Follow-up and Mediation Outcomes of a Movement-Based Mental Health Promotion Intervention for Refugee Children in Uganda

ABSTRACT

Objective

There is limited evidence for the effectiveness of mental health promotion interventions in low- and middle-income settings, especially for longer-term benefits. This study evaluates the 5-month follow-up outcomes of a movement-based mental health promotion intervention (TeamUp) for refugee children in Northern Uganda (West Nile) and further investigates what explains longer-term benefits.

Methods

This quasi-experimental study was conducted in four primary schools, randomly allocated to an experimental or a control condition. Enrolled in the study were 549 children ages 10–15 years (n = 265 TeamUp; n = 284 control group). Primary outcomes were psychosocial well-being (Stirling Children’s Wellbeing Scale), attitude toward school, and satisfaction with friendships (Multidimensional Students Life Satisfaction Scale [Friends and School subscales]). Mediators included social connectedness and sense of agency.

Results

At 5 months postintervention, 467 (85.1%) children were retained. Intent-to-treat analyses demonstrated sustained benefits for TeamUp for well-being (estimated mean difference −1.88, 95% CI −3.14 to −0.66, p = .003, effect size Cohen’s d = 0.25) and friendships (−1.52, 95% CI −2.55 to −0.48, p = .005, d = 0.25). There were no significant between-group differences for attitude toward school. Secondary benefits were shown for traumatic stress (2.18, 95% CI 0.45 to 3.91, p = .014, d = 0.21), quality of life (−1.29, 95% CI −2.31 to −0.30, p = .014, d = 0.21), bullying (0.53, 95% CI 0.08 to 0.97, p = .020, d = 0.20), and depression symptoms (1.31, 95% CI 0.09 to 2.52, p = .035, d = 0.18). Increased sense of connectedness mediates the effect of TeamUp on improving well-being (indirect effect = 0.30, SE = 0.13, p = .001), explaining 15% of variance.

Conclusion

This study shows sustained benefits of a mental health promotion intervention 5 months postintervention. Prolonged benefits are explained by an increase in social connectedness.

Introduction

The negative mental health impact of war, migration, and displacement on children and adolescents is both commonsensical and well established through research (Attanayake et al., 2009; Charlson et al., 2019). However, due to a shortage of mental health professionals in most low- and middle-income countries (LMIC), the need for mental health care services largely goes unmet. Having nonspecialists trained in mental health services, an approach that is called task-shifting, aims to address the gap between need and availability of mental health care. As a result, we have recently seen a burgeoning of new psychological interventions implemented by nonspecialists. The evidence base for such task-shifted psychological interventions for children and adolescents is growing (Purgato, Gastaldon, et al., 2018; Venturo-Conerly et al., 2022). However, this evidence base comes with some shortcomings. First, randomized controlled trials with the highest degree of scientific rigor are still scarce and tested psychological interventions highly heterogeneous (Uppendahl et al., 2020). Second, psychological interventions focusing on children and adolescents in humanitarian settings show less convincing results. Meta-analyses for this group of interventions have demonstrated effectiveness for reducing posttraumatic stress disorder (PTSD) symptoms as an outcome and not for other common mental health conditions such as depression, anxiety, and behavioral disorders (Barbui et al., 2020). Third, the large majority of RCTs (randomized controlled trials) evaluate treatments. While such treatment is very much needed, relying on treatment alone is insufficient and not cost-effective to achieve population-wide improvements in children’s mental health. Rather, a continuum of mental health promotion, prevention, and treatment is increasingly being advocated (Patel et al., 2018). Finally, there is a scarcity of evaluation of longer-term effects of interventions for children (Purgato, Gross, et al., 2018).

The World Health Organization (WHO) defines mental health promotion as interventions that improve or optimize positive mental health (WHO, 2023). Mental health promotion aims to address determinants of positive mental health and increase protective factors before the onset of symptoms or a disorder, thereby, improving population-level mental health and well-being. Moreover, mental health promotion interventions are characterized by the delivery platform in natural settings where people work, live, or learn (i.e., schools, the workplace, or at home). The distinction between mental health promotion and (universal and selective) prevention interventions is that the primary outcomes of the former are the increase of positive mental health or protective factors; whereas, the primary focus of the latter is the prevention of future disorders.

For children, mental health promotion is commonly addressed in school-based programs. There is empirical support for the benefits for such school-based mental health promotion interventions in both high-income countries (Wells et al., 2003) and, albeit less convincing, in low- and middle-income countries (Barry et al., 2013). Interventions typically include a focus on problem solving and coping skills, physical activity, classroom climate, improving social and emotional skills, conflict resolution, encouraging cooperation, and life skills. Based on systematic reviews of mental health promotion interventions in LMIC (Barry et al., 2013; Purgato et al., 2020), there still is a need for (i) a stronger evidence base given some mixed results, mixed quality of studies and heterogeneity of interventions, (ii) assessment of medium- to long-term effects of mental health promotion interventions, and (iii) understanding how mental promotion interventions contribute to (sustained) changes. The current study aims to address these questions, in relation to the mental health promotion intervention called TeamUp. In responses to these needs, the World Health Organization has developed guidelines specifically for mental health promotion and prevention interventions for adolescents in LMIC (i.e. Helping Adolescents Thrive) (WHO, 2020).

The need for better understanding of how interventions contribute to change is not exclusively relevant for mental health promotion interventions. With a reliance on task-shifting mental health interventions that are characterized by their brevity and often transdiagnostic, it is important that we know how we can fine tune and optimize results in resource-scarce settings. Yet, a recent report by the WellcomeTrust concludes that we have too little evidence about what helps prevent or treat youth anxiety and depression (WellcomeTrust, 2021). There is a need for more research into an intervention’s mechanisms-of-action (or key ingredients)—that is, those aspects of an intervention that drive the effect are theoretically defined and link to specific hypothesized outcomes. While dismantling studies is still rare, several investigations into mediational pathways of psychological interventions have been published. These either include studies to understand (i) what aspects of an intervention leads to outcomes or (ii) how an intervention leads to outcomes. An example of the former is a study of Group Problem Management Plus (gPM+) in Nepal that demonstrated the level of adoption of gPM+ specific therapeutic techniques that partially mediated (31% of variance explained) the outcomes in a cluster of randomized controlled trials (Jordans et al., 2021). An example of the latter is a study of the same intervention in Jordan, in which investigators demonstrated that for refugees participating in gPM+ results included a reduction in disciplinary parenting, which mediated a decrease in their children’s attentional and internalizing problems (Bryant et al., 2022).

TeamUp is a movement-based mental health promotion intervention developed for children affected by war and displacement, aimed to improve the psychosocial well-being of children ages 6–17 years. The largely nonverbal, or language-agnostic, intervention combines movement-based group games, dance and creative movement, physical activity, routines, songs, body awareness, and breathing practices (Bleile et al., 2021). A quasi-experimental study among children (n = 549) in refugee and host communities in Uganda aimed to evaluate TeamUp when compared to care as usual. Prior reporting on this study found that at postintervention, TeamUp participants saw significantly better outcomes on psychosocial well-being, attitude toward school, traumatic stress symptoms, perceived quality of life, social connectedness, and sense of agency—though not on bullying, quality of friendships, irritability, and depression symptoms (Bleile et al., 2024). The current paper reports on the 5-month follow-up of this quasi-experimental study. In addition, we aim to assess whether social connectedness and a sense of agency mediate outcomes at follow-up, using a sequential mediation model.

Methods

Design

This two-armed quasi-experimental study was conducted in a refugee settlement in West-Nile, Northern Uganda. The study was conducted in four schools, randomly allocated to either arm (1:1). With approximately 1.5 million refugees, primarily from South Sudan and Democratic Republic of Congo, Uganda is the largest refugee-hosting country in Africa. This study was conducted in the Bidi-bidi settlement and implemented by War Child Alliance, an international humanitarian nongovernmental organization. Details of the study and main results have also been published elsewhere (Bleile et al., 2024).

Sample

Schools were selected based on the following criteria: (i) predominantly serving refugee children; (ii) no other psychosocial interventions being implemented at the time of selection; and (iii) distance between schools and other settings where TeamUp might be implemented (child-friendly spaces) being ≥4 km, to avoid the risk of contamination. Schools were randomly allocated to either study by flipping a coin with multiple team members present. From the selected schools, two Grade 4 classes were randomly selected. Inclusion criteria for recruiting children were ages10–15 years, being able to participate in TeamUp and interviews, and speaking one of the most commonly spoken languages in the settlement (Bari, Juba Arabic or Aringa).

Intervention

TeamUp is a structured movement-based mental health promotion intervention (https://www.warchild.net/intervention-teamup/) that consists of activities aiming to improve children’s psychosocial well-being by strengthening social connectedness and friendships, improving emotion-regulation, promoting self-agency and a positive outlook, mapping onto essential principles of psychosocial care for people confronted with adversity and trauma (Hobfoll et al., 2007). TeamUp combines group movement games, dance and creative movement, routines, body awareness- and breathing-practices. Trained facilitators offered children 45–60 minute session twice weekly. (Note that the twice-per-week schedule entails an adaptation from the original intended schedule of once per week, an adjustment that was made due to changes in the planning and shortening of the school term as a result of COVID-19 restrictions.)

Each session follows a structure with an opening activity, a check-in, a warm-up, main activities, a cooling down and a check-out. This allows facilitators to include energy modulation through both energizing and more settling or relaxing activities to assure regulating and grounding opportunities. Activities are centered around eight psychosocial themes (i.e. fear, conflict, respect, assertiveness, anger, bullying, friendship, stress & tension). TeamUp aims to offer a safe space with trustworthy adults, engaging in social interaction with peers, strengthening resources while releasing stress in their bodies and experiencing settling and grounding, informed by trauma-informed care principles and guidelines for using creative and movement techniques when working with refugee populations (Bareka et al., 2019; Dieterich-Hartwell & Koch, 2017; Dieterich-Hartwell et al., 2020; Hobfoll et al., 2007). Control group participants received usual physical education (PE) lessons, or engaged in self-initiated non-guided play if PE classes were not offered. Referral to mental health services were offered in case of severe psychological distress for all participants irrespective of the study arm.

Procedures

Data collection was conducted by a group of 12 research assistants, who received a 5-day training. The training included sessions on basic communication and interviewing skills, the content of the instruments, practicing administration of the interviews through role-plays, ethics of research (including consent procedures), and the protocol for reporting in case of (serious) adverse events. Interviews were conducted in one of three selected languages following written consent from caregivers and assents from children. Data collection took place between April and May 2021 and October and November 2021 using the KoboCollect app for data entry. All research assistants and the lead statistician [GKG] and principal investigator [MJ] were masked to group allocation. The level to which masking was maintained was assessed by research assistants guessing group allocation of respondents at the end of each interview.

Instruments

The following instruments were administered at all time points (with presented Cronbach’s alphas based on baseline sample): (1) Stirling Children’s Wellbeing Scale (SCWBS) to assess psychosocial well-being (12 items, Cronbach’s α = .81) with subscales to assess positive outlook (6 items, α = .70) and emotional state (6 items, α = .70) (Liddle & Carter, 2015); (2) Multidimensional Students Life Satisfaction Scale (MSLSS), Friends subscale (9 items, α = .73), to assess satisfaction with friendships, and the School subscale (5 items, α = .64) to assess attitude toward school (Huebner, 1994).

Besides the aforementioned primary outcomes, we also included the following instruments to measure secondary outcomes: (3) Children’s Revised Impact of Event Scale (CRIES-8) (8 items, α = .80) to assess traumatic stress, with subscales intrusion (4 items, α = .81) and avoidance (4 items, α = .74) (Perrin et al., 2005); (4) KIDSCREEN subscales to assess quality of life (10 items, α = .70), physical health (5 items, α = .73), and bullying (3 items, α = .63) (Ravens-Sieberer et al., 2005, 2010); (5) Multidimensional Assessment of Interoceptive Awareness (MAIA) Noticing subscale to assess interoceptive awareness, capturing the ability of noticing sensations in the body (4 items, α = .70) (Mehling et al., 2012); (6) Affective Reactive Index (ARI) to assess irritability (7 items, α = .75) (Stringaris et al., 2012); (7) Short Moods and Feelings Questionnaire (SMFQ) to assess depression symptoms (13 items, α = .84) (Burleson et al., 2006; Costello & Angold, 1988).

In addition to the outcomes instruments, we were also interested in assessing the TeamUp mechanisms of action as a potential mediator of outcomes: In order to assess the degree to which TeamUp participants adopt or engage with the key hypothesized mechanisms of action that underly the intervention, we developed an instrument consisting of 24 items (α = .78) covering children’s (i) social connectedness, (ii) sense of agency, (iii) sense of safety, (iv) emotional regulation, and (iv) interoception. The instrument assessed the strategies and socioemotional skills that map onto the intervention content and theory of change, yet were formulated so that it could be scored independent of knowledge or experience of TeamUp in order to ask the same questions to control group participants. The items for social connectedness and sense of agency demonstrated good enough psychometric properties to be treated as subscales in the mediation analyses.

All instruments used a 5-point Likert scale, except ARI and SMFQ used a 3-point scale and CRIES-8 used a 4-point scale. All measures were systematically adapted for use in Uganda, following guidance for the preparations of instruments across cultural settings (van Ommeren et al., 1999). This included translation and back-translation into the three local languages (Bari, Aringa, and Juba Arabic). Cognitive interviewing was conducted with bilingual adults (n = 8) who were resident to the local area of the study and had themselves children in the age range of the study sample as well as speaking English and at least one of the three local languages used in the study. This group was selected from the pool of research assistants. The cognitive interviewing process involved an item-by-item review to ensure semantic and cultural adequacy and understandability for the study sample, asking the participants (2–3 per language) about (i) their understanding of the questions, (ii) any difficulties with understanding the question for a 10-year-old, and (iii) any suggested changes.

Analyses

The intraclass correlation (ICC) was assessed to detect clustering at the school and research assistant level. With ICCs ranging from almost zero to .05, with most around .02, we conducted linear regression analysis with follow-up scores as the response variable, group (TeamUp versus control) as the predictor, and baseline scores as covariate. All the analyses were conducted using an intention-to-treat (ITT) approach, dealing with missing values at follow-up by multiple imputation of five data sets. The sequential mediation model included group allocation as predictor, the TeamUp mechanisms of action subscales sense of agency and social connectedness at endline as mediators, well-being at follow-up as outcome, and well-being at baseline as confounder. Additionally, the same model was run with the TeamUp mechanisms of action total score at endline as the mediator. The total effect, direct effect, and indirect effects were used for interpretation, their 95% CIs were estimated using the bias-corrected percentile bootstrap method with 1,000 replications (Biesanz et al., 2010). Missing data at follow-up was handled by Full Information Maximum Likelihood. The results were depicted in a path plot including the parameter estimates, their 95% CIs, and p-values (Figures 2 and 3).

Ethics

The study protocol was approved by the Makerere University School of Health Sciences Institutional Ethics Review Board (MAKSHS-REC) and the Uganda National Council for Science and Technology (UNCST) (HS941ES). Caregiver consent and child assent were obtained prior to data collection. School directors and local authorities gave permission for the study to happen within their respective schools and area.

Results

At baseline, 549 children met inclusion criteria and were enrolled into the study, divided over the experimental arm (n = 265) and control arm (n = 284). Sample characteristics at baseline are summarized in Table 1 and described in more detail in a prior publication (Bleile et al., 2024). The follow-up assessments were conducted with 223 participants in TeamUp (84.1%) and 244 in the control condition (85.9%). See Figure 1 for the study flowchart and reasons for loss to follow-up.

Figure 1. Study flowchart.

* Reasons for loss to follow-up at endline are reported elsewhere (Bleile et al., 2024)

Table 1. Sample characteristics at baseline.

Demographic	Total sample	TeamUp	Control	χ^2(df)
Caregiver education				7.3 (3)
no schooling	149 (29%)	84 (33%)	65 (26%)
primary (completed or not)	263 (52%)	129 (51%)	134 (53%)
secondary (completed or not)	75 (15%)	29 (12%)	46 (18%)
certificate, degree or other	18 (4%)	11 (4%)	7 (3%)
Caregiver nationality				4.1 (1)*
Ugandan	31 (6%)	21 (8%)	10 (4%)
South Sudanese	474 (94%)	232 (92%)	242 (96%)
Child nationality				4.7 (2)
Ugandan	59 (11%)	36 (14%)	23 (8%)
South Sudanese	479 (87%)	223 (84%)	256 (90%)
other or missing	11 (2%)	6 (2%)	5 (2%)
Child separated from caregiver
No	428 (78%)	207 (78%)	221 (78%)	1.1 (1)
Yes	114 (21%)	54 (20%)	60 (21%)
missing	7 (1%)	4(2%)	3 (1%)

*p < .05.

The primary and secondary outcomes are presented in Table 2. We include only data from baseline and follow-up in line with the aim of this paper to evaluate sustained and mediating effects at follow-up. Baseline to endline comparisons have been presented elsewhere (Bleile et al., 2024).

Table 2. Means, standard deviations, and between-group differences of outcomes at baselines and 5-month follow-up.

Outcomes		Mean (SD)			Estimated Treatment Effect
		Total (n = 549)	AAU (n = 284)	TU (n = 265)	Mean Difference^# (95%CI) T0-T2	Cohen’s d	p
Primary outcomes
Stirling Wellbeing Scale	Baseline	30.50 (7.7)	30.89 (7.4)	30.09 (8.1)	−1.88*	0.25	.003
	Follow-up	31.08 (6.9)	30.24 (6.6)	31.97 (7.2)	(−3.14;–0.66)
Positive outlook	Baseline	14.88 (4.5)	15.25 (4.4)	14.48 (4.7)	−0.95**	0.24	.006
	Follow-up	15.23 (4.0)	14.82 (3.8)	15.67 (4.1)	(−1.63;-0.28)
Emotional state	Baseline	15.62 (4.1)	15.64 (3.8)	15.60 (4.3)	−0.90*	0.22	.014
	Follow-up	15.85 (3.7)	15.42 (3.6)	16.31 (3.9)	(−1.62;-0.19)
MSLSS Friends	Baseline	25.12 (5.9)	24.88 (6.0)	25.39 (5.9)	−1.52**	0.25	.005
	Follow-up	24.86 (5.6)	24.06 (5.6)	25.72 (5.5)	(−2.55;-0.48)
MSLSS School	Baseline	17.15 (2.6)	17.18 (2.4)	17.12 (2.8)	−0.26	0.12	.178
	Follow-up	17.11 (2.3)	16.99 (2.3)	17.24 (2.3)	(−0.65;0.12)
Secondary outcomes
CRIES-8	Baseline	14.03 (9.6)	14.38 (9.4)	13.66 (9.7)	2.18*	0.21	.014
	Follow-up	11.04 (10.3)	12.13 (10.2)	9.85 (10.3)	(0.45;3.91)
Intrusion	Baseline	5.99 (5.6)	6.44 (5.9)	5.50 (5.2)	1.38**	0.26	.003
	Follow-up	4.71 (5.3)	5.46 (5.4)	3.91 (5.1)	(0.46;2.30)
Avoidance	Baseline	8.05 (5.9)	7.94 (5.6)	8.16 (6.2)	0.76	0.11	.197
	Follow-up	6.33 (6.7)	6.69 (6.4)	5.95 (6.9)	(−0.40;1.93)
Kidscreen – health	Baseline	27.91 (5.6)	27.86 (5.3)	27.97 (5.9)	−1.29*	0.21	.014
	Follow-up	28.78 (5.6)	28.14 (5.4)	29.45 (5.7)	(−2.31;-0.30)
Kidscreen Phys. Act. Health	Baseline	13.72 (3.7)	13.99 (3.7)	13.43 (3.7)	−0.62	0.15	.082
	Follow-up	14.17 (3.7)	13.90 (3.7)	14.45 (3.7)	(−1.33;0.08)
Bullying	Baseline	3.93 (2.7)	4.08 (2.8)	3.79 (2.6)	0.53*	0.20	.020
	Follow-up	3.50 (2.5)	3.77 (2.4)	3.21 (2.4)	(0.08;0.97)
Interoception– noticing	Baseline	10.39 (3.5)	10.42 (3.5)	10.35 (3.5)	−0.40	0.11	.184
	Follow-up	11.17 (3.2)	10.99 (3.1)	11.37 (3.3)	(−0.99;0.19)
ARI–Irritability	Baseline	3.94 (3.1)	4.67 (3.1)	3.16 (2.9)	0.73*	0.18	.041
	Follow-up	4.06 (3.5)	4.59 (3.5)	3.48 (3.3)	(0.03;1.43)
Moods and feelings	Baseline	7.62 (5.4)	8.36 (5.5)	6.84 (5.2)	1.31*	0.18	.035
	Follow-up	6.99 (6.2)	7.76 (6.3)	6.16 (5.9)	(0.09;2.52)
MoA total score (24)	Baseline	60.52 (12.0)	59.86 (11.4)	61.22 (12.6)	−4.94***	0.41	<.001
	Follow-up	62.91 (11.3)	60.36 (10.0)	65.64 (11.8)	(−6.97;-2.91)
MoA–Social connectedness	Baseline	9.36 (3.8)	9.19 (3.6)	9.53 (3.9)	−1.08***	0.33	<.001
	Follow-up	9.74 (3.3)	9.20 (3.2)	10.32 (3.2)	(−1.63;-0.53)
MoA–Sense of agency	Baseline	10.92 (4.0)	10.89 (4.0)	10.95 (4.1)	−0.87**	0.25	.004
	Follow-up	11.32 (3.4)	10.89 (3.3)	11.78 (3.5)	(−1.45;-0.28)

Note: Mean difference at follow-up is corrected for baseline scores.*p < .05, **p < .01, ***p < .001.

At follow-up, participants of TeamUp demonstrated greater benefits compared to control-group participants for psychosocial well-being (estimated mean difference M_diff= −1.88, 95% CI −3.14 to −0.66, p = .003, effect size Cohen’s d = 0.25), friendships (M_diff= -1.52, 95% CI −2.55 to −0.48, p = .005, d = 0.25), traumatic stress (M_diff= 2.18, 95% CI 0.45 to 3.91, p = .014, d = 0.21), quality of life (M_diff=-1.29, 95% CI −2.31 to −0.30, p = .014, d = 0.21), bullying (M_diff= 0.53, 95% CI 0.08 to 0.97, p = .020, d = 0.20), depression symptoms (M_diff= 1.31, 95% CI 0.09 to 2.52, p = .035, d = 0.18). There were no significant between-group differences for attitude toward school, physical health, and interoception. Furthermore, we saw a significant difference in the opposite direction for irritability (M_diff= 0.73, 95% CI 0.03 to 1.43, p = .041, d = 0.18). Finally, for the hypothesized mediators, TeamUp participants showed significantly greater improvements in social connectedness (M_diff= -1.08, 95% CI −1.63 to −0.53, p < .001, d = 0.33), sense of agency (M_diff= -0.87, 95% CI −1.45 to −0.28, p = .004, d = 0.25), and in the total score of the mechanisms of action assessment (M_diff= -4.94, 95% CI −6.97 to −2.91, p < .001, d = 0.41).

Mediation analyses show that the total effect of TeamUp on improving well-being at follow-up was significant (c = 1.96, SE = 0.61, p = .001), and from this total effect, a significant proportion, 18%, was mediated by social connectedness and sense of agency at endline (total indirect effects = 0.35, SE = 0.14, p = .016). Specifically, the mediation analysis indicated a significant indirect path between receiving TeamUp and greater improvements in well-being via increased social connectedness (indirect effect = 0.30, SE = 0.13, p = .001), explaining 15% of variance. The indirect path via increased sense of agency is not significant (indirect effect = 0.05, SE = 0.09, p = .606). There is a significant direct effect from TeamUp to well-being at follow-up (c’ = 1.61, SE = 0.62, p = .009), which means that the above-mentioned model concerns partial mediation. See Figure 2 for details.

Figure 2. Mediation results for social connectedness and agency.

Of the total effect of TeamUp on well-being at follow-up, a significant proportion, 26%, is mediated by the Mechanisms of Action total score at endline (indirect effect = 0.51, SE = 0.17, p = .002). The direct effect after accounting for this mediation is also significant (c’ = 1.43, SE = 0.62, p = .021) (Figure 3).

Figure 3. Mediation results of total mechanisms of action score.

The TeamUp implementation indicators were all positive, with 91.3% attendance across sessions and 97.1% fidelity across individual facilitators (Bleile et al., 2024). Furthermore, assessment of the maintaining masking of research assistants at follow-up showed 57% guessing correctly which children were in the control group and 69% guessing the TU participants correctly (reference is chance level—i.e., 50%). A total 18% of the research assistants found it difficult to guess and responded “I don’t know.”

Discussion

This follow-up study of a quasi-experimental study evaluates the outcomes of a mental health promotion intervention, TeamUp, for refugee children in Uganda. A prior publication demonstrated positive results immediately following the intervention (Bleile et al., 2024). The current study has two distinct contributions: (1) evaluating sustained effects at 5 months postintervention and (2) explaining how a mental health promotion intervention results in improving well-being, based on a sequential mediation model exploring a priori hypothesized mechanisms of action.

First, at 5 months postintervention we still see clear benefits of participating in TeamUp compared to the care-as-usual control group. Children had improved psychosocial well-being and friendships (two of three primary outcomes), and a range of secondary outcomes (i.e., reduction in traumatic stress and depression symptoms, improved quality of life, and reduction in bullying), with effect sizes ranging from 0.18 to 0.25. The small effect sizes are in line with a recent meta-analysis of universal and selective interventions that aim to promote good mental health, which showed average effect sizes to be between 0.18 and 0.49 for most outcomes (de Pablo et al., 2020). When comparing these to the outcomes reported at endline (Bleile et al., 2024), we see that some of the benefits are maintained (improved well-being, reduced traumatic stress, improved quality of life, increased social connectedness, and sense of agency), with similar magnitudes of change. In addition, some benefits occur at the follow-up time point, that were not seen at the endline time point—namely, improved friendships and reduced bullying and depression symptoms. It appears that participating in TeamUp—which emphasizes values such as respect, collaboration, and connectedness—strengthens friendships that mature over time and might have translated to reduced bullying. The compressed period of implementation (as a result of COVID-19) of 6 weeks may have been too short to result in immediate changes in complex social dynamics involved in friendships and bullying, but it may well have planted the seeds for these changes to happen following the intervention. The benefits in terms of reduction in depression symptoms at follow-up, in absence of such reduction at endline, can be explained by a similar mechanism—that is, the immediate postintervention increase in positive outcomes contributes to downstream reduction in depression symptoms after some time. A similar delayed effect is seen in the evaluation of a universal school-based depression prevention program in China (Lai et al., 2016). This finding also points to the complementarity and overlap in the concepts of mental health promotion and universal prevention (Singh et al., 2022). Further research into TeamUp could involve investigating effectiveness in the prevention of depression or PTSD in refugee populations.

Finally, there are some indicators that no longer show significant improvements that did at the endline time point (i.e., attitude toward school and perceived physical health) or show even a slight worsening (i.e., irritability). It is plausible that the loss of these intervention gains are directly the result of stopping the intervention. Also, between endline and follow-up measures, schooling was not offered due to COVID-19 restrictions, which undoubtedly impacted children’s perspectives on schooling at large, and increase in irritability specifically, as TeamUp was embedded with the school structure. The loss of gain in reported physical health also maps neatly onto stopping the intervention, given the primary focus of the intervention on the body, movement, activation, and regulation.

Efforts to promote the mental health of children in LMIC can build on accruing evidence in general (Barry et al., 2013) as well as specifically for life-skills interventions (Singla et al., 2020) and physical education (Purgato et al., 2023). Yet, there is a scarcity in evidence of longer-terms effects (Purgato, Gross, et al., 2018). A review by O’Reilly and colleagues of mental health promotion interventions in schools concluded that results on maintenance of benefits are scarce and unclear (O’Reilly et al., 2018)—a conclusion that is shared specifically for mental health promotion evaluations in LMIC (Barry et al., 2013). The current study of a movement-based mental health promotion intervention adds to the potential for such interventions to have effects that maintain and expand over time.

Second, this study has demonstrated that postintervention (endline) improved social connectedness and the combination of skills and strategies that make up the hypothesized mechanisms of action of TeamUp partially mediate (18% and 26% variance explained, respectively) the effect of the intervention on improving psychosocial well-being at 5 months (follow-up). Improved sense of agency at endlines does not mediate this effect significantly. Importantly, the improvements on both social connectedness and the total mechanisms of action score are sustained at follow-up, in fact with the largest effect sizes (d = 0.33 and d = 0.40, respectively). This finding, combined with the mediation results, make it plausible for the demonstrated effects to be sustained even beyond the 5-month follow-up. This should be investigated in future studies.

These results confirm the well-established association between social ties and mental well-being (Kawachi & Berkman, 2001), as well as the potency of social connectedness as an intervention ingredient that has been well established in the literature to contribute to positive outcomes (WellcomeTrust, 2021). Moreover, these results substantiate the emphasis that is given in TeamUp to (i) creating a safe space for children and (ii) teamwork and collaboration-focused activities and exercises that are central to the intervention. This also points to the added value of knowing how interventions result in the desired outcomes. Demonstrated outcomes can be maintained or optimized in future implementation of TeamUp, with adequate and potentially increased focus on the elements that contribute to social connectedness.

Another implication of the mediation results is the relevance of assessing intervention-specific mechanisms of action (Miller et al., 2023). While many interventions have an underlying theory of change and aim to instill specific skills or (therapeutic) strategies among participants, the degree to which those are obtained are rarely assessed. To do so, we recommend developing instruments that allow the assessment of adoption of intervention-specific strategies and skills. An example is the Reducing Tension Checklist, which is an instrument to assess the level of adoption of strategies that map onto the key ingredients of the WHO’s Group Management Plus (gPM+) intervention, demonstrating to explain 31% of the obtained benefits (Jordans et al., 2021). Once researchers have demonstrated that the intervention is associated with desired outcomes, as we have done with the mechanisms of action scale in the current TeamUp study, then such an instrument can be used for routine monitoring during the implementation of the intervention. In turn, such information provides useful data to see whether participants are on track in adopting the desired strategies, skills, and behaviors and, if not, implementation through mentoring or supervision mechanisms can be adjusted.

Some limitations are important to note. To demonstrate (long-term) effectiveness of TeamUp, a cluster randomized-controlled-trial design is needed. While our quasi-experimental study did use a large sample of individuals, had a good level of retention and fidelity, and showed the effect of belonging to a cluster to be minimal, future study will need to be powered to allow randomization of an adequate number of clusters. Another limitation is that the time frame for assessing sustained effects of TeamUp was “only” 5 months. This period was not a priori set, but dictated by COVID-19 restrictions or the loosening thereof. A future study should assess even longer-term effects, for example, at 12-months postintervention. A third limitation is that there appeared to be some level of unmasking of the research assistants, with higher-than-chance levels of accurate guessing of group allocation, which may have introduced some bias. Finally, we believe that it is a strength that we used a sequential approach to investigating mediation—that is, by using endline outcomes mediating results at follow-up. Yet the study was not powered to assess the mediation pathways reported in this study.

Conclusion

This study demonstrates that a largely nonverbal movement-based mental health promotion intervention outperforms a control group even 5 months after completing the intervention. Some intervention benefits were maintained from previously reported endline results (i.e., improved psychosocial well-being and quality of life, reduced traumatic stress, and improved social connectedness and sense of agency), while new gains (i.e. improved friendships, reduced bullying and depression symptoms) and some benefits drop off over time directly related to stopping the intervention (i.e., improved attitude toward school and physical health). Increased social connectedness and the adoption of strategies, skills, and behaviors that TeamUp aims to instill, at endline, explain the medium-term intervention benefits in terms of improved well-being at follow-up. With few evidence-supported mental health promotion interventions in LMIC, this study shows that TeamUp holds potential to have immediate and longer-term effects. This is salient given that TeamUp requires only relatively brief training, can accommodate large groups of participants, and works mainly through nonverbal means—making it useful for large-scale implementation in multicultural/multilingual, mixed refugee, and host populations. Additionally, with a better understanding of how the intervention contributes to these sustained outcomes future implementation and cRCT can aim to optimize benefits.

Author Contributors

MJ and AB designed the study. BO and AB were responsible for the overall coordination and oversaw the data collection. MJ, GKG, and AB developed the data analysis plan; GKG and AB analyzed the data; MJ drafted the manuscript; and AB, BO, AB, GKG, and KV critically revised the manuscript. All authors commented on and approved the draft and final manuscript.

Acknowledgments

We thank all children participating in our study, school teachers, head-teachers, and community leaders (mobilisers) and our dedicated research assistants. We highly appreciate insights we received from Dr. Rosco Kasujja and Samuel Ouma (Makerere University), Emmanuel Ngabirano (TPO Uganda), Kevin Aciro (Save the Children Uganda) and Mathew Atibuni (Lutheran World Federation). We also thank the TeamUp facilitators and supervisors, the TeamUp Global team, and War Child staff in Uganda for all their support.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Data Availability Statement

Data sets are available upon request via the corresponding author.

Additional information

Funding

This work was supported by the Templeton World Charity Foundation under grant number [TWCF0409].