Abstract
Objective:
This study examined the efficacy of Hands-On Parent Empowerment-20 (HOPE-20) program.
Methods:
Eligible participants were parents residing in Hong Kong with target children aged 2 years attending nursery schools. Cluster randomized control trial was adopted, with 10 schools (110 participants) assigned to intervention group and 8 schools (63 participants) to control group using random number table, without blinding of participants. Intervention group attended 20 parent training sessions based on social learning theory. Children were individually assessed on preschool concepts and language skills. Parents completed child behavior and parental stress questionnaires.
Results:
Intention-to-treat analysis (173 participants and 18 schools) indicated improvement in primary outcomes: child preschool concepts (d = 0.70, 95% confidence interval [CI]: [0.38, 1.01]) and child behavior problems (d = 0.67, 95% CI: [0.35, 0.99]), together with language skills (d = 0.98, 95% CI: [0.65, 1.30]), and parental stress (d = 0.71, 95% CI: [0.39, 1.02]).
Conclusions:
The results suggested that HOPE-20 program was beneficial to Chinese families.
Keywords
Webster-Stratton and Taylor (2001) highlighted three levels of support to parents: Universal preventive programs serve children or parents without selective criteria; selective programs serve children or parents at risk of various factors, which might affect the children’s development; and indicated programs serve children or parents with confirmed diagnoses. From a health promotion perspective, Rose (1992) argued for the importance of a population or universal approach. Following his argument, the strength of universal preventive programs in parent training is that it can shift the distribution of parenting in a more positive direction in the population and brings cumulative benefit to the population (Rose, 1992). Focusing on a small group of high-risk parents with parenting difficulties might not bring lasting effects to the population unless we can achieve a widespread change in population parenting standards. Malti, Ribeaud, and Eisner (2011) also argued that universal programs could reduce the stigma of participation in selective or indicated programs.
Though universal programs are beneficial to the population, Malti et al. (2011) noted that most child mental health services were targeted services and there were few universal programs. Universal programs for parents with young children are useful in supporting parents to promote the development of their children. Heckman (2000) demonstrated that investment in programs for young children was more cost effective than investment in later remediation for early disadvantage among adolescents/adults to achieve the same outcome. He pointed out that early support to parents to help children develop social and cognitive skills could promote schooling, reduce crime, and improve workforce productivity. In the case of children, universal programs should aim to positively change the socializing environment for children as well as children’s behavior so as to shield children from exposure to risk factors contributing to child behavior problems (Spoth, Guyll, & Shin, 2009). Malti et al. (2011) also suggested that universal programs for children should target risk factors for child behavior problems and maximize protective factors. In their study on the effectiveness of a universal program for parents of first graders, they found a reduction in child behavior problems (Malti, Ribeaud, & Eisner, 2011). Webster-Stratton, Reid, and Hammond (2001) also reported fewer conduct problems and more positive parenting using the Incredible Years program as a universal program.
This study reported the development and evaluation of a universal program for Hong Kong Chinese parents with preschool children. In the following sections, the Hong Kong situation and service gaps would first be described, to be followed by details on the development of the program and its intended outcomes.
The Population and Current Provisions in Hong Kong
According to the Hong Kong 2011 census, there were 196,350 children aged 1–4 years (Census and Statistics Department, 2012) or approximately 50,000 children in each age-group. In a survey on local preschool children (Leung, Leung, Chan, Tso, & Ip, 2005), the prevalence of behavior problems was 10.5%, which was comparable to international figures (Campbell, 1995). It was also found that parental stress was associated with child behavior problems. Another study on local preschool children found that about 10.6% of the sample was identified to be developmentally vulnerable in language and cognitive development (Ip et al., 2013).
Maternal and Child Health Centers (MCHCs) provide free immunization, health and developmental surveillance services for all Hong Kong children from birth to 5 years, and universal parent training for their parents, including single-session talks and leaflets on child development, childcare, and parenting. However, there is no published information on the effectiveness of the universal parent training. The MCHCs are also not efficient in reaching parents of children beyond 18 months, as visits to MCHCs are less frequent after completion of the immunization schedule by 18 months. MCHCs also offer Positive Parenting Program (Triple P; Sanders, 1999), which has been shown to be effective for Hong Kong Chinese families (Leung, Sanders, Leung, Mak, & Lau, 2003). However, this program is only offered to parents who experience difficulties in managing their children.
Social services to families in Hong Kong are offered through over 65 Integrated Family Services Centers (IFSCs) spread over the territory, which provide a continuum of preventive, supportive, and remedial services. While a variety of universal, selective, and indicated programs are offered by individual IFSCs, there is limited published information on the effectiveness of these programs.
For education, kindergartens serve children 3–6 years of age. The kindergarten attendance rate is about 91.3% for children aged 3–5 years (Census and Statistics Department, 2012). Nursery schools provide care and education services for children under 3 years. At the time of the study, about 34,000 nursery school places were available for 2-year-old children (Education Bureau, 2014). Kindergartens and nursery schools organize occasional parenting talks for parents as universal programs, but there is little published information on their content or effectiveness. The Education Bureau developed a seven-session group parent training manual for parents of 3- to 6-year-old children (covering self-esteem, language, learning, and behavior management). Again, published information on its effectiveness or usage is absent.
From the above review, two service/research gaps are evident. First, while universal programs for families are offered in various sectors, there is little or no published information on their effectiveness. Second, there is no systematic population-level strategy for the delivery of universal programs for children after 18 months. The Hands-On Parent Empowerment-20 (HOPE-20) was designed as a universal program targeting parents with children aged 2 years to address these service/research gaps.
The HOPE-20 Program
The HOPE-20 program aimed to (i) equip parents with the skills and knowledge to promote child learning and to manage child behavior, (ii) to reduce parental stress, and (iii) to increase parent social support. Problematic parenting and childhood conduct problems were considered risk factors for adolescent conduct problems and thus it was important to enhance parenting skills and to prevent childhood conduct problems. As academic difficulties were associated with behavior problems (Webster-Stratton & Taylor, 2001), it was important to enhance children’s learning and decrease problem behavior. Further, as child behavior problems were associated with parental stress, reduction in child behavior problems should also reduce parental stress. Social support has been found to be associated with child behavior problems and parental stress (Leung et al., 2005). The peer support generated from group attendance of a multi-session parent education program should provide useful social support to the participating parents.
The content of HOPE-20 took reference from evidence-based parent training programs found to be effective with Hong Kong Chinese parents. Development of HOPE-20 also followed closely the National Institute for Health and Clinical Excellence (NICE, 2006) guidelines and the literature on effective parenting programs (e.g., Small, Cooney, & O’Connor, 2009).
For program content, HOPE-20 was based on the evidence-based 30-session HOPE program, a selective program designed by the first and second authors for new immigrant parents with preschool children aged 3–5 years in Hong Kong. HOPE was based on social learning theory and the theories of Piaget and Vygotsky. They emphasized the importance of interaction with a significant adult (parent[s]) to develop child competence, that children learn best within the zone of proximal development and active learning through experience-based learning. As social learning theory findings indicated that coercive parent–child interaction patterns could result in antisocial behavior (Patterson, 1982), positive child management skills were taught to enhance positive parent–child interaction. HOPE was found by a randomized controlled trial (RCT) study to be effective in reducing child behavior problems and parental stress and enhancing social support (Leung, Tsang, & Dean, 2011). Another reference was the Healthy Start Home Visit Program (HSHVP), a selective program which was also based on HOPE. The HSHVP was designed for disadvantaged families with preschool children aged 3–5 years in Tuen Mun area, an outlying area in Hong Kong. The effectiveness of the HSHVP was demonstrated by an RCT study, which showed increase in child learning and parent social support and decrease in child behavior problems and parenting stress in local families (Leung, Tsang, & Heung, 2014). In addition to these locally developed programs, the HOPE-20 took reference from Triple P and Parent–Child Interaction Therapy, both being selective programs found to be effective for Hong Kong families (Leung, Sanders, et al., 2003; Leung, Tsang, Sin, & Choi, 2015).
Similar to HOPE, the content of HOPE-20 included child management techniques which were based on social learning theory such as using calm, specific, and clear commands to increase compliance, use of praise and reinforcement to increase positive behavior, and using planned ignoring and time out to deal with misbehavior (Pearl, 2009). Strategies for enhancing parent–child relationship, learning basic preschool concepts, enhancing language skills, and paired reading were also included (see Table 1 for details). Being a multi-session group program including role-play practice and discussion, it was expected that social support among the group members could be enhanced through the opportunity to share with other parents who were neighbors with similar background.
Program Content.
HOPE-20 was different form HOPE in two ways. First, the target children of HOPE were 3- to5-year-olds, while that for HOPE-20 were 2-year-olds. The content of HOPE-20, derived from HOPE, was adjusted according to the local preschool curriculum guide (Curriculum Development Council, 2006) to make it developmentally appropriate. HOPE-20 focused on primary colors, basic shapes (e.g., circle and square), matching and sorting, and removed content too advanced for the age-group such as categorization. Second, the number of sessions was reduced from 30 to 20 to make it less time demanding for parents. The target participants of HOPE-20 were parents in the general population who might find it easier to grasp the session content than new immigrant parents in HOPE, some of whom might have difficulties with the local language. The child management strategies in HOPE were condensed in the HOPE-20 program, so they could be covered in fewer sessions.
The NICE guidelines stated that the program should be structured and be theoretically driven, using social learning theories, be comprehensive, developmentally and culturally appropriate, with sufficient dosage, and with a focus on parenting strategies, as they were the mediators for change in child behavior (Gardner, Burton, & Klimes, 2006). In-session role play and homework between sessions should be used. The program should be delivered by trained personnel and be conducted in group format. Correction strategies should be in place to help parents attain the target learning outcomes.
HOPE-20 was a structured program and the package included a facilitator’s manual detailing the content of every session, with accompanying PowerPoint slides for lecture to parents, parent notes, and homework activities for parents. For dosage, the program consisted of 20 2-hour weekly group sessions. In each session, role play was used to help parents master the homework skills. Parents had to spend 5 minutes each day between sessions to do homework practice with their children. The facilitators were registered social workers with degree training in social work, who received supervision from the first and second authors (psychologists) through regular team meetings, phone, and e-mail consultations.
Each session lasted for 2 hours. Each session consisted of (i) review of the previous session and homework, (ii) mini-lecture on the topic to be covered, (iii) explanation and demonstration of the homework for the coming week by the facilitator, and (iv) role play by participating parents to practice the homework, with feedback by the facilitator. The children were not present during the group sessions. Parents could bring their spouses where possible though the same parent from each family who completed the pre-intervention measures would complete the post-intervention measures in the study.
For cultural appropriateness, HOPE-20 was based on locally developed programs for preschool children, HOPE and HSHVP, as well as overseas programs found to be effective for Hong Kong families. Input from parents on their needs was collected from focus groups with local parents. Examples suited to the local context were used for illustration of social learning theories in parenting and the learning theories of Piaget and Vygotsky. For comprehensiveness, the program covered essential elements in child psychosocial and cognitive development.
HOPE-20 was delivered in group format in nursery schools with about 10 parents per group. As a fair percentage of 2-year-old children attend nursery schools, nursery schools are strategic platforms of universal programs. They are accessible by parents and stigma-free in service attendance. They share the same goals as HOPE-20 in promoting child development (Gross & Grady, 2002). The support of preschool teaching staff has demonstrated importance for sustaining parent attendance in Hong Kong families (Leung, Tsang, & Dean, 2012).
For program fidelity, a session checklist on program content and activities was developed and the facilitators were required to complete the checklist for each session. The authors conducted visits to the program and used the checklist for fidelity check. Peer observation among the facilitators was also conducted using the checklist.
The Present Study
The present study aimed to examine the efficacy of the HOPE-20 using cluster RCT design. Based on HOPE and HSHVP, the hypotheses were:
Method
Design and Setting
This study adopted parallel cluster RCT design. Randomization was at nursery school level. As parents within the same nursery school might share information about the intervention, randomization within nursery schools might lead to contamination. Further, there might not be enough parents within each nursery school to be randomized into intervention and control groups. Allocation ratio was 1:1. The program was delivered in nursery schools.
Participants and Procedures
The inclusion criteria were (i) the target child should be attending nursery school, (ii) the target child should be between the age of 2 years to 2 years 11 months at the commencement of the study, and (iii) the participating parent should be residing in Hong Kong with the target child. Children with developmental disabilities were excluded. As the first efficacy trial of this universal program, it was decided to test the program on children with no known developmental disability. All local preschools with nursery class provisions registered with the Social Welfare Department/Education Bureau were eligible to participate.
For a medium effect size (d = 0.5, power = 0.80, α = .05), a sample size of 64 per group was needed to test for differences in means between two groups (Cohen, 1992). For cluster RCTs, the sample size should be adjusted according to the design effect. The average intracluster correlation (ICC) for parent and child outcomes was .0265 in the present study and the average cluster size was 9.6. Thus, the design effect was 1.23 (design effect = 1 + [cluster size −1] × ICC). The sample size required was 158 (Campbell, Mollison, Steen, Grimshaw, & Eccles, 2000). The average ICC for teacher reports was .277 (design effect = 3.38) and the required sample size was 433. The ICCs were comparable to cluster RCTs with school as the unit of randomization (e.g., Malti et al., 2011).
All nursery schools registered with the Social Welfare Department/Education Bureau were eligible to enroll. Information about HOPE-20 was distributed by the investigators to all eligible schools through e-mail, and schools were informed that they would be randomized into intervention and waitlist control group. Schools voluntarily enrolled in HOPE-20 through e-mail. Upon the enrolment deadline, 18 nursery schools enrolled. The participating schools were then randomized into intervention (10 nursery schools) and waitlist control (control) groups (8 nursery schools) by the first author, using a random number table, without blocking. Each school was informed of its status, and they sent letters to all parents of nursery grade children in the school to invite them to join. Participation rate was 28.6% in the intervention and 27.6% in the control schools. The mean parent group size in the nursery schools was 9.6 (range: 5–19).
There were 105 parents (110 children) in the intervention group and 59 parents (63 children) in the control group. Among the intervention group participants, there were six families where the spouse participated as an observer for at least 50% of the sessions (13–19 sessions). There were 10 participants whose spouses attended between one to four sessions as observers.
One week before program commencement, upon parent consent, the intervention group children were assessed in their schools on preschool concepts and language skills (both being direct assessment of children), by trained research assistants (psychology graduates) and postgraduate psychology students, one of whom being the third author. None of them were involved in program delivery. The parents and teachers were given the questionnaires for completion and return by the first session of the program. At the second last session, the parents and teachers were given the post-intervention measures for completion and return in the last session. The children were assessed within 1 week of program completion (direct assessment by trained research assistants and postgraduate psychology students including the third author, with none involved in program delivery) in their own nursery schools. For the control group, upon parent consent, the children, parents, and teachers completed the two waves of assessment within a 5-month interval, which corresponded to the length of the 20-week HOPE-20 program. The two waves of control group parent and children assessment were conducted within the same month of the intervention group pre- and post-intervention assessment, respectively. The research assistants and the postgraduate psychology students (including the third author) who conducted the child assessment were not informed of the group status of the children they assessed. The third author did not conduct assessment for children in nursery schools, where he had conducted session observations for fidelity checks.
The control group parents were offered HOPE-20 after the intervention group parents had completed the program. While the intervention group parents were attending HOPE-20, no extra services were offered to the control group parents.
To find out whether the intervention group parents could maintain the gains after program completion, they were requested to complete the parent ratings on parenting, child behavior, and learning 3 months after program completion. A reunion meeting was organized and parents completed the questionnaires on site. For those who did not attend the reunion meeting, the questionnaires were sent to them by post and they could return the completed questionnaires to the investigators using reply-paid envelopes supplied. As the children had left the nursery schools to attend other kindergartens by the time of the reunion meeting, there were logistic problems in seeking the consent and collaboration of the kindergartens, which were not involved in the program initially. Thus, direct assessment of children and teacher reports were not obtained for the follow-up data. As the control group parents had started the HOPE-20 program by this stage, such data were not collected from the control group.
For program fidelity, facilitators were required to complete a session checklist on program content and activities for each session. The three authors conducted several visits to the program and used the checklist for fidelity check, as a measure of independent assessment. Furthermore, the three facilitators who delivered the program conducted several rounds of peer observation using the fidelity checklist.
This study was approved by the ethics committee of The Hong Kong Polytechnic University. Recruitment commenced in November 2012 and intervention commenced in January 2013. Collection of follow-up data was completed by November 2013.
Measures
To enhance the validity of findings, multiple method and multiple informants were used for data collection. These included direct assessment of children’s preschool concepts and language skills by trained research assistants and postgraduate psychology students, parent rating on their own parenting, parent rating on their children’s learning and behavior, as well as teacher rating on children’s learning. All measures were collected at pre-intervention and post-intervention for all parents/children. In addition, all parent-rating measures were collected for intervention group parents at 3-month follow-up.
Child learning
Cognitive Subscale of Preschool Developmental Assessment Scale (PDAS-Cognitive)
This scale was designed for Chinese children aged 39–75 months (Leung, Mak, Lau, Cheung, & Lam, 2013; Liu, 2013). It correlated with the Wechsler Preschool and Primary Scale of Intelligence–Revised. The current extension included items suitable for 2-year-old children such as colors, shapes, body parts, and quantity. This extension correlated positively with the Griffiths Mental Development Scales–Extended Revised (GMDS-ER; Luiz et al., 2004). It could differentiate 2-year-old children from 3-year-old children and children with developmental disabilities from children with typical development. In the validation study, the reliability (Kuder-Richardson [KR]-20) was .92 and test–retest reliability was .89 (Liu, 2013). This test was administered to children individually. Children were shown stimulus pictures and they had to point to the correct answer or provide answers verbally. The scale consisted of 34 items. Correct responses were scored as “1” and incorrect responses were scored as “0.” Higher scores indicated higher attainment of preschool concepts.
Language Subscale of PDAS (PDAS-Language)
This scale was designed for Chinese children aged 39–75 months (Wei, 2013; Wong, Leung, Siu, & Lam, 2012). It correlated with Hong Kong Oral Language Assessment Scale (T’sou et al., 2006) and Reynell Developmental Language Scales–Revised (RDLS-R; Reynell & Huntley, 1985). The present extension for 2-year-old children could differentiate 2-year-old children from 3-year-old children and children with developmental disabilities from those with typical development. It correlated with the comprehension subscale of RDLS-R and GMDS-ER language subscale. In the validation study, the reliability (KR-20) was .90 and test–retest reliability was .89 (Wei, 2013). This test was administered to children individually. Children were shown stimulus pictures and they had to point to the correct answer or provide answers verbally. The scale consisted of 41 items. Correct responses were scored as “1” and incorrect responses were scored as “0.” Higher scores indicated higher attainment of language skills.
Behavior academic competence (BAC)
This was based on the Behavior Rating Scale of Presented Academic Self-Esteem in Young Children (Fuchs-Beauchamp, 1996; Lin, 2013). It consisted of 13 questions on sense of competence and coping, with parent and teacher versions. Teachers and parents were to rate their students/children on each statement on a 4-point scale (0 = never, 1 = seldom, 2 = sometimes, and 3 = always). Both versions correlated positively with GMDS-ER (Luiz et al., 2004) and could differentiate children with typical development from children with developmental disabilities. In the validation study, the reliability estimates (Cronbach’s α) of the teacher and parent versions were .94 and .84, and the test–retest reliability estimates were .86 and .60, respectively (Lin, 2013). In the current study, parents and teachers each completed this measure for the target child. Higher scores indicated higher academic competence.
School Readiness Inventory
This scale was developed to assess the school readiness of children entering first grade, to be rated by preschool teachers (Gumpel, 1999). The Chinese version was adapted for 2-year-old and 3-year-old children, with 9 items on a 4-point rating scale from 0 (never behaves in this way) to 3 (always behaves in this way), with a parent version and a teacher version. Both versions could differentiate between 2-year-old and 3-year-old children as well as children with developmental disabilities and children with typical development. Both teacher and parent versions correlated positively with GMDS-ER (Luiz et al., 2004). In the validation study, the reliability estimates (Cronbach’s α) for the teacher and parent versions were .91 and .83, while test–retest reliability estimates were .88 and .62 (Li, 2013). In the current study, parents and teachers each completed this measure for the target child. Higher scores indicated higher school readiness.
Effort and task motivation
This instrument was based on the Inventory of School Motivation (Lai, 2013; McInerney & Ali, 2006) and consisted of a parent version and a teacher version. The Effort Subscale measured effort attribution and effort-focused motivation and consisted of 6 items. The Task Subscale measured task orientation and involvement and consisted of 4 items. In this Chinese version, teachers and parents were to rate their students/children on each statement on a 4-point scale (0 = never, 1 = seldom, 2 = sometimes, and 3 = always). Both teacher and parent versions of the Effort and Task Subscales correlated positively with GMDS-ER (Luiz et al., 2004). Both teacher and parent versions of the Effort and Task Subscales could differentiate children with developmental disabilities from those with typical development. In the validation study, for the Effort Subscale, the reliability estimates (Cronbach’s α) of the teacher and parent versions were .93 and .88, whereas the test–retest reliability estimates were .74 and .43. For the Task Subscale, the reliability estimates (Cronbach’s α) of the teacher and parent versions were .88 and .83, whereas the test–retest reliability estimates were .70 and .52 (Lai, 2013). In the current study, parents and teachers each completed these two subscales for the target child. Higher scores indicated higher effort and task motivation.
Child behavior
Eyberg Child Behavior Inventory (ECBI)
This was a 36-item measure of parent perception of child disruptive behavior, with two scales, to be completed by parents (Eyberg & Ross, 1978). The Intensity Scale was a measure of frequency of disruptive behavior rated on a 7-point scale. The Problem Scale measured the extent to which the parents were troubled by children’s behavior and was rated as yes (1) and no (0). In the Chinese validation, it correlated positively with the Child Behavior Checklist and negatively with the Parenting Stress Index, with reliability of .94 and .93 for the two scales (Leung, Chan, Pang, & Cheng, 2003). In the current study, parents completed this measure. Higher scores indicated more frequent child behavior problems and more parent concern over child behavior problems.
Strength and Difficulty Questionnaire (SDQ)
The SDQ was a brief behavioral screening questionnaire for children and adolescents aged 4–16 years (Goodman & Scott, 1999). In terms of its validity, the SDQ scores correlated with the Child Behavior Checklist and it could discriminate between psychiatric and dental cases (Goodman & Scott 1999). The SDQ consisted of five subscales (each with 5 items): Emotional Symptoms, Conduct Symptoms, Hyperactivity Symptoms, Peer Problems, and Prosocial Behavior. In the present study, parents completed this measure by rating each item on a 3-point rating scale from 0 (not true) to 2 (certainly true), with higher scores showing higher endorsement of the behavior domain. A Total Problem Behavior score could be computed by adding the raw scores from the subscales of Emotional Symptoms, Conduct Symptoms, Hyperactivity Symptoms, and Peer Problems. The Chinese version of the scale has been validated by Lai et al. (2010). It has been used with preschool children in Hong Kong (Leung, Chan, et al., 2003). In the current study, only the Prosocial Behavior Subscale and Total Problem Behavior scores were used. The former was used as a measure of appropriate behavior with higher scores indicating higher prosocial behavior, while the latter was used as a summary of problem behavior, with higher scores indicating more problem behavior.
Parenting
Parental Stress Scale
This 17-item scale measured the stress of parenting on a 6-point rating scale (Cheung, 2000). The Chinese version was validated by Leung and Tsang (2010). It correlated with the Parenting Stress Index and ECBI. It could distinguish between parents of children with attention-deficit hyperactivity disorder and children without special education needs. The reliability in the validation study was .86 (Leung & Tsang, 2010). In the present study, parents completed this questionnaire. Higher scores indicated higher parental stress.
Duke-University of North Carolina (UNC) Functional Social Support Questionnaire
This was an 8-item questionnaire on perceived social support in various areas and was used as a measure of parents’ social networks (Broadhead, Gehlbach, de Gruy, & Kaplan, 1988). Participants rated their perceived degree of support on a 5-point scale. The Chinese version of the questionnaire has been used with Chinese new-immigrant parents of preschool children with satisfactory reliability (.84; Leung, Leung, & Chan, 2007). In the present study, parents completed this questionnaire. Higher scores indicated more perceived social support.
Parenting Sense of Competence Scale
This 16-item questionnaire assessed parents’ views of their competence as parents on a 6-point scale (Gibaud-Wallston & Wandersman, 1978). The Chinese version was validated by Ngai, Chan, and Holroyd (2007). In the validation study, the reliability was .85 (Ngai, Chan, & Holroyd, 2007). In the present study, parents completed this questionnaire. Higher scores indicated higher sense of competence.
The Intervention
HOPE-20 was delivered to participating parents in nursery schools, in groups of about 10 parents. HOPE-20 was a structured program with a detailed facilitator’s manual spelling out the content of every session, PowerPoint slides for lecture to parents, bring-home notes for parents, and daily homework sheets. The program consisted of 20 2-hour weekly sessions delivered by trained social workers who developed the program under the supervision of the first and second authors. The third author drafted the content of three sessions under the supervision of the first and second authors, but he was not involved in program delivery. The three authors were not involved in the direct delivery of the program. Each session began with a review of the previous session and homework, followed by a mini-lecture on the topic to be covered, then explanation and demonstration of the homework for the coming week by the facilitator, and finally, role play by participating parents to practice the homework, with feedback by the facilitator. The details of session content are in Table 1.
Data Analysis
Missing post-intervention data were estimated using multiple imputation (five imputations) with multivariate normal regression based on all pre-intervention measures. The imputation was conducted without incorporating information on clusters because it has been found that standard multiple imputation (ignoring clustering) could produce valid results in cluster randomized trials, when compared to methods which incorporate cluster information such as regression models in conditions similar to this study, in terms of number of clusters and ICC (Taljaard, Donner, & Klar, 2008).
Analysis was by intention to treat. Hayes and Moutlon (2009) recommended the use of cluster level data for analysis with small samples, but Galbraith, Daniel, and Vissel (2010) maintained that with this method, individual observations would be lost during the process. They recommended analyses based on individual-level data using linear mixed model regression and demonstrated that this approach could be used reliably with nine clusters and a sample size of 48. In this study, following the advice of Galbraith et al. (2010), linear mixed model regression was used, so individual observations would not be lost. The fixed factors were group status and pre-intervention measures. The random factor was nursery school. The dependent variables were the post-intervention measures. SPSS (Version 21) was used for data analysis.
Effect size was presented in terms of Cohen’s d. Both adjusted and unadjusted estimates were provided. Adjusted effect sizes were calculated using pooled predicted post-intervention score values from mixed method regression, and unadjusted effect sizes were calculated using pooled post-intervention values from multiple imputation. For outcome variables where the intervention group post-intervention scores were hypothesized to be lower than the control group, the numerator was calculated as mean control group post-intervention score minus mean intervention group post-intervention score. For outcome variables where the intervention group post-intervention scores were hypothesized to be higher than the control group, the numerator was calculated as mean intervention group post-intervention score minus mean control group post-intervention score. The unadjusted effect sizes are presented in Table 3.
Reliable change “determines whether the magnitude of change for a given client is statistically reliable” (Jacobson & Truax, 1991, p. 12). This is determined by the difference between the pre-intervention and post-intervention scores divided by the standard error of the difference between the two scores. Logistic regression with random effects was used to analyze the outcomes measures, where there were significant differences between intervention and control groups. The independent variable was group status and the dependent variable was reliable change status (achieved = 1 and not achieved = 0), and the random factor was nursery school.
To test treatment maintenance effect, the pre-intervention, post-intervention, and 3-month follow-up parent-rating measures of the intervention group were analyzed using repeated measures analysis of variance (ANOVA). Post hoc tests (Bonferroni adjustment) were conducted, where the overall ANOVA result was significant to find out whether the follow-up results differed significantly from the pre- and post-intervention measures.
As there were multiple outcomes, the correlations among outcome variables were examined to check for multicollinearity. According to Tabachnick and Fidell (2007), statistical problems created by multicollinearity occur, where correlation is above .90. In the present case, all correlations among outcome variables were less than .90.
Results
Ten intervention group parents (11 children) did not complete the program mainly due to work or childcare reasons. Post-intervention data were collected from them where possible, with 107 parent questionnaires and 108 child assessment completed. For the control group, eight parents (six children) refused to join post assessment, with 55 parent questionnaires and 57 child assessment completed (see Figure 1 for the flow of parents/children through each stage of the study). For attendance, 80 (76.2%) of the 105 intervention group parents attended 15 or more sessions. For fidelity, a checklist covering the core topics of each session was developed and all facilitators had to check their coverage of core topics for each session using the checklist. All facilitators covered all aspects of the session checklist in every session.

Flow of participants.
There were no significant differences in pre-intervention measures or demographic characteristics between parents/children with complete and incomplete data, except a significant difference in terms of sex of target children, χ2(1) = 4.18, p = .041. There was a higher percentage of boys (n = 9, 81.8%) among those with incomplete data than that in those with complete data (n = 81, 50.0%). There was a significant difference between the intervention and control groups in terms of parents/children with complete data, χ2(1) = 6.69, p = .010, with a higher percentage of parents/children in the control group (n = 8, 12.7%) with incomplete data than that in the intervention group (n = 3, 2.7%). For nine cases where a parent had two target children, the parent completed a separate questionnaire for each target child. Due to the small number of cases, the parent factor was not treated as a cluster.
There was no significant difference between the intervention and control nursery schools in terms of school/class size or median household income of the districts of the schools. There was no significant difference in baseline measures (demographic characteristics and pre-intervention measures) between the intervention and control groups (see Table 2), indicating that randomization was effective. The mean post-intervention and 3-month follow-up scores, reliability estimates of measures, ICC, and unadjusted effect sizes were reported in Table 3. All reliability estimates were above .70 except SDQ–Prosocial Behavior scores.
Baseline Characteristics of Parents, Children, and Schools.
Note. PDAS = Preschool Developmental Assessment Scale; ECBI = Eyberg Child Behavior Inventory; SDQ = Strength and Difficulty Questionnaire.
aIntervention: 108, control: 63. bIntervention: 96, control: 61. cIntervention: 87, control: 49. dIntervention: 102, control: 61.
Reliability, Intracluster Correlation, Post-Intervention and Follow-Up (FU) Scores, and Effect Size for Treatment Effect.
Note. CI = confidence interval; PDAS = Preschool Developmental Assessment Scale; ECBI = Eyberg Child Behavior Inventory; SDQ = Strength and Difficulty Questionnaire; NA = Not applicable.
aIntervention: 102, control: 61.
Intention-to-Treat Analysis
Linear mixed model regression was used. The fixed factors were group status and pre-intervention measures. The random factor was nursery school. The dependent variables were the post-intervention measures. For child learning, the results were significant for child preschool concepts (PDAS-Cognitive; F = 9.82–14.48, p = .006–.001, d = 0.70, 95% confidence interval [CI]: [0.38, 1.01]) and language skills (PDAS-Language; F = 13.90–23.21, p = .002–p < .001, d = 0.98, 95% CI: [0.65, 0 1.30]), after adjusting for pre-intervention measures and school effect. Intervention group children attained higher achievement on preschool concepts and language skills than control group children at post-intervention. The results were not significant for parent rating of school readiness (F = 0.03–1.27, p = .866–.262, d = 0.36, 95% CI: [0.05, 0.67]), BAC (F = 0.05–0.28, p = .832–.600, d = 0.14, 95% CI: [−0.17, 0.45]), effort motivation (F = 0.88–2.70, p = .349–.102, d = 0.47, 95% CI: [0.15, 0.78]), and task motivation (F = 2.08–4.86, p = .152–.029, d = 0.43, 95% CI: [0.12, 0.74]) and teacher rating of school readiness (F = 1.78–2.82, p = .201–.113, d = 0.75, 95% CI: [0.42, 1.07]), BAC (F = 1.26–3.42, p = .278–.083, d = 0.67, 95% CI: [0.34, 0.99]), effort motivation (F = 1.06–2.76, p = .320–.116, d = 0.61, 95% CI: [0.28, 0.93]), and task motivation (F = 1.58–2.47, p = .227–.136, d = 0.57, 95% CI: [0.24, 0.89]), after adjusting for pre-intervention measures and school effect. With regard to child behavior, the results were significant for frequency of child behavior problems (ECBI-Intensity; F = 7.49–18.82, p = .016–.001, d = 0.67, 95% CI: [0.35, 0.99]), parent concern over child behavior (ECBI-Problem; F = 11.01–18.23, p = .005–p < .001, d = 0.63, 95% CI: [0.31, 0.95]), and SDQ-Total Problem Behavior (F = 17.72–22.97, p ≤ .001, d = 0.83, 95% CI: [0.51, 1.15]) but not for SDQ-Prosocial Behavior (F = 0.64–2.34, p = .424–.128, d = 0.26, 95% CI: [−0.05, 0.57]), after adjusting for pre-intervention measures and school effect. Intervention group parents reported lower child behavior problems at post-intervention, compared with control group parents. For parent outcomes, the results were significant for parental stress (F = 6.89–10.03, p = .018–.005, d = 0.71, 95% CI: [0.39, 1.02]) and parenting sense of competence (F = 10.86–13.57, p = .006–p < .001, d = 0.83, 95% CI: [0.50, 1.04]) scores, after adjusting for pre-intervention measures and school effect. Intervention group parents reported lower parental stress and higher parenting sense of competence than control group parents at post-intervention. The results were not significant for social support (F = 0.23–0.75, p = .640–.401, d = 0.26, 95% CI: [−0.05, 0.57]), after adjusting for pre-intervention measures and school effect.
For reliable change, logistic regression with random effects was used to analyze the outcome measures, where there were significant differences between intervention and control groups. The independent variable was group status and the dependent variable was reliable change status (achieved = 1 and not achieved = 0), and the random factor was nursery school. The results indicated significant differences between the intervention and control groups in terms of child preschool concepts (PDAS-Cognitive), F = 7.86–11.41, p = .013–.004; child language skills (PDAS-Language), F = 10.75–16.95, p = .005–.001; and ECBI-Intensity, F = 7.66–14.69, p = .015–p < .001. The percentages for achievement of reliable change (multiple imputation estimates) were 65.5% (n = 72) for child preschool concepts (PDAS-Cognitive), 56.4% (n = 62) for child language skills (PDAS-Language), and 18.2% (n = 20) for frequency of child behavior problems (ECBI-Intensity) in the intervention group. The corresponding figures for the control group were 39.7% (n = 25), 25.4% (n = 16), and 1.6% (n = 1). More children in the intervention group were able to achieve reliable change in increase in preschool concepts, language skills, and decrease in child behavior problems.
Parents and Children With Complete Data
Linear mixed model regression was used. The fixed factors were group status and pre-intervention measures. The random factor was nursery school. The dependent variables were the post-intervention measures. Among those with complete data, the linear mixed model regression results were similar to the intention-to-treat analysis. There were significant improvements in child preschool concepts (PDAS-Cognitive), child language skills (PDAS-Language), parent report of task motivation, and parenting sense of competence scores and decrease in frequency of child problem behavior (ECBI-Intensity and SDQ-Total Problem Behavior), parent concern over child behavior (ECBI-Problem), and parental stress, after adjusting for pre-intervention measures and school effect. For reliable change, logistic regression with random effects was used to analyze the outcomes measures, where there were significant differences between intervention and control groups. The independent variable was group status and the dependent variable was reliable change status (achieved = 1 and not achieved = 0), and the random factor was nursery school. The logistic regression with random effects results were again similar to that in the intention-to-treat analysis, except there was also a significant difference in parenting sense of competence, with more parents in the intervention group achieving reliable change.
Follow-Up Results
Among the intervention group parents, 89 (80.9%) returned complete follow-up data. Missing data were estimated using multiple imputation with pre-intervention and post-intervention scores as predictors. Only participants with complete post-intervention data (n = 107) were included in the multiple imputation, as it was not meaningful to have follow-up data estimated from estimated post-intervention data. There were 107 participants included in the analysis.
For child behavior, repeated measures ANOVA results were significant for ECBI-Intensity, F(2, 212) = 20.34–21.68, p < .001, ηp 2 = .16–.17; ECBI-problem, F(2, 212) = 25.79–27.44, p < .001, ηp 2 = .20–.21; and SDQ-Total Problem Behavior, F(2, 212) = 22.50–30.64, p < .001, ηp 2 = .18–.22, indicating significant differences across the time points. In all cases, post hoc tests (Bonferroni adjustment with family-wise error rate = .05) indicated that the post-intervention and follow-up measures were significantly lower than the pre-intervention measures, with no difference between the post-intervention and follow-up measures, suggesting that program gains in child behavior could be maintained at follow-up. For SDQ-Prosocial Behavior, repeated measures ANOVA results were significant, F(2, 212) = 9.62–11.57, p < .001, ηp 2 = .08–.10. Post hoc tests (Bonferroni adjustment with family-wise error rate = .05) indicated that the post-intervention and follow-up measures were significantly higher than the pre-intervention measures, with no difference between the post-intervention and follow-up measures, suggesting that program gains in child prosocial behavior could be maintained at follow-up. For child learning, repeated measures ANOVA results were significant for parent rating of BAC, F(2, 212) = 7.52–9.84, p ≤ .001, ηp 2 = 0.07–0.09; school readiness, F(2, 212) = 74.49–84.10, p < .001, ηp 2 = .41–.44; effort motivation, F(2, 212) = 6.67–7.41, p = .002–.001, ηp 2 = .06–.07; and task motivation, F(2, 212) = 9.94–11.83, p < .001, ηp 2 = .09–.10, indicating significant differences across the time points. For task motivation and school readiness, post hoc test (Bonferroni adjustment with family-wise error rate = .05) indicated that the post-intervention and follow-up measures were significantly higher than the pre-intervention measures, with no difference between the post-intervention and follow-up measures, suggesting that program gains in child learning could be maintained at follow-up. With regard to effort motivation and BAC, post hoc tests (Bonferroni adjustment with family-wise error rate = .05) indicated that the post-intervention measures were significantly higher than the pre-intervention measures, with no difference between the post-intervention and follow-up measures, suggesting that program gains in child learning could be maintained at follow-up. For parenting measures, repeated measures ANOVA results were significant for parenting sense of competence, F(2, 212) = 21.92–25.29, p < .001, ηp 2 = .17–.19, indicating significant differences across the time points. Post hoc test (Bonferroni adjustment with family-wise error rate = .05) indicated that the post-intervention and follow-up measures were significantly higher than the pre-intervention measures. There was no difference between the post-intervention and follow-up measures, suggesting that parents could maintain the program gains in parenting competence at follow-up. The overall results were significant for parental stress, F(2, 212) = 7.02–7.86, p = .001, ηp 2 = .06–.07, and social support, F(2, 212) = 6.42–8.05, p = .002–p < .001, ηp 2 = .06–.07, indicating significant differences across the time points. In both cases, post hoc tests (Bonferroni adjustment with family-wise error rate = .05) indicated that the post-intervention scores were significantly different from the pre-intervention scores, with no difference between the post-intervention and follow-up measures, suggesting that program gains in parenting could be maintained at follow-up.
Among participants with complete data (n = 89), the patterns of results were similar to that in the intention-to-treat analysis. For child behavior, repeated measures ANOVA results were significant for ECBI-Intensity, ECBI-problem, SDQ-Total Problem Behavior, and SDQ-Prosocial Behavior. In all cases, post hoc tests (Bonferroni adjustment with family-wise error rate = .05) indicated that the post-intervention and follow-up measures were significantly different from the pre-intervention measures, with no difference between the post-intervention and follow-up measures, suggesting that the program gains in child behavior could be maintained at follow-up. For child learning, repeated measures ANOVA results were significant for parent rating of BAC, school readiness, effort motivation, and task motivation, indicating significant differences across the time points. For BAC, school readiness, and task motivation, post hoc tests (Bonferroni adjustment with family-wise error rate = .05) indicated that the post-intervention and follow-up measures were significantly higher than the pre-intervention measures, with no difference between the post-intervention and follow-up measures. With effort motivation, post hoc test (Bonferroni adjustment with family-wise error rate = .05) indicated that the post-intervention measures were significantly higher than the pre-intervention measures, with no difference between the post-intervention and follow-up measures, suggesting that program gains in child learning could be maintained at follow-up. For parenting measures, repeated measures ANOVA results were significant for parenting sense of competence, parental stress, and social support, indicating significant differences across the time points. Post hoc test (Bonferroni adjustment with family-wise error rate = .05) results were similar to those in the intention-to-treat analysis for parental stress and social support. For parenting sense of competence, all measures (pre-intervention, post-intervention, and follow-up) were significantly different from one another. The results suggested that the parenting gains could be maintained at follow-up.
Discussion
This study aimed to examine the efficacy of the HOPE-20 program using cluster RCT design. The results indicated that compared with the control group, the intervention group showed significant improvement in post-intervention child learning and parenting sense of competence and decrease in child behavior problems and parental stress.
Hypothesis 1 on children’s learning was supported. Intervention group children achieved higher post-intervention scores on preschool concepts (PDAS-Cognitive) and language skills (PDAS-Language) than control group children, with medium effect sizes (unadjusted: 0.54–0.74 and adjusted: 0.70–0.98). These were consistent with and even slightly superior to the effect sizes for cognitive development and educational success (0.34 and 0.53) of a meta-analysis of early prevention programs for socially disadvantaged families (Manning, Homel, & Smith, 2010). The HOPE-20 program was a universal program, whereas the programs cited in the Manning, Homel, and Smith (2010) study were programs for disadvantaged families.
Hypothesis 2 on parent and teacher ratings of child learning was not supported. There was no significant difference in post-intervention scores between intervention and control groups. One explanation was that learning behavior might take longer to change, so teachers and parents might not have observed drastic changes at post-intervention. Another explanation was that the measures might not be sensitive enough to detect the differences, though they could differentiate delayed children from children with typical development.
Hypothesis 3 on child behavior was supported. Intervention group parents reported lower child behavior problems at post-intervention, compared with control group parents. The intervention group was able to maintain their gains at 3-month follow-up. The effect sizes were mostly in the medium range (unadjusted: 0.45–0.55 and adjusted: 0.63–0.83) and were consistent with the effect sizes (between 0.26 and 0.46) of a universal parent training program (Malti et al., 2011).
Hypothesis 4 on parent outcomes was partially supported. Among the intervention group parents, there was an increase in parenting sense of competence and decrease in parental stress at post-intervention, but there was little change in the control group. The effect sizes were mostly in the medium range (unadjusted: 0.48–0.51 and adjusted: 0.71–0.83). The parents were able to maintain their gains in parental stress and parenting sense of competence at 3-month follow-up. However, there was no significant difference between the two groups on social support. In a study of a universal parent training program, there was no significant change in parenting practices (Eisner, Nagin, Ribeaud, & Malti, 2012). One possible explanation for the lack of change in social support was that the parents were not recruited among socially isolated families. They might have their own established network of social support, and the social network in the present program might not impact their perceived support significantly.
In terms of attendance over the 20-session program duration, over 75% of the parents achieved at least 75% attendance. This provided evidence that the program dosage was acceptable to the parents. The delivery of the program in nursery schools and the support of school staff were probable reasons for the encouraging attendance. The results provided support for the strategy of delivery of universal programs in preschools.
The effect sizes were mainly in the medium range and were consistent with the average effect sizes of 0.42 on behavior problems and 0.45 on parenting practices reported in parent training programs (Kjobli & Ogden, 2012). For the outcomes with significant differences between the intervention and control groups, none of the effect size CIs included zero, which provided evidence for the robustness of the findings (Kjobli & Ogden, 2012).
There were some limitations to the present study. First, follow-up results included only parent ratings from the intervention group. This precluded attributing the results entirely to program effects. However, as the control group was a waitlist control group, there were ethical concerns about delay in offering the program to the control group parents. Further, due to logistic issues, direct assessment of children and teacher ratings could not be collected for follow-up measures. Second, the sample size was less than that required for teacher ratings, though the sample size was adequate for parent ratings and direct assessment of children. Third, although efforts were made to recruit nursery schools from various districts in Hong Kong, there were still districts where nursery school participation was absent. The participation rate of the schools was not satisfactory. In each participating school, the participation rate of parents was only about 30% and it was not possible to obtain information on the characteristics of parents who did not wish to participate. These might affect external validity. Fourth, the majority of parent participants were mothers. Gender-sensitive services and strategies need to be further devised. Fifth, three facilitators conducted different groups in different nursery schools and there might be facilitator effects. Facilitator effect was not investigated, as the facilitator effect was a cluster (school) level covariate and the number of clusters was small, and Hayes and Moulton (2009) specifically cautioned against such analysis. Sixth, the parent participants were only recruited by schools after the schools were informed of their intervention/waitlist control status. Parents who volunteered in the waitlist control status might have been more motivated, as they were willing to wait for the services, and this might have affected the internal validity of the study. There was, however, no significant difference between the intervention and control group participants in terms of demographic characteristics or pre-intervention scores on parenting, child behavior, and child learning. The participation rates in the intervention and control schools were similar. Finally, although parents could bring their spouses where possible, only one parent from each family would complete the outcome measures in the study. Out of 105 intervention group parents, the spouses of six participants attended at least 50% of the sessions. The spouses of 10 participants attended 20% or less of the sessions.
Discussion and Application to Practice
HOPE-20 was developed by the same researchers after HOPE and HSHVP. The efforts sustained a productive new prevention trend in Hong Kong in establishing locally developed, evidence-based early childhood parent education manualized programs suitable for wide use. This is a concrete response to international policy advocacies (e.g., Heckman, 2000) on the importance of early investment on children. HOPE-20 also successfully achieved to cater for a younger children group while reducing the number of sessions from 30 to 20, without compromising the outcome. This attempt to make a quality program even more compact best suits the fast pacing lifestyles of families in cities like Hong Kong. The success of this attempt should encourage more research to devise the optimal dosage for programs of similarly high quality.
Following the NICE guideline emphasis on adequate dosage and quality input, HOPE-20 ensures that the parents spend enough time on children, aside from polishing the quality of the parent–child interactions. The emphasis on mastering parenting strategies together with micro skills also ensures that the parents understand the rationale in parenting. Parents’ mastery of the rationale behind micro practices is particularly important because children grow and develop while family circumstances change. If the parents’ mastery of parenting principles allows them to generate age- and situation-sensitive skills to novel situations, the positive parenting impact of HOPE-20 can be sustained. In fact, when funding and collaborator support are available, future studies should assess the longer term maintenance of the program input, say for 1 or 2 years.
In terms of the cost, assuming that one social worker could conduct six HOPE-20 groups per week, it is estimated that one social worker can deliver 12 HOPE-20 groups within 1 year (each group lasting for 20 weeks). With about 10 parents per group, 120 parents and children should be able to benefit from the employment of one social worker for 1 year.
The successful accomplishment of HOPE-20 has positive implications for social work practice: Frontline practice on the parents: The project shows that social workers can competently deliver well-designed and manualized education programs for parents with children as young as 2 years to achieve intended outcomes. The social workers’ pre- and in-service mastery of the needed knowledge, attitude, and skills in parenting, conducting groups, and making presentations in groups are pivotal to the success. Frontline practice with community partners: In this project, social workers were the key coordinators in the long-term collaboration with their community partners, being preschool educational settings in this case. Strategies to engage their trust and cooperation to join and finish the program are also important aspects of the professional competence of social workers. Program development and adjustment: Making age-appropriate developmental adjustment from HOPE to HOPE-20, which serves a younger target group, was the concerted efforts of the lead psychologists and the supporting social work research team members. Their sensitivity to the parents’ needs and responsiveness to the client profile at the pre-assessment period and along the few months of program delivery enabled the session by session fine-tuning of the program materials to achieve the optimal effect. Their between-session support to parents who missed certain sessions, and their provision of corrective strategies to parents who had difficulties in completing homework, also greatly reduced parental stress and enhanced parents’ competence. Provision of stigma-free universal program to engage users: The fact that HOPE-20 is designed as a universal program for all offers a ready-to-use package for social work services like the IFSCs and Parents Resource Centers to position the service to attract a broad range of community users to begin to consider using public social services. This is particularly important in Chinese communities, where help-seeking is a taboo and that one should avoid revealing family weaknesses as far as possible. Given that it is widely accepted that all parents struggle with parenting and that parents who seek to learn and improve their parenting are seen to be better parents, the package with its very neutral and even positive image should be able to tone down some psychological hurdles in seeking help. Provision of systematic training for junior professionals: That HOPE-20 as a manualized and evidence-based program makes it a very useful package to train junior professionals, like social workers and even psychologists, who often have to deal with children and parents on child behavior and learning problems as well as parental stress issues. The systematic nature of the program with clear goals, steps, and examples provides a convenient bridge from compact fast-paced professional training to more detailed and practical application needed in actual practice. The program thus has great potentials to be used for family and child work staff orientation to equip them competently for their professional practice.
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The project was funded by Lo Ying Shek Chi Wai Foundation.
