The question being addressed is which interventions are most effective in reducing persistent absenteeism and how effectiveness varies across different levels of influence. This is outlined using the PICOS framework: the review focuses on secondary school students (Population), examining interventions designed to address persistent absenteeism (Intervention), with or without a comparator (Comparison), assessing school attendance as an outcome through various measures (Outcome). Only empirical peer-reviewed studies were included (Study Design), ensuring a rigorous evidence base.

The search strategy was developed in Ovid MEDLINE, informed by a review of key publications and refined with input from a librarian to optimise search terms and combinations. Once finalised, it was adapted for use in additional databases. The search incorporated terms related to persistent absenteeism (e.g., “emotionally based school avoidance”, “school refusal”, “non-attendance”), interventions (e.g., “program”, “mentorship”, “counseling”), and school culture (e.g., “school belonging”, “student-teacher relationships”, “classroom environment”). Boolean operators and database-specific subject headings were used to refine the search strategy. The full search strategy is available in the Supplementary Material. Citation and reference list searches were undertaken to explore potential additional studies.

Given the intersection of school absenteeism with both health and education research, multiple databases were searched to ensure comprehensive coverage; PsycINFO (Behavioural Sciences and Mental Health), ERIC (Education Research Information Centre), BEI (British Education Index), Child Development and Adolescent Studies, ASSIA (Applied Social Science Index and Abstracts) and MEDLINE (Medical Literature). The final searches were conducted on 16th April 2024.

Inclusion and exclusion criteria were structured using the PICO framework and informed by pilot searches and background literature. Eligible studies examined interventions targeting persistent absenteeism among secondary school-aged students (aged 11–18), defined using the UK threshold of missing 10% or more school sessions. Studies including mixed-age samples were included where at least 90% of participants met the inclusion criteria or where subgroup data were reported. Interventions designed to proactively support at-risk students were included, and interventions where attendance was a secondary outcome (for example, interventions targeting mental health or behaviour) were included where attendance outcomes were reported.

Studies were required to measure school attendance through any method, including school records, self-reports, or staff reports. Both qualitative and quantitative studies were eligible, provided they included primary data. Although attendance outcomes are often reported quantitatively (e.g., percentage of sessions missed), qualitative evidence was included to provide insight into mechanisms of change, student and staff engagement, and contextual factors influencing acceptability and feasibility. Qualitative data relevant to attendance could include student, parent, or staff descriptions of experiences of returning to school, perceived barriers and facilitators to attendance, and changes in attitudes, confidence, or school connectedness. This supports a more complete understanding of how and why attendance outcomes vary across settings. Studies reporting only feasibility or acceptability without attendance outcomes were excluded.

Commentaries, opinion pieces, and reviews were excluded to avoid secondary interpretation bias, but reference lists of reviews were screened for additional primary studies. Case studies with fewer than five participants were also excluded, as small samples limit the interpretability of findings and restrict generalisability.

The review focused on school-based interventions; therefore, pharmacological treatments, medical interventions, and those delivered in non-school settings were excluded. Given the broad and varied conceptualisation of school culture, the role of culture was assessed in result synthesis rather than used as a strict inclusion criterion.

To ensure relevance to the UK context, studies were limited to countries with comparable compulsory education systems, including Europe, the USA, Canada, Australia, and New Zealand; Comparable systems were pragmatically defined as those with universal compulsory schooling to at least age 16 and broadly similar primary and secondary structures. No date restrictions were applied, but studies were limited to English due to resource constraints.

Search results were uploaded and managed using the web-based reference management tool CADIMA (44) for efficient tracking and documentation (45). Duplicate studies were automatically removed using CADIMA and EndNote (46), followed by title and abstract screening to exclude remaining duplicates and obviously irrelevant studies. Screening was conducted by one reviewer, with 10% independently screened by a second reviewer to assess consistency in applying inclusion criteria; agreement exceeded 90%. Any uncertainties at this stage were taken forward to full-text screening.

Full texts were screened using the same process: one reviewer screened all studies, and a second reviewer independently screened 10% to ensure accuracy and reduce the risk of missing relevant studies (47). Disagreements were resolved through discussion. The screening process is outlined in the PRISMA flowchart (Figure 2).

A spreadsheet was used to extract and organise data from the included studies, with iterative development of categories. The final spreadsheet captured key details, including authors, year of publication, study title, study design, data type, population type, sample size, setting, intervention, comparison, outcome measures, time points of data collection, and findings, which were organised by attendance and other outcomes.

The Template for Intervention Description and Replication (TIDieR) guideline (50) was used to extract further details of interventions. This checklist provides a structured framework for describing interventions, including their rationale, materials and procedures, mode of delivery, personnel involved, setting, timing, and implementation details such as fidelity. These elements facilitated structured comparisons between strategies and their implementation, supporting a comprehensive understanding of the interventions reviewed.

Data extraction was conducted by a single reviewer, with the extraction framework refined iteratively to support consistency across studies.

Quantitative studies were assessed using the Effective Public Health Practice Project (EPHPP) Quality Assessment Tool for Quantitative Studies (51), a recommended tool for appraising public health interventions across different study designs (39). This tool evaluates selection bias, study design, confounders, blinding, data collection methods, and participant withdrawals, with each domain rated as strong, moderate, or weak. These ratings determine the overall risk of bias (low, moderate, or high). Two additional components, intervention integrity and analytical methods, further contextualise study limitations.

Qualitative studies were appraised using the Critical Appraisal Skills Programme (CASP) Qualitative Checklist (52), a widely recognised tool for assessing methodological rigour and the relevance of findings in health research (53). This checklist consists of ten prompts across three sections: validity of results, findings, and applicability to practice.

Critical appraisal was conducted by one reviewer, with a sample of studies independently assessed by a second reviewer as a quality assurance measure to ensure consistent application of criteria. A small number of discrepancies were discussed and resolved through consensus, which refined and clarified the application of appraisal criteria before continuing with the remaining studies.

Given the diversity in population characteristics, settings, and interventions, a narrative synthesis approach was chosen. This method integrates study results while considering statistical significance, confidence intervals, and risk of bias. A meta-analysis was deemed inappropriate due to the heterogeneity of study designs, which could obscure important contextual factors. Findings were interpreted in relation to study robustness rather than methodological exclusion, acknowledging the complexities of researching a highly mobile population with school engagement challenges (54).

The synthesis followed the Economic and Social Research Council (ESRC) guidelines for narrative synthesis (55) to ensure a systematic and comprehensive analysis. Interventions were grouped through iterative review of study characteristics, identifying common mechanisms of action. Eight categories were developed that reflected the dominant modes of support used to influence attendance [Multifaceted Interventions, Mentoring, Family Involvement, Counselling, School-based Healthcare, Social-Emotional Learning (SEL), Police Partnership and Incentive Schemes].

The KiTeS framework (38) was used to analyse intervention components across the microsystem, mesosystem, exosystem, and macrosystem. Intervention activities were coded to one or more ecological levels depending on their mechanisms of influence. For example, mentoring schemes typically operate at the microsystem through one-to-one relational support, but also influence the mesosystem by facilitating communication and problem-solving between school, family, and the student. These components were therefore coded across both levels. Coded intervention levels were then used to structure the narrative synthesis and to compare the distribution and focus of intervention strategies across studies.

Applying this framework allows an exploration of relationships between interventions and absenteeism, identifying key influences on attendance outcomes. Particular attention was given to school culture within the mesosystem and its interaction with exosystem-level factors such as school policies. The robustness of the synthesis was assessed to account for methodological variability.

The included studies were conducted between 1990 and 2017, with publication dates from 1993 to 2024. Fifteen were based in the US and one was in Australia.

Seven studies were experimental designs (all randomised controlled trials), four were quasi-experimental and five were observational. Thirteen studies collected only quantitative data, while the other three collected both quantitative and qualitative data. One study used a survey to assess students' willingness to attend, the other fifteen studies used school records to measure absence using various metrics. This included days absent, measures of consistent attendance, and percentages of students who were persistently absent. Two studies specifically assessed unexcused absences, and one subdivided absences by specific class attendance.

All studies were assessed using the EPHPP quality appraisal tool (51), with no studies achieving a global rating of “strong”. Eight were rated “moderate” and eight were rated “weak”.

This is summarised in Figure 3. Assessments of each study are detailed in Tables 4, 5.

The included studies had significant sources of bias, with common weaknesses in selection bias, blinding, and intervention integrity. No studies were rated as strong for selection bias, as participants were often not fully representative of the target population due to stratified sampling, risk-based selection, or exclusion of key groups. Recruitment challenges were also prevalent, with some studies reporting extended start-up periods and low participation rates.

Study design was rated strong for eight controlled trials and moderate for quasi-experimental and observational studies. Confounding risk varied; studies rated strong demonstrated no baseline differences between groups in key demographics such as sex, ethnicity, income, and education status. In contrast, weaker studies lacked baseline data or reported significant group differences.

Blinding was the most frequently rated weak domain, primarily due to unclear participant awareness of study aims and insufficient reporting on assessor blinding. None of the experimental studies detailed whether participants were informed of research aims, increasing the risk of reporting bias. Additionally, outcome assessors, often school staff, may have been aware of intervention status, leading to potential detection bias.

Data collection was also a concern, particularly for attendance measures, which were drawn from school records in most studies. While considered valid, such records may not always accurately reflect student presence due to tardiness, selective attendance, or procedural inconsistencies. No studies provided reliability checks for attendance data, leading to cautious assessments of validity.

Handling of withdrawals and dropouts varied, with nine studies rated strong for reporting attrition by group and maintaining over 80% completion rates. Some studies did not report dropouts, while others had attrition rates exceeding 20%, raising concerns about data completeness and potential bias.

Intervention integrity was inconsistently reported, with 12 studies failing to measure consistency, and nine acknowledging possible unintended intervention. Statistical analysis methods were generally appropriate, with nine studies following an intention-to-treat approach.

Overall, all studies presented some risk of bias, often linked to challenges faced by at-risk populations, such as conflicting priorities, low school engagement, and circumstantial barriers. Insufficient reporting undermined the certainty of quality assessments. These limitations are consistent with challenges reported in other school-based intervention research and are considered in the interpretation of results.

The qualitative findings from the three mixed studies were appraised using the CASP tool, which revealed potential limitations in validity, and consistently identified a lack of reporting on ethical issues, as shown in Table 5. The aims of this review were not necessarily addressed by all qualitative aspects of these studies, so further examination of limitations is detailed according to relevance in the synthesis.

The level influence of most interventions was individual, focusing on the needs and characteristics of students. For instance, mentoring and counselling schemes provided direct student support, incentives aimed to increase individual student motivation, and healthcare services supported individual health. Some interventions also targeted the family level of influence, through improving communication and links between schools and families, although no interventions explicitly targeted family dynamics or culture as a primary aim. There were no interventions that explicitly addressed school level in the interventions, with no direct strategies to influence the school environment. At the community level, there was involvement of community services within several interventions, and a police-school partnership represented an example of collaborative efforts (66). These initiatives still primarily served individuals, whilst recognising the importance of community support in addressing school attendance issues. Finally, at the societal level, one multifaceted approach advocated for creating a supportive societal environment for schools (59). Links to community services represented a potential role for society as a whole, but again, these were mainly focused on supporting individual students.

Given this distribution with a heavy emphasis on individual-level approaches, results are presented according to the eight intervention categories previously outlined. Table 6 provides a summary of outcomes, grouped by these categories. Aspects of the interventions will be applied to the bioecological KiTES framework (38) to enable an understanding of their outcomes, which is further explored in the discussion.

The effectiveness of multifaceted intervention strategies showed mixed results. These interventions, which align with the bioecological model to target multiple system levels, are gaining attention due to their potential to address the complex and diverse needs of students (34). However, the evidence supporting their effectiveness remains limited and inconsistent, particularly because of short durations, single-site implementation, and lack of detailed reporting of intervention components. While some studies support the use of these approaches (34, 73, 74), others have failed to demonstrate their superiority over simpler interventions (26).

A significant limitation in interpreting findings here is the lack of detail regarding the individual components of the intervention strategies, coupled with insufficient fidelity monitoring and limited efforts to mitigate or document contamination. This issue was particularly evident in the investigation of Full-Service Community Schools (FSCS), where the bespoke nature of each school's approach, based on specific needs, made it difficult to draw definitive conclusions. Although qualitative data from a small sample of schools provided valuable insights into what might make certain strategies effective, the credibility of these findings was compromised by concerns during quality appraisal. The researcher-participant relationship was not explored, and the philosophical stance and positionality of the researchers were not addressed. The analysis was described only as a “qualitative analysis using NVivo software”, with no further detail around the approach used (59).

It is unclear which strategies were used by schools with less successful outcomes, making it difficult to confidently attribute positive outcomes solely to the strategies discussed. Without stronger evidence of variation in strategy implementation and its corresponding effects, assessing how FSCS influenced students across the six implementing high schools becomes challenging. This also limits our ability to identify which system levels of the bioecological model were engaged by the intervention. Additionally, non-FSCS schools may have adopted similar strategies, further complicating the understanding of outcomes and the FSCS approach.

A second multifaceted strategy showed some promise in improving attendance, and provided valuable evidence of selective attendance, revealing a wide variation in how a student may attend between different classes (54). It also offered insights into the potential impact of implementation factors, with significant differences detected in the per-protocol analysis but not in the intention-to-treat analysis. However, the study was single-site with a short duration (nine weeks), limiting depth and confidence of conclusions.

Overall, the two multifaceted interventions included in this review showed mixed effects on attendance, highlighting that evidence for this approach remains inconclusive.

This review found weak to moderate evidence supporting mentoring as an effective strategy to improve attendance for at-risk students, with five of the eight mentoring studies demonstrating significant reductions in absenteeism, particularly when programmes were sustained over longer periods. Mentoring provides individualised support to students with poor attendance, fostering personal relationships and guidance in areas like communication and study skills. These interventions primarily operate within the microsystem of the bioecological model, with direct student-mentor interactions. Mentors also influence the mesosystem by connecting students, families, and school personnel. For example, mentors can advocate for students, helping resolve problematic student-teacher relationships, while also working with families to emphasise the importance of education and prioritising school. Such mesosystem interactions contribute to a more supportive learning environment that may help re-engage absent students, including those experiencing emotionally based school avoidance, where relational trust and reduced anxiety are key to re-engagement.

One study highlighted the importance of mentor-mentee rapport as a key implementation factor in the effectiveness of mentoring (68), with institutional barriers also playing a role. However, a previous meta-analysis of school-based mentoring did not establish a significant effect on attendance (75). The discrepancy may lie in the specific populations targeted in this review, where mentoring was focused on at-risk students, rather than a broad approach. Mentoring may be particularly effective in addressing barriers to attendance, academic struggles, or personal issues. Building trust and rapport with a mentor could provide a positive school relationship for at-risk students, potentially boosting engagement and attendance (76).

While mentoring theoretically could influence a student's exosystem, through mechanisms such as work with policymakers or school administrators, this was rarely discussed in the studies. Similarly, the potential impact on cultural norms or societal values in the macrosystem was not a primary goal of the mentoring programmes described.

In conclusion, mentoring primarily operates at the microsystem and mesosystem levels of the bioecological model and may be an effective intervention for at-risk students when sustained over a long duration. However, the evidence is inconclusive. Barriers such as lack of universal school staff support and institutional policies may hinder its effectiveness. Successful mentoring programmes can foster supportive relationships, increase student engagement, and potentially improve school culture, but there is no conclusive evidence to confirm these effects.

Family involvement was a key feature in five of the mentoring interventions and was also present in other strategies, which showed weak to moderate evidence of effectiveness in reducing absenteeism. Previous research has highlighted the value of family engagement in addressing absenteeism (77), and the evidence in this review supports this. Interventions that facilitated direct contact between schools and families have been found to be particularly effective (78).

Family involvement operates at the microsystem and mesosystem levels of the bioecological model. While mentoring programmes sought to increase communication with families, other studies placed greater emphasis on community links and practical support, contributing to a more comprehensive approach. This model, described as “overlapping spheres of influence”, sees the family, school, and community working together to support a student (79). Collaborative efforts from community partners create a support network that can enhance the effectiveness of interventions. A similar approach is seen in Full-Service Community Schools (FSCS), which have shown promise in building community partnerships and reducing persistent absenteeism in elementary school students (79). However, there was no clear pattern of effectiveness based on the extent of family involvement in the studies reviewed here.

The exosystem was less directly engaged, although sharing information between families and external parties could potentially impact factors like workplace policies or community resources (80). This potential was not explored in the included studies, limiting the interpretation of broader societal effects of family involvement.

This review found evidence suggesting that accreditation with national counselling standards in the US may reduce persistent absenteeism in some student groups. One study reported a 1.9% reduction in absenteeism for Hispanic students and a 1.7% reduction for economically disadvantaged students (56).

Counselling interventions likely operate at similar system levels as mentoring and family involvement, aiming to improve engagement through support. However, the accreditation programme also emphasises improving school culture, engaging both the mesosystem and exosystem through broader school policies and environments.

It is unclear whether the observed effects are linked to changes in school culture, as the study did not specify which accreditation components were implemented. The retrospective observational design limits the ability to establish causation, and there are concerns about potential bias from confounding factors. Nevertheless, the reductions in persistent absenteeism for Hispanic and economically disadvantaged students are notable, as these groups are known to be at higher risk of absenteeism (23, 81), making such interventions valuable.

This review found weak to moderate evidence suggesting that students' use of school-based healthcare (SBHC) may be associated with lower absenteeism for at-risk students. However, no clear causation could be established, and differences in healthcare access between the US and UK must be considered.

One study linked healthcare clinic outcomes with other strategies in a multi-faceted approach, while another retrospective observational study had high risks of confounding and attrition bias. Additionally, the second study was conducted over 30 years ago and may reflect outdated practices, such as recommending bedrest, which is a rare approach in modern medical practice. The remaining two studies were both retrospective, focusing on mental health services and prenatal care in school clinics (57, 64). These also faced confounding risks, with different rates of absenteeism and health status between groups at baseline.

In the bioecological model, SBHC impacts a student's microsystem through direct healthcare provision. Convenience and availability of care were considered key factors in included studies, alongside increased social support and positive role models, which reflect changes at the mesosystem level.

Existing literature indicates that the use of SBHCs, rather than simply their presence, can improve academic outcomes (82), and may also enhance school connectedness (83). However, the complexity of the mechanisms involved, including potential exosystem effects through health-positive school policies, requires further investigation.

One study suggested that delivering Social Emotional Learning (SEL) through a cognitive behavioural social skills curriculum may increase absenteeism. However, this finding was not consistently replicated across all time points, and the study's limitations (single-site, high risk of bias, confounding, and attrition) weaken the evidence (60). In this study, the curriculum was delivered to a class within a school with poor discipline and attendance. Improvements in self-reported behaviour were noted, which can be related to the students' mesosystem through peer relationships, influences in the school environment, and impacts on a students' sense of safety and belonging. The increased absences in the intervention group was unexpected, and is not replicated in other research. A previous meta-analysis evaluated the effect of universal school-based SEL interventions to find positive effects on similar social and emotional outcomes, alongside positive academic effects including improved attendance (84). They understood findings with a focus on environment, indicating the importance of universal implementation. This is in contrast to the intervention in this review, which was delivered to a single class. Delivery to a single class may overlook the benefits of continuity and synergy that can be achieved with wider implementation (85).

The authors identified existing norms and the schools' “expectation climate” as factors that undermined implementation. Despite the study limitations and the uncertainty about the effect on attendance, it is apparent that mesosystem factors, school culture and exosystem factors impacted both the implementation of the intervention and its effects.

A police partnership intervention led to a positive change in students' willingness to attend school, influencing all levels of the bioecological system. The intervention not only affected the microsystem (direct student action) and mesosystem (family and school involvement) but also extended to the exosystem through police presence, signalling legal support for school attendance. Additionally, the use of co-production and a procedurally just approach reflects a macrosystem change, aligning societal values with the need for legal compliance in education. Co-production fosters collaboration and more sustainable outcomes, and has been shown to be effective in tackling attendance issues (86). A multi-agency approach, including councils, healthcare providers, social workers, and community organisations, is key to overcoming barriers and improving support for at-risk students (87).

The final category of intervention studied was an incentive scheme, which resulted in an improvement in attendance from 82.4% to 92% (70). Students who did not receive family involvement participated in daily check-ins, celebrations, and rewards, applying behaviourist principles at the microsystem level through positive reinforcement (88). In this context, students with persistent absenteeism receive rewards for attending, which can create a motivating and encouraging environment. This approach contrasts with previous punitive measures they may have experienced due to absences, which can worsen attendance (89).

Additionally, group celebrations, viewed through social learning theory (90), encourage students to emulate positive behaviour observed in peers. While incentive schemes have had mixed results (30), they are valued for fostering a positive culture (91) and play a key role in absence prevention in multi-tiered strategies (34). However, the short study period and single-site design require caution in generalising outcomes and assessing long-term effectiveness.

The interventions across included studies were widely varied, and understanding links to outcomes is complex. Mapping interventions in the context of the KiTES bioecological framework (38) provides a basis to explore differences in direction and size of effects further. The included studies provide examples of interventions with influence across the bioecological system. Figure 4 provides an overview of results with application to the framework.

This summary highlights the lack of definitive conclusions regarding the application of these findings to a bioecological framework. There is no clear evidence that interventions must focus on one level of the system to be effective, nor a consistent correlation between engagement at multiple levels and effectiveness. While some multi-level interventions showed positive outcomes, others did not. Similarly, both comprehensive and targeted interventions have had mixed results, with some showing significant improvements in attendance and others having little or no effect. The effectiveness of strategies seems to depend on various factors, some within the scope of the interventions and others beyond it. What works in one context may not be effective in another.

This review applied a systematic methodology, enhancing the reliability and validity of findings. The use of transparent reporting and adherence to established frameworks provides clarity in decision-making and allows for reproducibility. A narrow focus on secondary school students, particularly those at risk of persistent absenteeism, ensured that the findings were relevant and specific. This targeted approach allowed for more precise insights applicable to real-world educational settings.

The search strategy spanned both health and education databases, with expert input to ensure coverage of relevant literature. This breadth allowed identification of studies across multiple intervention types and school settings. Incorporating a bioecological framework for analysis provided a structured lens, helping to interpret the findings more coherently. Additionally, the narrative approach allowed for capturing nuances in the interventions that quantitative methods might have missed, offering policymakers valuable context.

The inclusion of various study designs, from randomised controlled trials (RCTs) to observational studies, helped strengthen the overall evaluation. Combining study type provided insight into both effectiveness and real-world implementation. While RCTs offered stronger internal validity, their findings were mixed: several mentoring trials showed only small improvements in attendance, and in some cases no statistically significant effect. In contrast, quasi-experimental and observational studies more frequently reported larger reductions in absence, particularly where interventions were embedded in broader school environments (e.g., school-based health centres or counselling programmes). However, these designs are more vulnerable to bias, making it difficult to determine whether improvements were due to the intervention or to other contextual factors. Bringing these approaches together allowed the review to balance certainty in effect estimates with understanding of how interventions operate in diverse educational settings, and highlighted that outcomes may depend heavily on local implementation, student needs, and school context.

One of the main limitations of this review is the small evidence base, with most studies originating from the US and one from Australia. This limits generalisability to the UK context, especially considering cultural and policy differences. Absenteeism is recognised as a concern across Europe and other comparable education systems, but evidence remains limited internationally, as well as specifically in the UK. While a mix of study designs was used, the total number of studies (16) is insufficient to provide comprehensive insights into all intervention strategies within the bioecological framework.

There are also methodological limitations relating to the review process itself. The search strategy focused on school-based terms combined with “absen*” and related concepts. While this approach captured a wide range of studies addressing absenteeism, it may have missed interventions described solely using terms such as “persistent absence” without explicit reference to “school”. Additionally, terms such as “truancy” were not separately included, which may have limited retrieval of studies using older or justice-focused terminology. This is only partly mitigated with citation and reference searching.

Inclusion criteria required a measure of attendance for inclusion in this review; although there was no limitation on data type, a quantitative approach was usually adopted. Limited qualitative data may have overlooked deeper insights into the mechanisms behind interventions. Some studies failed to report key details on intervention integrity, such as frequency and duration, making comparisons difficult. The lack of detail in reporting limited the understanding of inconsistent findings, for example as seen with the implementation of Check & Connect (C&C) mentoring.

The heterogeneity of interventions presented a challenge in synthesis. Many programmes operated across more than one level of the bioecological model and contained multiple components spanning different intervention categories. As a result, some interventions were represented in more than one category. This reflects the inherently multi-component nature of attendance interventions rather than duplication of findings. However, it does mean the boundaries between categories should be interpreted as indicative rather than discrete. In some cases, separating components for categorisation risked oversimplifying how interventions functioned in practice, while retaining multi-level programmes within single categories could obscure important mechanisms of action. Although the synthesis aimed to balance capturing nuance with drawing out broader patterns, categories with only one or two interventions limit the strength of conclusions. A larger evidence base would allow more confident comparison within and across categories.

It is also important to note that the 10% threshold commonly used to define “persistent absenteeism” represents an administrative boundary rather than a distinct behavioural category. Interventions act both preventively, targeting pupils at risk of reaching this threshold, and responsively, supporting those already classified as persistently absent. Consequently, the included studies span a continuum of need, and effectiveness is likely to vary within that context.

Another concern is the risk of bias, which was moderate to high in several studies. This was mainly due to non-representative samples and significant dropout rates, common in populations at risk of absenteeism. Inconsistent attendance data and selective reporting further compounded the challenges in establishing the effectiveness of interventions. Additionally, conducting interventions in real-world settings introduced challenges such as recruitment and variable implementation, which further limits the generalisability of findings. Many studies also lacked detailed demographic reporting, such as ethnicity and socioeconomic status, which are known risk factors for absenteeism and can introduce bias into outcomes when unmatched.

As outlined in the introduction, similar quality findings have been reported in other reviews of school-based interventions using a range of quality appraisal tools. Reviews applying the EPHPP in comparable contexts have also found few studies rated as strong (92, 93). This suggests that the lack of high-quality studies reflects broader methodological constraints in evaluating real-world school-based interventions, particularly those involving at-risk or hard-to-engage populations, rather than an artefact of specific appraisal tools.

The process of study screening and appraisal also introduced potential bias. While quality appraisal was carried out systematically, only a small proportion of studies (3 out of 16, 18.8%) were independently assessed by a second reviewer. This falls short of gold-standard systematic review practice and may have influenced judgements of study quality.

Finally, where descriptions of interventions and their implementation was not reported in detail there was a need for interpretation. Conclusions are inherently shaped by judgements, which were in turn influenced by biases from author positionality.

This review offers valuable insights for policymakers in planning interventions to address persistent absenteeism in secondary school students. Several interventions showed positive impacts on attendance, with four studies improving attendance by over 5%, equivalent to more than 8 extra days in school over an academic year (54, 57, 62, 63). Additionally, two other studies showed improvements of over 2%, approximately 4 extra days (61, 68). These findings suggest that targeted interventions can make a difference, although the practical significance requires careful interpretation.

The meaning of any attendance change is highly context dependent. For some students, a few additional days in school may have limited academic impact, particularly where disengagement is deep-rooted. For others, such as those returning from emotionally based school avoidance or prolonged absence; small, incremental gains can reflect important steps toward re-engagement, trust-building, and reduced anxiety. Attendance outcomes should therefore be considered alongside baseline attendance, reasons for absence, level of need, and the nature of support provided.

Students at risk of persistent absenteeism often face challenges that also impact their ability to participate in interventions, which are in turn reflected in enrolment and attrition issues in research. Effective implementation planning should address these challenges to ensure interventions are accessible and have maximal impact. It's crucial to consider specific subgroups of students, including those with economic disadvantage, minority ethnic groups, and those with special educational needs. Interventions need to be tailored to address the unique barriers faced by these students.

Differential findings between intent-to-treat and treatment-on-treated groups emphasise the importance of implementation fidelity. Ensuring high-quality, consistent implementation is key to achieving programme success. Policy planning should include strategies for ensuring realistic yet ambitious implementation levels, including co-production, clear guidelines, robust training, and ongoing support.

This review highlighted the limited evidence on effective strategies for addressing persistent absenteeism in secondary schools. The scarcity of eligible studies and the lack of UK-based research underscore the need for more robust studies in this area. Persistent absenteeism continues to be a pressing educational and public health issue, presenting an opportunity to further investigate interventions with well-designed, high-quality research. Cost-effectiveness research will also be vital to guide resource allocation and ensure interventions benefit vulnerable populations.

There was limited evidence to understand the relationship between school culture, interventions that impact it, and their effect on student attendance. This reflects that many interventions may focus on broader objectives such as school culture, rather than directly addressing persistent absenteeism. In the full-text review, almost 30% of the studies (41 out of 138) were excluded due to not focusing specifically on persistent absenteeism or at-risk students. Future reviews could benefit from broader inclusion criteria to capture the full scope of cultural interventions, especially given the UK's focus on promoting a safe, supportive learning environment in schools. More primary research is needed to explore how interventions that aim to improve school culture specifically impact students with persistent absenteeism.

Future research should incorporate longer follow-up periods and consider outcomes beyond attendance alone, including attainment, wellbeing, and post-16 transitions. Understanding whether any change is maintained, and under what conditions, is essential for informing long-term policy and resource planning.

Evaluations should account for baseline attendance, reasons for absence, level of need, and demographic and contextual factors, including culturally specific barriers, accessibility, and economic influences. Demographic subgroups should be recorded, and qualitative research could provide valuable insights into designing more tailored interventions. This would include understanding culturally specific barriers, improving accessibility, and addressing economic factors that influence attendance.

The involvement of students experiencing persistent absenteeism in the development and refinement of interventions was also largely absent from the studies reviewed. Given the complex and individual nature of school avoidance, co-design approaches involving young people, families, and school staff could help ensure that interventions are acceptable, relevant, and responsive to lived experience. Future research and programme design including participatory or co-production methods can support more tailored and impactful interventions.

The application of the bioecological framework in this synthesis provided a useful basis for exploration of intervention efficacy and underlying mechanisms, although limited. Adopting this approach with a larger evidence base in future research can establish a deeper understanding of how interventions interact with different levels of the bioecological model to influence persistent absenteeism.

Finally, future research should focus on the quality, depth, and role of intervention components, rather than trying to separate them artificially, as students' needs are interconnected and require multifaceted responses. In particular, examining how specific components function would strengthen understanding of how interventions work in practice, for example, the way family involvement ranges from one-way communication to genuine partnership. Realist evaluation approaches could support this by exploring what works, for whom, and in what contexts, helping to clarify mechanisms and inform more tailored, impactful interventions.