Attrition

Fifteen children dropped out of the study, which is shown in Table 2. The dropping out rate did not differ by treatment group (χ2 = 0.01, p = 0.91). There was no significant difference in terms of sociodemographic and background variables at baseline between the retained sample (n = 555) and the dropout sample (n = 15). Dropping out wasn’t associated with the following factors at baseline: anxiety (GAD subscale) (t = − 1.12, p = 0.26) and depressive symptoms (MDD subscale) (t = − 0.68, p = 0.5).

Sociodemographic and background information

The descriptive statistics for the sociodemographic and background information by group at baseline are presented in Table 2. At baseline, no difference was found between the intervention and control groups except that fewer children in the intervention group lived with both parents (p = 0.01), and fewer children in the intervention group reported they had good relationship with their mothers (p = 0.003).

Table 2 Sociodemographic and background information at baseline by treatment group. N (%)Change in anxiety and depressionThe scores of anxiety and depression at three assessments

The descriptive statistics for anxiety and depression by treatment group at each assessment are presented in Table 3. At baseline there was no significant difference between the intervention and control group for anxiety, while the intervention group reported higher depression levels (t = − 2.33, p = 0.02).

Table 3 Descriptive statistics of anxiety and depression at three assessments by treatment groupIntervention effects on anxiety and depression

This part presents detailed results of two linear mixed models (LMMs) evaluating the intervention effects on anxiety and depression, with the pre− test depression scores controlled as a covariate. The report of the models in Table 4. followed the best practice guidance for LMMs (Meteyard and Davies [25]).

Table 4 Main and interaction effects in the linear mixed models of the intervention on anxiety and depression

There was no condition × time interaction effect (F = 0.22, p = 0.802) on anxiety, which suggested no difference between intervention and control groups of the changes in anxiety scores from Assessment 2 to 1, B = 0.24, 95% CI [− 0.49, 0.97], p = 0.516, or from Assessment 3 to 1, B = 0.17, 95% CI [− 0.56, 0.89], p = 0.656. There was no significant main effect of condition (intervention vs. control) (F = 0.8, p = 0.37). However, there was significant main effect of time (assessments) (F = 5.27, p = 0.005) on anxiety, which suggested children across the whole sample (both in intervention and control school) reported higher anxiety at Assessment 3, B = 0.51, 95% CI [0.05, 0.98], p = 0.03 than Assessment 1.

There was no condition × time interaction effect (F = 0.25, p = 0.78) on depression, which suggested no difference between intervention and control groups of the changes in depression scores from Assessment 2 to 1, B = 0.048, 95% CI [− 0.83, 0.92], p = 0. 915, or from Assessment 3 to 1, B = 0.29, 95% CI [− 0.58, 1.17], p = 0.514. There was no significant main effect of condition (intervention vs. control) (F = 2.65, p = 0.1) on depression. However, there was significant main effect of time (assessments) (F = 11.67, p < 0.001) on depression, which showed children of the whole sample reported higher depression at Assessment 3, B = 0.85, 95% CI [0.29, 1.41], p = 0.003 than Assessment 1.

To summarise, there was no intervention effect on anxiety or depression, and children both in intervention and control school reported higher anxiety and depression at Assessment 3 compared to Assessment 1.

Moderation analyses of the intervention effect on mental health

A series of linear models showed no interaction effect for short- or long- term effect on anxiety in terms of Condition × Gender, Condition × Age, Condition × Whether living with parents, Condition × Parents’ migration status, or Condition × other sociodemographic/background variables. However, a model showed “Condition × Whether living with parents” interaction short-term effect on depression, B = − 2.54, 95% CI [− 4.96, − 0.11], p = 0.04, which implied that children living with neither parent (compared with both parents) in the intervention school reported lower depression than the control school at immediate post-intervention compared to baseline. A model showed “Condition × Whether living with parents” interaction long-term effect on depression, B = − 2.96, 95% CI [− 5.76, − 0.15], p = 0.04, which implied that compared to children living with both parents, those living with neither parent in the intervention school reported lower depression than the control school at 5-month follow-up compared to baseline. A model showed “Condition × Relationship with father” interaction short-term effect on depression, B = − 2.53, 95% CI [− 4.73, − 0.34], p = 0.02, which implied that children having poorer relationships with fathers (compared to good) in the intervention school reported lower depression than the control school at immediate post-intervention compared to baseline.

Implementation fidelity of the programme

There were overall 145 activities conducted in the programme. A summary of the number and percentage of activities fully implemented, partially implemented and not implemented is shown in Table 5, which compares programme fidelity assessed by programme facilitators and coordinators. According to the facilitators, the fully implemented rate was over 90% throughout the six grades, which was very similar to the coordinators’ assessment, except for the first grade. This indicated that the implementation fidelity was high, contributing to the internal validity of the study.

Table 5 Implementation fidelity assessed by facilitators and coordinators. N (%)Qualitative interviews

All 8 school teachers from 2nd -6th grade in the intervention school were interviewed. The average age of the interviewed teachers was 38 years (range 26 to 52), seven of the school teachers had received college education and one high school education. Nine caregivers, five mothers and four grandmothers volunteered to participate in interviews. The average age of the caregivers was 44 years (range 35 to 57). Six caregivers had secondary school education, and three had primary school education. Six were homemakers and three worked part-time. We planned to randomly select 30% of children in the intervention school as interviewees, which is sufficient to reach data saturation. In the end we interviewed 36% of children because many children volunteered to be interviewed. Eighty-three children (36 boys and 47 girls) aged 8–12 years were interviewed, and 33 (40%) were left-behind children.

Two main themes were identified from the interviews. Following the SEL programme (1) Children could cope with negative emotions better. (2) Children were more likely to communicate thoughts and feelings. The results of the interviews informed a conceptual framework, which is shown in Fig. 1.

Fig. 1

Conceptual framework of the SEL intervention on change of emotion management and communication

Theme 1 Children could cope with negative emotions better.

Over one quarter of the children (n = 23) reported they learned how to cope with negative emotions. The most commonly used coping mechanisms described were taking a walk, skipping, running, watching cartoons, and painting. A boy age 11, reported “When I feel nervous now, I take a deep breath; when I am angry or sad, I think about something happy, or I do something I like.” A girl age 10, said “When I am unhappy, I write it down and then tear it up, which helps me forget.” Three children learned to seek help from others. A boy age 12, stated “Before, if I failed an exam and felt down, I didn’t know what to do, but now I talk to my parents, friends or go to a quiet place yelling out the pressure, which is really helpful.” Meditation practiced in the SEL classes was also used by three children. A girl age 11, described “When I feel sad, I listen to comforting music, close my eyes and imagine some peaceful images as the SEL teacher told me. This makes me forget the negative thoughts.”

Two teachers reported that the SEL classes had helped children control their emotions. The 2nd grade teacher said “One of the boys used to lose his temper and became aggressive with other children. Now after the SEL classes, when he gets angry, I ask him what the SEL teacher taught him. Now he takes a deep breath and calms down.” The 3rd grade teacher reported “After the SEL classes, children could control their emotions better, and are less likely to cry or shout.”

Three caregivers reported that after the SEL classes children were less likely to lose their tempers. A mother of three children revealed “My first daughter had a hot temper before and always lost control, crying and shouting over small issues. But she is much better now. She calms down and listens to me.” Children learnt some skills to manage emotions. A mother of two daughters stated “When my daughters feel angry, I ask them what the SEL teacher taught them. They take a deep breath, which makes them calm down and then they can solve the problem.” Two caregivers reported that children dealt with negative feelings better than before. The grandmother of a boy age 10, reported “I know that my grandson is often unhappy, but he is more positive after the SEL classes.” Two caregivers reported children had fewer quarrels with them after the SEL classes. The grandmother of a girl age 10, said “My granddaughter had a hot temper before, always lost control and we always had quarrels. However, nowadays she doesn’t shout when we have arguments. Now she is polite.”

Theme 2 Children were more likely to communicate thoughts and feelings.

Nineteen children reported that since the SEL classes, they could express themselves better and show kindness to others. A girl age 10, said “In the past, when my elder sister came home, I just sat there and didn’t say anything. But now I talk to her and show I am glad to see her.” Four children reported that the SEL classes improved relationships with their families after they learned to communicate better. A boy age 9, stated “Because of the SEL classes, I know how to comfort my mother when she is sad. So my mother talks to me more than before and we are getting closer. I am really happy about this.” A boy age 10, told “In the past, if I was feeling down, I kept it to myself and didn’t talk to others. But now I talk to my mother, and she listens carefully. This has made our relationship closer. My mother also thinks the SEL classes are very good for this.” Five children commented that communicating more and not keeping things to oneself brings them more friends.

Four teachers revealed children were better able to express themselves after the SEL classes. A 5th grade teacher reported “The SEL classes have made children happy and relaxed, and encouraged children to express themselves. They are now more likely to speak their thoughts and feelings. Some children never answered my questions in class and didn’t talk to others before, but they do now.”

Two caregivers stated children started to talk more after the SEL classes. The mother of a girl age 12 reported “My daughter really changed a lot in personality since the SEL classes. In the past, she was extremely introverted and didn’t tell me about what was going on unless I asked her, but now she tells me even if I don’t ask. She told me she wrote letters to the SEL teacher and always got reply, which helped her a lot.” Two caregivers said they had closer relationships with their children after the SEL classes. The mother of a boy age 9 stated “When we are together now, my son always has something to say to me. He never shared much about his life before the SEL classes. In the past, he only talked when I asked him. This has helped me to get to know him better. We are closer now.”

In summary, following the SEL programme, children’s emotion management skills improved, thus they showed self-control and had fewer arguments with peers and family members. Children were also better able to communicate their thoughts and feelings, which may help to improve relationships. While only a limited number of children (just one quarter) expressed the changes.