Feasibility pilot of an adapted parenting program embedded within the Thai public health system

Violence against children (VAC) is a violation of child rights to protection under several international human rights treaties [1‐4]. Estimates of the global prevalence of past year violence, including “moderate” forms such as spanking, is at least 50% – or exceeding 1 billion of the world’s children [5]. Estimates of such rates tend to be higher in low- and middle-income countries (LMICs) than in high-income countries [6], with minimum rates of past-year violence against children aged 2–14 years the highest in Asia, at 68% [5]. The most common perpetrators of VAC are household members [7], with this pervasive form of domestic violence similarly prevalent in Thailand. In a nationally representative household survey, adults reported that family members subjected 75% of children aged 1–14 years to at least one form of physical or emotional punishment in the past month [8]. The consequences of VAC are both immediate and long lasting, including delinquency, criminal activity, low educational performance, perpetration and victimization of intimate partner violence, and adverse mental health in childhood and adulthood [9‐11].

Although harsh parenting and corporal punishment are often deemed normative in Asian LMICs [12], only a minority of adults in many LMICs believe such practices are necessary for childrearing [13] – suggesting they may be open to alternative forms of discipline. A well-established evidence base shows that social learning theory-based parenting programs can effectively reduce harsh parenting and prevent child maltreatment [14‐17]. Such programs augment parenting skills by providing practical instruction on positive-parent child interactions, non-violent discipline techniques, socio-emotional coaching, problem solving, positive encouragement, and responsive supervision [18, 19], and may offer the non-violent methods that Asian primary caregivers want to learn [20, 21]. Interventions targeting difficult child behaviors and parental aggression are often similar to those which aim to prevent child maltreatment, and function by helping to disrupt coercive cycles of parent-child interactions [16, 18].

Several international development agencies are promoting an emphasis on the prevention of VAC and encouraging dissemination at scale [22]. However, such efforts face a number of challenges. First, many evidence-based parenting programs designed in high-income countries often charge expensive fees for licensing, materials, training, and support [23]. Second, while both transported and homegrown parenting programs can effectively reduce child maltreatment and harsh parenting in LMICs [16, 24, 25], some adaptation may be necessary [26]. Adapted interventions require empirical testing in new contexts, while mechanisms of change need to be examined in order to understand parent experiences and how this can inform efforts to maximize intervention effectiveness [27]. Finally, many LMICs must confront gaps in institutional ‘readiness,’ including whether there are sufficient human and technical resources, adequate institutional linkages and infrastructure, and sufficient funding and political will to assure scalability and sustainability [28].

In response, the Universities of Oxford, Cape Town, Stellenbosch, and Bangor, together with the WHO and UNICEF, developed Parenting for Lifelong Health for Young Children (PLH-YC), one of a suite of evidence-based, low-cost parenting programs freely available to LMICs [29]. PLH-YC is a group-based program that targets primary caregivers of children aged 2–9 years. It has shown reductions in child maltreatment and child behavior problems in randomized controlled trials (RCTs) in South Africa and the Philippines, along with improvements in positive parenting in the former and sustained child maltreatment effects at one-year follow-up in the latter¹ [30]. Initially developed for low-income families in Cape Town, it is conducive to cultural and contextual adaptation in different LMIC settings [31, 32].

In Thailand, the University of Oxford, UNICEF, and the Thai Ministry of Public Health (MOPH) developed a partnership to adapt and test PLH-YC in accordance with the UK Medical Research Council guidelines on developing and evaluating complex interventions [33]. This iterative approach allows for the systematic adaptation of an intervention, designed to examine uncertainties, explore user and deliverer acceptability, and gauge outcomes through piloting prior to more stringent testing. Feasibility pilots may improve the quality of subsequent RCTs by focusing on the processes of the main study, such as recruitment, treatment, and assessments, while also exploring the suitability of outcome measures and the willing involvement of participants and deliverers [34]. In 2018, a formative evaluation informed study recruitment and adaptations to PLH-YC, taking into account potential barriers and opportunities for scaling up within routine public health services.² During the current study, we conducted a single group pre-post feasibility pilot of the adapted PLH-YC program with low-income families in rural Thailand (N = 60) through the local public health system. Results from this pilot aim to inform further adaptation and testing in a pragmatic RCT using real-world conditions (N = 120) prior to wide-scale dissemination, if shown to be effective.

This study represents one of few evaluations to test the feasibility of an evidence-based parenting program embedded within routine public health service delivery in a LMIC. It is also the first known scientific study of such a program in Northeastern Thailand, a region that is home to nearly half (47%) of the country’s ‘skipped generation’ households – in which grandparents raise grandchildren in the absence of a mother and father [35]. The key research questions were: 1) Is a rigorous evaluation of a parenting program feasible within the public health system in terms of evaluation approaches, outcome measurement reliability, and adverse events monitoring?; 2) What is the feasibility of delivering an adapted version of PLH-YC to low-income primary caregivers with children aged 2–9 years, as measured by enrolment, attendance, completion, dropout, and fidelity?; 3) What are the indicative effects of the program on reducing VAC (primary outcomes) and associated risk factors (secondary outcomes) according to a hypothesized theory of change (see Fig. 1)?; and 4) What are the perceptions of caregivers and facilitators of program content associated with key mechanisms of change?

This mixed-methods feasibility study was pre-registered with ClinicalTrials.gov (NCT03539341). It is reported in accordance with the CONSORT 2010 extension to pilot and feasibility trials [36], as well as best practice guidelines on mixed methods research [37].

We embedded a qualitative component within a larger, quantitative study to provide an enriched understanding of the experimental pre-post outcomes, and to examine why the intervention may or may not have worked [38]. We collected qualitative and quantitative data in participants’ homes in parallel, with analysis for integration beginning after data collection had been completed [39], merging data by reporting results together and using each type to support or refute respective findings [37, 38]. Strategies for resolving discordance between findings included re-analyzing data, identifying possible theoretical explanations, and challenging construct validity [40]. Screening and pre-test data were collected during December 2018, 1 month prior to intervention delivery, with post-test quantitative and qualitative data collected approximately 1 month after program completion (late March to April 2019). Parent daily report assessments were conducted four times: at pre-test, by phone during the weeks following program sessions 2 and 6, and at post-test.

This feasibility study is one of few evaluations of parenting programs in Thailand and the wider East and Southeast Asian region that aim to reduce VAC [25]. It is also one of few studies that included grandparents as the majority of participants and tested the feasibility and indicative effects of a parenting intervention embedded within the public health system. The potential for program sustainability and scale-up under real world conditions was strengthened by the utilization of existing health system capacities and partnerships, such as health worker facilitators, local and multi-sector recruitment channels, and community-based primary care [85].

This study highlights several lessons and challenges regarding study and intervention feasibility that require consideration. One such issue was the screening procedures used to identify caregivers at risk of perpetrating violence against children, in which only one out of 63 participants did not meet the cut-off criteria. While this may suggest that the threshold was too low, it may also have been due to Village Health Volunteers and teachers seeking out higher risk families during the identification and referral process. Nonetheless, the screening interview should either be eliminated, or the threshold increased to target caregivers reporting higher rates or more severe forms of maltreatment.

Further, recruitment procedures should also be carefully revisited. First, the relatively moderate consent rate and rationales given for declining to participate suggest that many working caregivers were not available to attend the program, even on Sundays. Second, the careful presentation of the program to potential participants was important. During recruitment, a Village Health Volunteer in one area pointed out a caregiver to research staff as abusive and “needing to participate in the program.” This led to clarification by study researchers that the public shaming of caregivers during the referral process may contribute to stigmatization, as well as increase the likelihood that caregivers would decline to participate due to perceptions that the program would be punitive rather than supportive. Third, teachers in one sub-district were important sources of caregiver referrals, suggesting that Early Child Development Centers and primary schools should also be included in future recruitment efforts. Fourth, the use of both phone calls and home visits accompanied by familiar Village Health Volunteers were necessary in many cases, as caregivers contacted via “cold calling” alone were more suspicious of study researchers or simply did not answer calls. Finally, only four male referrals were provided, with only one consenting to participate. Low rates of male engagement in parenting programs are a common challenge across countries [86‐88]; in Thailand, as in South Africa, this may reflect the social norm that childrearing is a role primarily reserved for women [89, 90]. Targeted outreach to male primary caregivers should thus be incorporated into RCT study procedures, while research on barriers to male enrolment and attendance in Thailand should be conducted [91].

The study also underscored the necessity of combining the program with well-articulated child protection and adult welfare referral procedures, as well as the importance of familiarity with local laws and regulations and close collaboration with local service providers. During the course of data collection and program delivery, 14 adults and five children were referred for child protection, social welfare services, or clinical assessments, with most referrals requested by caregivers themselves.

The study also provided a valuable opportunity to assess outcome measurement reliability and suitability. This was of particular relevance given that most of the outcome measurements had not been validated in Thailand. Examinations of Cronbach’s alpha, showing that only 59.3% of values fell within acceptable ranges (see Additional file 2) [83], may reflect weak internal consistency of scales. However, Cronbach’s alpha itself does not provide sufficient evidence of instrument quality, and is less informative for scales that measure several constructs at once [92]. Despite such considerations, the questionable reliability of these instruments suggests that metric invariance should be assessed prior to further use. Furthermore, the pilot tested the suitability of the HOME Inventory as a mixed interview and observational assessment tool. It was substantially adapted and shortened from the original, allowing for application to the range of child ages in the sample and administration within 30–60 min. Although the use of videotaped observations to establish inter-rater reliability between independent assessors would strengthen instrument validity, the study findings suggest that the instrument can sensitively detect observed and reported changes in maltreatment rates and caregiver-child relationships.

Study findings also support the feasibility of delivering an adapted version of PLH-YC for low-income families – including grandparents – in Udon Thani. High rates of program enrolment, attendance, and completion, with no dropouts, suggest that the intervention is relevant and acceptable to participants across a wide age range [93]. Given that one in five children in Thailand reside in households without either biological parent, with many of their primary caregivers being grandparents, it is important that the program is able to meet the caregiving needs of this older population [94]. However, several aspects should be considered prior to future delivery. The provision of travel compensation for all parenting group participants, as well as transport for two groups, may have contributed to high program attendance. This travel compensation in the amount of 150 Baht (4.95 USD), was nearly half of the daily minimum wage (315 Baht or 10.40 USD) for Udon Thani province [95]. Removal of these benefits may result in lower rates of engagement [96]. Moreover, given the large proportion of older participants with visual and mobility impairments, as well as low levels of education and literacy, program materials (e.g., illustrated story posters, parent handbook), the use of mobile text messaging, and disability-inclusive transport need further review to reduce potential obstacles to accessibility.

Following participation in PLH-YC, caregivers reported reductions on all primary outcomes of interest, with moderate size effects on overall child maltreatment and physical abuse, and small effects on emotional abuse. The HOME Inventory instrument also demonstrated moderate size effects on abusive and harsh parenting. Analyses of secondary outcomes showed promising results across almost all outcome measures (22 out of 24). Caregivers reported decreases in dysfunctional parenting; poor child monitoring and supervision; neglect; depression, anxiety, and stress; attitudes supporting physical punishment and harsh discipline; daily and past month child behavior problems; parent sense of inefficacy; and intimate partner violence. There were also improvements in caregiver reports of positive parenting, daily parenting behavior, as well as combination interview and observational assessments of caregiver-child relationships. Qualitative findings converged with quantitative results, with perceived key mechanisms of change including strengthened caregiver-child relationships; increase child compliance and reduced child behavior problems; improved attitudes and strategies toward discipline; as well as improved management of parental stress. There was no evidence of adverse effects.

Although the study lacked a control group and was not designed to test causal impact, findings suggest that this 8-session version of PLH-YC could lead to an array of positive effects, despite a 33% reduction in dosage from the original 12-session version, and is worth testing more rigorously in a subsequent RCT. The program’s overall hypothesized theory of change thus appears viable and complements the mechanisms of change perceived by caregivers and facilitators. However, the study raises two questions. One is regarding the lesser impact on reported rates of child emotional abuse in comparison to child physical abuse, along with qualitative findings indicating that some caregivers may still be utilizing or switching to such practices. This suggests that program content may need to specifically address this form of violence, and that ICAST-TC measures should be expanded to encompass a broader range of emotional abuse items (e.g., telling the child “I do not or will not love you”). Another question concerns the effects of the intervention on intimate partner relationships, as domestic violence was significantly reduced yet coercion and negotiation did not significantly change. This requires empirical examinations of potential moderators of impact and possible diffusion effects.

This study has a number of additional limitations. Foremost, findings from a pre-post, non-randomized study cannot reliably be attributed to the intervention, and can only be indicative of potential program results [97]. Second, we used a relatively limited range of data sources. Due to their young age, we did not utilize child reports or interview children, which could have corroborated findings from caregivers, observers, and facilitators. The inclusion of other reports and individual interviews, such as with other family members, as well as caregiver report and observational assessments with other children in the household, would have provided yet further data on potential impacts outside the dyadic caregiver-target child relationship. While we also included observational measures and periodic daily reports, our reliance mainly on self-report measures and a limited set of individual caregiver interviews therefore raises concerns of recall bias and the influence of social desirability. A third limitation is that post-test assessments and individual interviews were conducted only 1 month after intervention; thus, it is uncertain whether effects were sustained, diminished, or delayed over the medium and long term [98]. Fourth, while internal consistency is only one aspect of instrument validity, the low Cronbach’s alpha scores for many of the measures suggest cautious interpretation of study findings. Fifth, a large number of outcomes were analyzed, which raises the risks of false positive effects. However, no statistical adjustments for multiple tests were undertaken due to the potential for overcorrection and obscuring of treatment effects [99]. Nonetheless, checking with a conservative Bonferroni correction (α = 0.05/32), including different time point comparisons using the PDR, 23 out of 28 effects remained statistically significant, except for proactive parenting, poor child monitoring and supervision, parental anxiety and stress, and IPV. Finally, given the time constraints for this study, we did not conduct qualitative interviews with caregivers who did and did not do well following program completion based on outcome analyses. This would have yielded deeper insights on underlying mechanisms of change and allowed us to better isolate the parenting techniques perceived as having the most and least utility.

Despite such limitations, the study has many strengths. First, the strong involvement of high-level public health officials and local practitioners raises prospects for the scaling up of the intervention – should positive effects be sustained in subsequent RCT testing. Second, the indication by caregivers through individual interviews that they shared acquired program skills with household members may engender diffusion effects and public demand that fosters scalability. A third strength was the inclusion of observer ratings, regarded as a “gold standard” in objective evaluations of changes in caregiver-child interactions [100], as well as periodic daily reports, which avoid the demands of aggregate recall over multiple days or for estimates of behavior frequency [101]. Fourth, significant reductions in child maltreatment rates converged with qualitative data. This demonstrates the value of adopting mixed methods approaches to feasibility pilots, as it allows for the maximizing of strengths from each type of data. Finally, the inclusion of audio-CASI to administer more sensitive items regarding child maltreatment and IPV assisted in encouraging disclosures and reducing risks of social desirability bias.

This mixed methods study is the first known evaluation of an evidence- and group-based parenting intervention delivered within the public health system in Thailand, with several notable findings. First, it shows the feasibility of study evaluation approaches, including data collection methods and the absence of adverse effects, although the reliability of many outcome measurements is uncertain. Second, it also demonstrates program feasibility and the preliminary effectiveness of a parenting intervention in reducing child physical and emotional violence and other related outcomes within low-income families in Thailand, with perceived key mechanisms of change corroborating some of these effects. Finally, it highlights that close collaboration with policymakers and local practitioners is possible, thus improving prospects for program scalability and sustainability through the utilization of existing routine public health service staff and delivery systems. While further research is necessary to determine program effectiveness before such expansion efforts should be undertaken, these initial findings suggest a promising contribution toward violence prevention in Thailand and other LMICs more broadly.