were given a choice of program (Aalborg et al., 2012; Byrnes et al., 2012; Miller et al., 2011). These studies used preventive parenting programs with significantly different presentation styles. The first, Family Matters (FM), required parents to complete four booklets at home with their adolescents, and families received weekly phone calls from the research team (Aalborg et al., 2012; Byrnes et al., 2012; Miller et al., 2011). The second, the Strengthening Families Program (SFP), required families to attend seven two-hour weekly group sessions at a medical facility (Aalborg et al., 2012; Byrnes et al., 2012; Miller et al., 2011). These studies found that compared to parents who were randomised to the corresponding no-choice condition, parents in the choice condition who chose FM completed the booklets over a significantly shorter period, and parents who chose SFP attended more sessions (Aalborg et al., 2012; Miller et al., 2011). In addition, parents who self-selected into the two different programs demonstrated some significantly different characteristics. Parents who chose the FM program were more likely to be educated, whereas parents who chose the SFP program described their adolescent’s behaviour problems as more severe (Miller et al., 2011). Miller and colleagues (2011) hypothesised that this difference between parents’ program choice could be because parents who rated their teenager’s behaviour problems as more severe felt they needed a more personalised level of intervention (SFP).

Finally, four studies used engagement enhancement methodologies to recruit all participants (Bjørknes, Jakobsen & Nærde, 2011; Bjørknes & Manger, 2013; Carpentier et al., 2007; Eisner & Meidert, 2011; Helfenbaum-Kun & Ortiz, 2007). One study discussed the different number and type of participants recruited from (1) local information meetings which included talks by cultural leaders, (2) public service professionals, and (3) the recruitment team’s personal and professional networks, and compared this to the number of hours required to recruit these participants (Bjørknes, Jakobsen & Nærde, 2011). The authors reported that local information meetings were the most successful recruitment approach, with 57% of their sample recruited through these meetings (Bjørknes, Jakobsen & Nærde, 2011). In addition, these meetings were also the most cost-effective and least time-intensive approach per participant (Bjørknes, Jakobsen & Nærde, 2011). Recruitment through public service professionals was seen to be the least effective recruitment approach, accounting for less than 15% of recruited participants (Bjørknes, Jakobsen & Nærde, 2011). Further analysis revealed that parents recruited from local information meetings and the recruitment team’s networks had significantly poorer Norwegian language skills (Bjørknes, Jakobsen & Nærde, 2011).

The remaining three studies that used enhanced strategies to recruit all participants included writing personal letters to parents in their own language, promoting programs via meetings at the schools or child care centres, and making phone calls to parents using experienced interpreters. These studies were found to have enrolment rates of 31.3% (Eisner & Meidert, 2011), 62% (Carpentier et al., 2007) and 85%, respectively (Helfenbaum-Kun & Ortiz, 2007). Eisner & Meidert (2011) reported similar levels of ongoing engagement in comparison to other preventive parenting programs that did not use enhanced recruitment strategies, whilst Carpentier and colleagues (2007) and Helfenbaum-Kun & Ortiz (2007) reported a large proportion of parents dropping out of the study, either before attending the first session or during the program.

Summary of evidence

Recommendations for future research

Strengths and limitations of this review

To our knowledge, this is the first systematic review of studies with the primary outcome of measuring and predicting parent engagement in programs that are specifically focused on the prevention of child mental health problems. Unlike other reviews that focused on reviewing literature for one specific type of parenting programs (e.g., Behavioural Parent Training; Chacko et al., 2016), this review did not place any restrictions on the type of preventive parenting program or type of mental disorder the program aimed to prevent. Additionally, we placed no restrictions on the age of the child at the time the program was delivered. Employing wider inclusion criteria allowed us to draw together the sparse literature and develop recommendations both for increasing parental engagement, and for the reporting of such research. However, even with this wide inclusion criteria, only 21 studies were identified. This limited our ability to draw firm conclusions and as such, all findings stemming from this review should be viewed as preliminary in nature.

Furthermore, some limitations of our findings should be noted. Firstly, there were not enough studies included in this review that consistently defined variables (both predictors of engagement, and stages of parental engagement), and that employed similar methods of analysis, to permit a meta-analysis to estimate effect sizes. As such, the Stouffer’s p analysis was adopted to estimate the reliability of associations between investigated predictors and parental engagement. Nonetheless, the Stouffer’s p method is unable to weight studies according to sample sizes (Darlington & Hayes, 2000). Furthermore, there has been a shift within the academic community away from reporting p-values as an indicator of significant results (Thomas & Pencina, 2016). This is due to the prevalent misuse of p-values to arbitrarily divide studies into significant and non-significant, which was not the intention of the founders of statistical inference (Sterne & Smith, 2001). Effect size measures along with confidence intervals have also been demonstrated to be more clinically relevant than a stand-alone p-value (Thomas & Pencina, 2016). In light of this, the quantitative results of this review should be interpreted with caution, and be considered instead as hypothesis-generating findings to guide future research.

As observed in other reviews (Chacko et al., 2016; Ingoldsby, 2010; Yap et al., 2016), study quality could not be accurately assessed due to poor reporting, particularly the selective reporting of ongoing engagement measures. No studies included in the current review could be considered to have low risk of bias across all six domains assessed, hence the results should be interpreted cautiously. All studies included in the current review were either RCT’s or experimental trials, resulting in an inability to assess if parents’ engagement differed for open access versus RCTs of preventive parenting programs. Many of the studies included in this review did not have a rigorous measure of the child’s current or previous mental health diagnoses. Given that mental health issues are common and potentially under-diagnosed in community-based samples (McManus et al., 2009; McManus et al., 2016), it is possible that included studies did not have truly preventive samples. Nonetheless, by including these community-based samples, this review may provide a more ecologically valid review of the potential predictors of parental engagement in prevention programs. Future studies employing rigorous diagnostic assessments and excluding data from participants with past or current diagnoses are required to verify whether predictors of parental engagement may differ across various populations. Finally, included studies consisted of articles written in English and published during or after 2004, therefore the current review findings may not generalise to studies published outside these dates or to literature published in other languages.