Study design

This cohort study involved the comparison of two groups of families with young children in Grades 1–3 at the time of study initiation. All families in TFES were eligible and invited to participate in the HEP (referred to as "intervention families") and a comparison group was selected from families in Tung Chung and Kwai Chung who did not join the TFES. Kwai Chung is an emerging suburban community in Hong Kong with limited access to healthcare services and therefore has a similar sociodemographic background to Tung Chung [36]. Participants were recruited between July 2013 and March 2016. Families were recruited when they met the following criteria: (1) One or more family members were employed, either on a full-time or part-time basis, (2) had a child or children enrolled in a primary school program for grades 1 to 3, (3) had a monthly household earnings that did not exceed 75% of the median monthly household income in Hong Kong during that period. Each intervention and comparison family could have more than one child included in the study. Children in both groups had cognitive skills assessments by qualified clinical psychologists at enrollment. In addition, a comprehensive health evaluation and a phone survey were administered for their parents as a baseline and again after approximately 5 years as a follow-up. During the 5-year follow-up, intervention families were offered regular HEP activities and could participate if they wished. Meanwhile, families in the comparison group were not asked to participate the intervention activities.

Components of HEP

The HEP included intercalated routine health evaluations, self-management education, health knowledge talks, and health ambassador training. Among these, the first two types of activities were available to children participants together with their parents, while the latter two types of activities targeted the parents.

HEP activities for both children and parents

The annual health assessment included a telephone survey about health and health service use and an in-person clinical health assessment. In the telephone survey, parents answered the survey questions about their children’s health, healthcare service utilization, behavior, and HRQOL. The clinical health assessment of children was carried out by registered nurses, trained technicians, and research assistants. Children identified as having substantial health risks or abnormalities were offered counseling by a nurse or doctor from the project team or were directed to suitable services for additional management. The yearly count of individuals facing major health issues or risks, along with the referrals they obtained, can be found in Supplementary Table 2. Self-care enablement activities (see Supplementary Table 3) included workshops and training courses on nutrition, cooking and exercise, family gym, and hiking groups in which the children participated together with their parents. Participants in these activities reported significantly improved self-care enablement after engaging in these activities. Additionally, over 90% of participants expressed satisfaction and indicated they would recommend the activities to their family and friends. The self-care components also created a mobile app, FamilyMove, delivered to these parents and their children to provide coaching after the face-to-face training.

HEP activities for parents

In addition to annual health assessments and activities available to both children and parents, recurring health lectures and talks on typical issues detected in the health evaluations were delivered to parents to enable better self-care (see Supplementary Table 4). The participants’ understanding of the health topic was assessed before and after each seminar, revealing a significant improvement. Additionally, several parents received training to serve as leaders for nutrition and physical activity classes and to manage group exercises following the classes. These adults acted as health ambassadors for their acquaintances and families.

Outcome measures

The Chinese Strengths and Difficulties Questionnaire (SDQ) [37] was used to assess children’s behavioral issues. The SDQ is one of the most widely used questionnaire for child emotional and behavioral screening globally. The measure has been translated into over 80 languages and serves various purposes, including clinical assessment, outcome evaluation, research, and screening [38]. The SDQ was used to assess 5 domains of children’s behaviors, including emotional issues, behavioral misconduct, attention difficulties, socialization difficulties and prosocial conduct. Each of these domains has a 5-item subscale. The measure is valid and reliable in the context of Hong Kong [37]. A 3-point Likert scale, in which 0 = “did not apply,” 1 = “apply to a certain extent,” and 2 = “applied very much”, is used to record the participants’ answers. The overall score for each issue/behavior is computed by adding the scores of all items in the corresponding subscale. Thus, the score on each problem/behavior ranges from 0 to 10. The score of total difficulties is computed by adding up the scores on the four problem subscale scores [39] and ranges from 0 to 40. A greater value for the specific difficulty subscale and the total score of difficulties indicate more severe problems, while a higher score for prosocial behavior subscale is indicative of better prosocial behaviors.

The Chinese Child Health Questionnaire Parent Form 28 (CHQ-PF28) [40] was administered to assess children’s HRQOL. The CHQ-PF28 is a widely recognized pediatric health survey designed to evaluate HRQOL in children and adolescents aged 5–18. This measure has been developed and validated, demonstrating its effectiveness across various regions and populations. It has been utilized in over 600 peer-reviewed journal articles [41]. The measure has shown good validity and reliability in a Chinese setting [42]. It consists of 28 items, which cover 12 domains, including general health, physical functioning, emotional/behavioral limitations, physical limitations, bodily pain, general behavior, mental health, self-esteem, parent impact on emotion, parent impact on time, family activities, and family cohesion. The scores of these domains range from 0 to 100, which can be computed using a weighted summation algorithm detailed in the CHQ manual [43]. The aggregated scores comprise a physical summary score (PHS) and psychosocial summary score (PSS), which can be translated into norm-based scores where the mean score of the general population is 50, with a standard deviation of 10 [44]. A greater value indicates better HRQOL.

Baseline covariatesCognitive skills

The Wechsler Intelligence Scale for Children–Fourth Edition (WISC–IV) [45] was used to assess cognitive skills of children in both groups and the assessment was conducted by clinical psychologists. This instrument has shown good validity among Hong Kong children [46, 47]. The measure can generate an intelligence quotient (IQ) score, and a higher value represents a higher intellectual capacity.

Socioeconomic and health status

The following variables were also included as baseline covariates: children’s age, gender, body weight status, known doctor diagnosis of learning disability and chronic diseases, parents’ marital condition, monthly household revenue, and governmental comprehensive social security assistance (CSSA) scheme reception.

The SDQ, CHQ-PF28, and the other baseline covariates questionnaires were carried out by skilled interviewers, either in face-to-face or telephonic settings, and responses were provided by one or both parents of the children through self-reporting. The face-to-face interviews were administered by trained Research Assistants (RAs), and registered nurses. All interviewers received training from the project investigators and registered nurses on health assessment and survey administration, and their performance was monitored by the nurses during the initial data collection sessions. The telephone interviews were conducted by the Social Science Research Centre (SSRC) of the University of Hong Kong, a research center with over 10 years of experience in telephone surveys. SSRC staff received training in social science research methods, including phone survey techniques. All interviews adhered strictly to the interview protocol.

Statistical analysis

We used intention to treat and complete-case analysis in that all participants with baseline and follow-up assessments being included in the analysis. Stata version 16.0 (StataCorp LP, College Station, Texas) was employed for all statistical analyses. Statistical significance was determined using two-tailed tests, with p < 0.05 as the threshold for significance.

Descriptive statistics were employed to display the baseline characteristics of the participants. The p-value was used to evaluate the balance of baseline covariates between the two groups, where p > 0.05 indicated that the two groups were significantly different from each other. The change in each measure between baseline and follow-up was computed by subtracting the baseline score from the follow-up score. The statistical significance of the within-group change was assessed via paired samples t-test, and the effect size was calculated by Cohen’s d.

Multiple linear regressions were performed to investigate the independent impact of HEP on the changes in SDQ and CHQ-PF28 scores after 5-year follow-up, adjusting for covariates and baseline values. For each model, F-ratio and significance of the F-ratio were used to indicate the extent to which the predictors can explain the variance in the outcome variables in a statistically significant manner. The adjusted R2 was used to indicate the extent to which the variance in the predictor variances (as a whole) explains the variance in the outcome variable. Also, unstandardized B, p-value, and 95% confidence level were determined to show the influence of HEP on each outcome measure. Bonferroni correction was performed. The threshold for significance was adjusted by dividing the alpha level by the number of analyses, resulting in a threshold of 0.008 for the SDQ analyses and 0.025 for the CHQ-PF28 analyses. To provide a more detailed investigation, we repeated the above analyses to examine the effect of the HEP on changes in each of the 12 domains of the CHQ-PF28.

Subgroup analyses were further conducted based on baseline covariates. These analyses aimed to identify which groups benefited most from the intervention and to understand the impact of the intervention across different sociodemographic groups. Additionally, to evaluate the robustness of the study findings, a sensitivity analysis was performed by further adjusting for mothers’ sociodemographic and health status in addition to baseline covariates and values in multiple linear regressions. In addition, generalized estimating equations with exchangeable correlation structures were applied to account for the correlations of outcomes between children belonging to the same family in addition to baseline covariates and values.