Education in crisis

Global education and learning are in crisis. In low- and middle-income countries (LMICs), 70% of children are experiencing ‘learning poverty’, defined as being unable to read a simple text by the age of 10 or the end of primary school [1]. Of the 234 million children and adolescents worldwide living in fragile, conflict-affected, and violent settings, only 17% are achieving minimum academic proficiency levels, more than a third (37%) are out of school completely (52% girls), and the remaining 110 million are in school but not learning sufficiently to reach minimum proficiency targets [2]. Beyond academic learning, being in school has intrinsic benefits for children’s social and emotional wellbeing, as well as access to health and other interventions utilising schools as a delivery platform – a commonality in conflict settings [3–6].

However, only 58% of children in low-income countries (LICs) complete primary school, 35% complete lower secondary, and just 22% complete upper secondary education [7]. Compounding the issue of dropout is low and irregular school attendance, which is considered an indicator or “early warning signal” of wider system instability at the student, family, school and community levels [8]. An estimated 11% of in-school adolescents globally are chronically absent [9], defined as missing 10% or more of a school year. School attendance and retention are key to both cost-efficiency and addressing learning poverty; investment in school reform and school-based interventions is only worthwhile if children are consistently in school. Meeting minimum thresholds for attendance and retention is critical to ensuring intervention quality, impact, cost-effectiveness, and scalability [10].

Risk and protective factors for attendance and dropout in LMICs and conflict-affected settings

In LMICs and conflict-affected settings, the factors influencing school attendance and absenteeism are numerous [11]. Here we highlight the most prevalent at different levels of the socioecological system around children: the individual, the household, and the school.

At the individual level, school attendance and dropout are influenced by the child’s gender, age, disability status, mental health, and peer relationships. Boys are at higher risk of dropout in many contexts (e.g., over double the risk of girls in Pakistan, Afghanistan and Bangladesh [12]), mainly due to the desire or pressure to transfer into the workforce; however, these rates are calculated among enrolled students who are more likely to be boys to begin with [12]. Girls face numerous barriers to staying in school, including household obligations, discriminatory cultural beliefs, menstruation-related factors (e.g., insufficient hygiene materials and facilities), and early marriage and pregnancy [11,13]. Additionally, national policies in 2% of countries limit married, pregnant and parenting girls and women’s right to education, and one country (Afghanistan) bans girls’ access to secondary and higher education completely [14]. Older age and larger discrepancy between the student’s grade and age increase the risk of dropout, notably common in refugee contexts due to years out of school or delays in registration into host country’s education system [11]. Children with disabilities face infrastructural barriers and negative societal beliefs about their academic potential [11]. Experiencing stress and anxiety increases the likelihood of absenteeism by up to 60% [15], while being bullied results in 1.29 to 1.53 higher odds of being absent from school in low-middle income countries [9]. Conversely, peer support and friendships are protective against absenteeism, associated with 17% to 37% lower risk of missing school [9].

At the household level, poverty is one of the strongest drivers of school absenteeism and dropout [9,11,16]. Globally, students living in food insecure households are 43% more likely to be absent from school [17], increasing to 47% in low-middle income countries in Africa [9], and rising to 120% in Pakistan, Bangladesh, and Afghanistan, i.e., more than double the likelihood compared to children living with food security [12]. Family composition also matters; more siblings, later birth order, being an orphan or non-biological child, and living in a single parent household increase the risk of school dropout [11]. On the other hand, higher levels of parental education and employment are protective against school dropout [11]. Caregiver engagement, defined as the support of the parent or adult responsible for the child’s development towards their academic progression, is found to positively influence attendance and retention in school [18–20]. Higher caregiver support was associated with significantly lower odds of absenteeism (relative risk ratio = 0.6, p < 0.01) in Pakistan, Bangladesh and Afghanistan, and in African low-middle income countries by between 21 and 33% lower odds [9,12]. In Kenya, higher caregiver engagement was associated with student retention (r = 0.811, p < 0.01; [21]). In displacement settings, caregivers’ beliefs about the value of education drive their decision-making on their child’s schooling [19].

At the school level, physical violence on the way to or at school increases risk of absenteeism by 33–64% in African and Asian low-middle income countries [9]. Large class sizes, lacking infrastructure (e.g., desks, roofs), teacher absenteeism, and corporal punishment further undermine participation, as does sedentary learning with little interaction (e.g., OR 2.48, p < 0.001) [9,11,22]. Long distances to school risk dropout; Kremer et al. [23] found that girls’ enrolment fell by 19% for each additional mile to school, while boys’ fell by 14%. In refugee contexts, language barriers and unfamiliarity with the host country’s education system significantly deter school attendance [24–26].

Deeper understanding of the issues of chronic absenteeism and low attendance faces important limitations. First, attendance data in LMICs is sorely lacking and inconsistently measured; in an analysis of 27 national Education Management Information Systems, only eight recorded attendance [4]. Second, most literature focuses on adolescents, leaving the risk and protective factors for primary school students often assumed to be the same. Third, studies often focus on other educational outcomes, such as enrolment, grade progression, and learning, which are frequently used as proxies for attendance and retention. Fourth, there is a dearth of quantitative evidence from conflict-affected settings.

Interventions to increase school attendance in low-resource and conflict-affected settings

In a systematic review of interventions in LMICs aiming to increase school enrolment and participation, the mean effect across intervention types on attendance was d = 0.15 (95% CI 0.10–0.20) and on dropout was d = 0.05 (95% CI 0.02–0.-09) [27]. These are considered medium and small effect sizes, respectively, within the context of education interventions [28]. Another review, exclusively of randomised controlled trials, found a median effect of 0.07 [29]. Interventions to increase school attendance and retention in LMICs and conflict-affected settings fall broadly into four categories.

First is health and hygiene programmes. In low-income settings, one review found that school feeding had a small, positive effect on attendance--a projected increase of 4–6 days per year [30]--while another found mixed evidence from studies in the Philippines, Burkina Faso, Bangladesh, Ethiopia, and Peru [31]. Emergency school feeding in refugee contexts in Lebanon resulted in up to 14% (p < 0.001) lower school dropout and 5 additional days of attendance [32], while in Mali it increased enrolment by 10% (p < 0.001) but had no effect on attendance [33]. A systematic review of menstrual hygiene interventions in LMICs found a non-significant impact on attendance [34]; in some cases they have a negative impact [31]. Qualitative evidence from displacement settings points to the benefits of Water, Sanitation, and Hygiene (WASH) interventions on educational outcomes, a conclusion supported by reviews in LMIC, though inconsistent outcome definitions prevent clear estimates of their impact [19,31,35].

Second is economic and in-kind interventions, such as school fee waivers, cash transfers or provision of school uniforms, which constitute the majority of evaluations [31]. Fee waivers alone, without accounting for other costs (e.g., books, uniforms, lost opportunity), are rarely effective [31]. A meta-analysis of conditional cash transfer programmes in LMICs demonstrates significant positive effects on school attendance (3% increase for primary and 6% for secondary) and dropout (1% lower in primary and 3% lower in secondary) [36]. Unconditional cash transfers have been shown to have no effect on school attendance in some settings (e.g., Gambia and Mexico; [37,38]), and positive but small effects in others (e.g., Syria; [39]). Though many promote the use of cash transfer programming to improve educational outcomes (e.g., [23]), the Global Education Advisory Panel [40] consider them a ‘bad buy’ in their review of cost-effective approaches to improve global learning. Disbursement of school uniforms has been shown to temporarily reduce absenteeism (e.g., in Kenya by 37%), however this effect disappeared with time [41]. In Ecuador, distribution of uniforms was found to significantly, negatively affect attendance [42].

Third are caregiver- and community- focussed interventions, such as information campaigns on the importance of education or promoting caregiver engagement. While Petrosino et al. [27] found a small, non-significant effect (d = 0.06, 95% CI −0.09–0.05) of informational campaigns, the Global Education Advisory Panel [40] classify them as a ‘great buy’, likely driven by their low cost. There is some evidence for SMS reminders as effective nudges to increase attendance; for example, SMS messages in Chile resulted in 5% more children achieving attendance requirements for grade progression [43]. Spier et al. [20] found mixed evidence in a meta-analysis of the effects of caregiver and community engagement on children’s literacy outcomes but did not examine effects on attendance and dropout. In India, a three-armed trial compared the effects of sharing information on local services with school and community members, training sessions on caregiver assessment of a child’s learning, and teacher training; absenteeism increased in all groups [44]. In Mexico, providing information to caregivers increased their engagement and improved student behaviour, and in Malawi, a family-school-community intervention improved caregiver engagement for children with disabilities, but neither study recorded effects on attendance [37,45].

Fourth is remedial and supplementary education, such as accelerated education programmes, community-based education, and interventions aimed at improving teaching quality and learning. Considering their wide use in displacement contexts, evidence for effects on attendance and dropout is sparse. In a review of what works in displacement settings, Burde et al. [19] show that only one of five remedial education programmes had a positive effect on attendance; none measured dropout. Similarly, of the 20 evaluations of socio-emotional learning, education technology, teacher training, and school management programmes, only one included attendance as an outcome (and did not show an effect) [19,46]. However, evidence from emergency settings points to positive effects of technology on attendance and retention through motivating children to stay in school (e.g., [47,48]).

We observe several limitations and gaps in interventions promoted to improve school attendance and retention in LMICs and conflict settings. First, the majority do not primarily target attendance and dropout; these are often assumed to be collateral benefits. Second, quantitative evidence for such collateral benefit is limited and inconsistent. Third, the interventions often have large administrative burden and implementation costs, e.g., school feeding and cash transfer, which likely prohibit large-scale implementation. Fourth, there is a scarcity of evidence from conflict-affected settings. These support the need for low-cost interventions developed specifically to target attendance and retention in low-resource and conflict-affected settings.

Rationale and aim of currently study

As of July 2026, Uganda hosts over two million refugees and asylum-seekers, making it the largest refugee-hosting country in Africa [49]. Most refugees reside in 30 settlements in the North, West and Southwest of the country, bordering South Sudan, DR Congo, and Rwanda. Country-wide, 4% of children are out-of-school and the learning poverty rate is 82% [50]. Within the settlements, at least 49% of school-aged children are estimated to be out-of-school [51].

War Child (WCA) is an international non-governmental organisation which aims to improve the wellbeing and strengthen the resilience of conflict-affected children and their communities. WCA follows a systemic approach, consisting of a combination of interventions which are multisectoral (mental health and psychosocial support, protection, education), multi-level (population-wide and specialised care), and socioecological (targeting communities, schools, education staff, families, caregivers, and children, adolescents, and youth) [52]. The present study describes the collaborative design process of ‘SchoolLinks’, which is part of that systemic approach, aiming to increase students’ school attendance and retention in refugee settlements and conflict-affected communities. This research seeks to answer, “What complementary strategies, co-designed with caregivers and teachers, can positively influence children’s school attendance and retention in refugee settlements?” and “To what extent is SchoolLinks culturally and contextually acceptable, appropriate and feasible to implement?”

Phase 1: Group model building

Participants & recruitment.

In each of the four schools, teachers for Primary Three (P3) and/or Accelerated Education Program 1 (AEP 1; a refugee education response to accommodate over-age learners) and 10 caregivers were invited to participate in the research (see Table 1). Fewer teachers than caregivers were selected to lower the influence of teachers on the model that was being built and increase the variation in caregiver responses. The headteacher of each school was involved in the GMB preparations but not the workshops; this was done to mitigate power imbalances.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Participants in the three phases of intervention development.

https://doi.org/10.1371/journal.pone.0357215.t001

Recruitment started on 1^st November 2021 and completed on 18^th November 2021. Caregiver selection was based on previously collected data of 50 randomly selected students from P1 to P7 and AEP 1 on children’s home learning environment, which measured caregiver and community support towards children’s reading and numeracy skills inside and outside the home. Students were ranked from the lowest to highest home learning environment score, and the lowest, the highest, and eight interval positions in between, were selected to achieve a sample representing a spectrum of home learning environments. The caregivers of these children were invited to participate in the GMB workshops, totalling 10 per workshop. Their characteristics were checked against those of the rest of the P3 class to ensure approximate representativeness by gender, age, nationality, and relationship to child. At the start of each workshop, the research and the details of the consent form were presented and any questions answered, after which written consent was obtained from participants. In cases where a participant was illiterate, verbal consent was obtained in the presence of a witness known to the participant.

Phase 2: respondent validation & consultation

Phase 3: field-testing

Ethics

Ethical review and approval for Phases 1 and 2 was received from Makerere University’s School of Social Sciences’ Research Ethics Committee (03.20.394) and Uganda’s National Council of Science and Technology (SS451ES). Prior to each activity, information sessions were held in different languages to explain the aims, methods, risks, and requirements of the research, data management and confidentiality measures, and gather written or witnessed consent (the latter for non-literate participants only). Contact details of WCA staff and representatives of the ethical review boards were shared so that participants could directly communicate concerns and questions.

Ethical review and approval for Phase 3 was conducted by The AIDS Support Organisation (TASO-2022–136) and Uganda’s National Council of Science and Technology (SS1378S). Additionally, the research and SchoolLinks were presented to the refugee settlement and school management bodies to gather their approval before commencing the field test, including Office of the Prime Minister, community leaders, the School Management Committee, and Parent-Teacher Association. In the first meetings with school staff and caregivers, attendees were given the option to opt their children (and themselves) out of the intervention and the research. Contact details of WCA staff, field office, and representatives of the ethical review boards were also shared. Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in the Supporting Information (S1 File).

GMBs: Casual map and action ideas

Participants were highly engaged in the GMB workshops and the rich discussions contributed to detailed causal map for school attendance, retention, and caregiver engagement in refugee settlements in Uganda (see Fig 1). Time constraints meant that not all activities were systematically completed for each workshop; we prioritised gathering variables, developing the connection circles, and compiling action ideas over participant-led categorisation of variables and rating the actions ideas, which we did in the respondent validation workshops instead. Many examples and stories were shared (see Table 2 for example stories and the extracted variables, S1 Table for all variables generated, and S2 Table for the causal links) to provide nuance to the descriptions that participants provided. On average, each workshop produced 70 variables with 142 causal links. Where appropriate, variables were recoded; for example, the availability of food, soap, kerosene, and menstrual hygiene materials were recoded as ‘basic needs are met’, leaving 78 unique variables. Of these, only 13 were mentioned in a single location. To determine the variables included in the map, we identified those that were mentioned in three or more locations, resulting in a causal map consisting of 38 variables (see Fig 1).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Causal map.

Caregivers’ and teachers’ depiction of factors influencing school attendance, dropout, and caregiver engagement. Black arrows depict positive polarity, red arrows depict negative polarity.

https://doi.org/10.1371/journal.pone.0357215.g001

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Example stories and variables from GMB workshops.

https://doi.org/10.1371/journal.pone.0357215.t002

There were differences between the four locations’ categorisation of variables and causal loop diagrams (see S1 Fig for all causal loop diagrams), but several common themes emerged. First, the household financial status was a key determinant of caregivers’ support to their children’s education. Caregivers are constantly faced with decisions on how to allocate limited financial resources, balancing between school-related expenses (e.g., school fees, scholastic materials) and other basic needs. This strain is exacerbated in single-headed households, and linked to drug and alcohol usage, domestic violence, child labour and school dropout.

Second, caregivers’ attitudes and knowledge of their children’s education directly influence their prioritisation of spending money on school fees. Positive attitudes and increased engagement can lead to higher child motivation and attendance, whereas negative attitudes risk the child being kept at home to support chores or not encouraged to attend school.

Third, caregivers’ attitudes and knowledge are influenced by their communication and relationships with teachers, which in turn affect the teacher’s relationship with the student. However, language barriers between school staff and refugees (and between refugees) significantly challenge this communication. Teachers face this challenge in the classroom with their students, as well as large class sizes and limited resources (e.g., desks, teaching materials). The relationship between teachers and students affects children’s attendance and behaviour in class, influenced by teachers’ use of corporal punishment.

Fourth, low school attendance is a predictor of dropout, which in turn increases risk of early pregnancy, child labour, and future unemployment. Children’s attendance, motivation and academic progression have a cyclical reinforcing relationship, where the causal relationship between motivation and achievement is mediated by attendance. Children’s motivation and academic achievement influence caregivers’ attitudes and behaviour.

Many action ideas were shared (see S1 Text for full list). Those that aimed to increase communication between teachers and caregivers, teachers’ knowledge of children’s home context and individual needs, and caregivers’ knowledge of their child’s education were prioritised, as were those that addressed unsafety on the way to school. Some action ideas were combined to form strategies, for instance regular attendance monitoring by the teacher and gifting children with basic or scholastic necessities were combined into a conditional reward strategy. Eight strategies were identified for the next phase of development, including: 1) organising meetings between school staff and caregivers, 2) translation at these meetings, 3) digitised enrolment, attended by caregivers, 4) attendance monitoring and conditional rewards, 5) homework groups supervised by caregivers, 6) caregivers accompanying children to school, 7) caregiver volunteers in the classroom, and 8) support for a community income-generating activity. Other suggested action ideas, such as parental education on discussing sexual and reproductive health, formed a suggested agenda of topics for discussion in the meetings between caregivers and school staff.

Respondent validation & consultation

In the validation workshops, participants confirmed the validity of the variables included in the causal map as those affecting school attendance, as well as the relationships depicted between the variables (see S2 Text for the compiled workshop notes). Some suggested additional variables for inclusion, such as distance to school, cultural beliefs related to underage marriage, and disability. Where these were repeated across workshops, or reflected input from the GMB workshops, the additional variables were incorporated into the causal map.

The meetings strategy was considered the most feasible and impactful, rated ‘high’ on both criteria across participant groups (caregiver, teachers, partners) and locations. Translation was rated high impact by all, but some respondents questioned the feasibility of relying on volunteers. The strategy of caregivers enrolling their children was considered high impact by all except one group in one location, who believed it would have low impact on attendance. Most respondents rated it highly feasible, however teachers flagged that it would not be possible to do before the start of term, as teachers return home during school holidays. Attendance monitoring and rewards were universally rated as high impact and participants suggested soap, exercise books and pens, and menstrual hygiene products as rewards that would motivate attendance and have broader benefits. However, there was mixed feedback on its feasibility; many participants questioned how rewards would be decided, cautioned against the sustainability of the strategy, and proposed alternative systems for awarding rewards (e.g., needs-based or achievement-based).

Homework groups supervised by caregivers were rated by most participants as high impact, believing they would enable peer-to-peer learnings and foster friendships between children, but moderate or low feasibility due to reliance on volunteers, differing school policies on homework, and questions about appropriate locations. Similarly, accompanying children to school was rated as high impact but low feasibility by most participants, attributed to timing issues as many caregivers tend to their vegetable gardens in the early morning. Children walking to school together in groups or accompanied by older siblings was proposed as a more viable alternative. Volunteer support in classrooms by caregivers was unanimously voted high impact in the belief that it would support teachers’ workload and positively influence children’s behaviour and was rated feasible by teachers but unfeasible by most partners representatives and caregivers. Feedback stressed that caregivers would need to be paid and that they would face language barriers with students. The final strategy – an income-generating activity – was rated high impact but low feasibility by most participants. Concerns were raised about its financial sustainability, how many caregivers would be able to benefit through it, and the training it would require.

In the WCA workshop, two strategies were considered unfeasible from an organisational perspective: volunteer classroom assistants would need to undergo a recruitment process to ensure adherence to child safeguarding principles, which would substantially increase the intensity of the strategy. Similarly, support for an income-generating activity was considered high intensity and outside of WCA’s technical expertise. Concerning rewards for attendance, the group cautioned that providing menstrual hygiene products without accompanying education, community sensitisation, and facilities, created a risk of harm. The remaining six strategies were implemented for field-testing and reflected in a Theory of Change which was developed and refined over the course of the GMBs and respondent validation process (see Fig 2).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Theory of change of schoollinks.

https://doi.org/10.1371/journal.pone.0357215.g002

Field-test

Intervention.

Following a series of preparatory meetings with the school headteacher, SMC, PTA, community leadership, and translators, the first Meeting (Picture 1 in Fig 3) was held at the end of Term 2 to introduce SchoolLinks and discuss the importance of education and caregivers’ engagement. Translation (Picture 2 in Fig 3) was done by two volunteers in Juba Arabic and Bari, though additional translation needs were checked at the start of each meeting. Posters of the six strategies supported the discussion (see Fig 3).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Visual of the SchoolLinks strategies.

(1) Meetings, (2) Translation, (3) Enrolment week, (4) Star charts and rewards, (5) Homework Clubs, (6) Safety to School.

https://doi.org/10.1371/journal.pone.0357215.g003

Six teachers were given a half-day training on using a tablet for enrolment data collection a week before Enrolment week (Picture 3 in Fig 3), which was held in the first week of the school term, during which enrolment is normally done. Caregivers who arrived alone or had asked neighbours/friends to register their child were asked to return with their child. Teachers entered children’s registration data on tablets, including their name, date of birth, home address, caregivers’ contact details, the Washington Group Short Set on Functioning [65], and two questions on the home context and specific needs of the child. Most students (61%) were enrolled in the first 3 days, and then a tablet was left at the school for additional enrolments.

Students’ names were extracted from the enrolment data to place on the charts for monitoring attendance – called ‘star charts’ – which were made from Velcro and laminated sheets so that they were reusable. Teachers collected attendance every morning and afternoon, and on Fridays monitored weekly attendance; students who attended four days in a week were given one star to place by their name on the star chart, those who attended five days were given two stars (Picture 4 in Fig 3). After 4 weeks, students with seven or eight stars – equivalent to 95–100% attendance – were given a bar of soap. Two cycles of attendance monitoring were completed per term, in the second cycle children were rewarded with a pen and exercise books.

A second Meeting was held near the start of the third term to remind caregivers about attendance and rewards, homework clubs, and safety to school. Teachers used the enrolment data and their knowledge of their students to group children into Homework Clubs (Picture 5 in Fig 3), which met during the weekends and were supervised by volunteer caregivers (who volunteered themselves in the second Meeting). These same groups of children agreed on a time and place to meet each day and walk to school together every day to increase their Safety to School (Picture 6 in Fig 3). Table 3 depicts the implementation details of each strategy as well as the specific challenges the strategies are designed to address and their intended outcomes.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. SchoolLinks strategy description.

https://doi.org/10.1371/journal.pone.0357215.t003

Feasibility and practicality.

The field test helped identify challenges and inform direct adaptations to the intervention design (see Table 4). A key learning was the importance of adapting to caregivers’ schedules, for example, avoiding activities during gardening times or market days. Adjusting the Meeting time from morning to afternoon increased caregivers’ attendance from 70% to 90%. To account for Ugandan teachers travelling home during holidays, Enrolment Week was moved to the first week of term. Bad weather and low digital literacy slowed enrolment, but it sped up as familiarity with the process increased. Similarly, Attendance Monitoring was initially time-consuming as teachers did not know all children by name or sight, and children took time to find their names on the star charts. It was more difficult for children who lived further away to participate in Homework Clubs and Safety to School, though most persevered. Another learning was the need for implementation flexibility, for example, allowing caregivers to coordinate Homework Club meeting times and locations between themselves, and for children to do the same for Safety to School.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Field-test challenges and adaptations.

https://doi.org/10.1371/journal.pone.0357215.t004

Teachers took on multiple roles which could risk overburdening them over time or at a larger scale. Support from the WCA implementers, especially at the start, was necessary and valuable but it will be important to determine the precise role and responsibilities of such external implementers in the future. Small indications of task-sharing, such as school management taking over facilitation of the Meetings and supporting caregivers in the Homework Clubs, suggest that the implementer role could be withdrawn over time to increase programme sustainability.

Comparison to risk and protective factors

This research demonstrates the value of using systems modelling to inform intervention development. The group model building and resultant causal map reflect many risk and protective factors identified in the literature. At the child level, the positive influence of friendships and peer support was highlighted and incorporated into the intervention design, specifically Homework Clubs and Safety to School. Children’s mental health was not discussed in detail but their motivation to attend school emerged in the causal map as a key influencing factor on attendance. This reflects findings on the importance of children’s intrinsic motivation, for example in Mozambique [67] and South Africa [68]. Motivation is implicit in much literature, e.g., on safety or bullying, but the association between children’s motivation and school attendance in refugee settings would benefit from deeper investigation.

At the household level, poverty was discussed as a strong risk factor for absenteeism and drop-out, aligning closely with findings from research in low-income and conflict settings [9,12,17]. The choice of soap, exercise books, and pens as Rewards was guided by these discussions, and their conditional nature aligns more consistent evidence for conditional over unconditional in-kind incentives to influence attendance [31]. Caregivers’ knowledge of children’s schooling, attitudes towards education, and degree of encouragement were identified as important protective factors if strengthened. These were influenced by perceptions of quality and safety at the school, the child’s learning outcomes, and their relationship with teachers and school staff. This supports evidence from low-resource contexts on the protective role of caregiver engagement [9,12,20], and raises the importance of further investigation of this within refugee contexts.

At the school level, the language barrier and safety to school were consistently mentioned in the GMBs as challenges to attendance, aligning with findings of research investigating caregiver engagement, school attendance, and children’s schooling experience in refugee contexts [24,26,69]. The relationship between teachers and caregivers emerged as a central factor in the causal map, as it was closely linked to the relationship between teachers and students and their motivation to come to school, and influenced by communication between teachers and caregivers (or lack thereof). The relationship and communication between caregivers and teachers were considered key leverage points as, when positive, they boost caregivers’ knowledge of and encouragement towards the child, as well as teachers’ motivation, knowledge of and empathy for their students. The importance of this relationship is recognised through many global initiatives promoting home-school partnerships, such as Epstein’s School-Family-Community Partnership Model [70], though investigation of its influence on attendance in refugee contexts remains underexplored.

Design rationale

The SchoolLinks design draws on elements of other attendance interventions. First, the use of conditional incentives was noted to be effective by teachers. Second, increasing caregivers’ knowledge of their child’s schooling, with an emphasis on information exchange between the home and school rather than ‘sensitisation’ of caregivers, was appreciated by caregivers. Third, is the use of multiple strategies rather than a single intervention to address a complex issue, such as attendance [3,31,71]. Fourth, targeted intervention on safety and language barriers were demanded by participants, as observed in other refugee contexts [25,26,72,73]. Fifth, though the strategies are too overt to be considered ‘nudges’, the design draws on the premise of using multiple prompts or incentives to influence greater behaviour change [74].

Much research and programming on caregiver engagement and home-school partnerships is centred on Epstein’s [75] model, which describes six types of parental involvement in education: (1) parenting (e.g., waking and feeding the child before school), (2) home-school communication, (3) parental volunteering, (4) home learning support, (5) parent involvement in decision-making, and (6) community collaboration. We observe several similarities between Epstein’s model and the SchoolLinks design. For instance, encouraging caregivers to enrol their child in school rather than letting them enrol themselves promotes parenting. Additionally, increasing home-school communication is central to the SchoolLinks design, promoted through Meetings, Translation, and Enrolment Week. Parent volunteering and home learning support featured in the Homework Clubs as well as Meetings, where caregivers chose to prioritise their attendance of the meetings over other obligations and engaged in discussions with teachers on ways to support learning at home. Caregivers were involved in decision-making about the implementation of Homework Clubs and Safety to School, in terms of when, where, and who to conduct them, however their increased involvement in decision-making about school governance is not an explicit part of the SchoolLinks design. Lastly, community involvement was demonstrated through community leaders’ vocal support for SchoolLinks and importance of education, faith leaders making places of worship available for Homework Clubs, and caregivers’ willingness to host Homework Clubs. Similarly to other evidence generated on caregiver engagement amongst refugees [26], our research highlights the additional provisions required for effective implementation of Epstein’s model, including translation for multilingual communities, flexibility in implementation schedules, and socioeconomic support.

Gaps and further development

Gender- and disability-targeted strategies were not incorporated into the design, however we included the questions about disability and functioning in the enrolment form so that the teacher and school had increased awareness of children’s disability status, and two questions for the caregiver to be able to highlight additional needs of their children. Caregiver mental health, family wellbeing, income generation, and poverty were also beyond the scope of this intervention, underscoring the need for holistic, multisectoral intervention to have positive, systemic effect in conflict and low-resource settings [56]. While volunteer roles were accepted, and additional support voluntarily given by the PTA and SMC is promising, longer implementation would reveal if this will be sustainable. Implementation materials, including rewards and meeting refreshments, are small unit costs but may be prohibitive at scale and may lose their motivating power if given repeatedly. It is also important to recognise that the selection of strategies was shaped in part by WCA’s organisational mandate. While this was appropriate for the context of this intervention development, S1 Text (full list of action ideas) demonstrates the much broader scope of initiatives proposed by participants to promote school attendance and caregiver engagement.

The promising findings of this study warrant further research to validate the mechanisms of action and evaluate the effectiveness of SchoolLinks [10]. It would also be beneficial to assess SchoolLinks’ applicability in a different refugee context and potentially non-refugee contexts too. Considering the stagnation in government funding and reduction humanitarian funding for education [76,77], evaluation of the effects of SchoolLinks on learning outcomes would shed light on its potential to increase the cost-efficiency of standard education.

Limitations

This study had several limitations. First, omitting children and adolescents from the GMBs precluded their perspectives, which to some extent limits the validity of the causal map [78]. This was due to a combination of budgetary constraints, child protection considerations, and logistical parameters; however, we validated the model in subsequent work with adolescents. Second, it would have been beneficial to purposively include caregivers of children facing additional barriers to attendance (e.g., disability, longest distances). Third, power dynamics and social desirability between caregivers, teachers, partners, and WCA staff could have influenced input and feedback. We tried to limit social desirability by emphasising to enumerators and participants our need for genuine input and feedback, and that there were no right or wrong answers. We also encouraged participants to consider others’ perspectives (e.g., “what might your friends or neighbours suggest to improve this strategy?”) to lessen the perceived risk of giving ‘negative’ feedback. Fourth, the implementation was too short to measure student retention or dropout. We also acknowledge that the positive influence reported on attendance may be the effect of novelty and not sustained during longer implementation. Fifth, there was a risk of an “implementer effect”, where the thorough follow-up of the implementing staff member may have positively influenced (but therefore confounded) the uptake of the intervention.