Educational quality represents an essential pillar of a nation's social and economic Educational quality represents an essential pillar of a nation's social and economic progress, being closely associated with the formation of human capital and national competitiveness (Goczek et al., 2021). In this context, the success of the educational mission and student retention in Higher Education Institutions (HEIs) depend on the expertise, knowledge, and abilities of those involved (students, academic, and administrative staff). Both in the European Union and globally, the pressure to maintain high graduation rates and manage institutional performance requires that student retention strategies be continuously improved (Sá, 2023). This allows HEIs to consolidate their reputation, achieve strategic objectives, and offer exceptional educational performance (Vasilev et al., 2024).

However, the complexity of the student life cycle, along with the multitude of factors influencing dropout (of an academic, personal, social, financial, and institutional nature), makes designing effective interventions a difficult task (Pusztai et al., 2022).

To quantify the scale of this challenge, it is relevant to note that, according to a comparative study (Berlingieri and Bolz, 2024) based on PIAAC data in 18 European countries, individuals who leave university without obtaining a degree earn on average 8% more than those who have never attended higher education, but 20% less than university graduates. The results of this study suggest that these differences increase depending on the type of program: dropping out of academic programs is linked to incomes 25% lower than those of graduates from the same type of programs, while vocational higher education shows a difference of 9%. These figures highlight not only the economic dimension of the phenomenon but also its potential impact on the return on public and private investment in education.

Higher education dropout represents a persistent global problem, with average rates of 20–30% in many countries (Schnepf, 2017). Approximately 39% of bachelor's degree students in the US do not complete their studies within 8 years. In Europe, the EU average for early school leaving from education and training is 14.9%, with much higher values in countries such as Spain (31.9%) and Portugal (35.4%) (Schnepf, 2017). In the European context, the 2024 Bologna Process report indicates that over 25% of students in regions like Andalusia (Spain) drop out in the first year, a phenomenon often linked to socio-economic and academic factors. Globally, in developing countries such as Ecuador, the probability of dropout reaches 44.9% in private institutions, with 484 cases identified from a sample of 1,078 students over 5 years (Núñez-Naranjo, 2024). Recent studies, including an analysis of 17,328 students in Colombia, suggest that economic and academic factors are associated with 45% of the dropout risk, especially in the first 2 years, highlighting persistent inequalities in access to education (Barragán Moreno and González Támara, 2024).

A major obstacle to reducing dropout resides in efficient institutional management, namely how student support services are allocated, how pedagogical methods are adapted to student needs, and how at-risk students are identified (Barnes et al., 2024). Without a dedicated instrument for the effective analysis and prediction of dropout risk, an institution's capacity to successfully retain students and fulfill its social contract can be severely affected (Realinho et al., 2022).

Problems such as low academic performance, insufficient social integration, inadequate financial support, and dissatisfaction with the study environment are frequently associated with dropout, a significant loss of human capital, and substantial social costs (Gonzalez-Nucamendi et al., 2023). These issues are amplified in HEIs, where the student population is diverse, requiring varied support, and time and resource constraints for institutional staff are a frequent reality (Barragán Moreno and González Támara, 2024). Furthermore, student success often depends on multiple categories of interconnected factors (for example, of an academic, social, economic, or psychological nature), making optimal intervention difficult without an effective system that considers the multidimensional nature of student persistence (Nikolaidis et al., 2022; Matz et al., 2023). Student retention and dropout prediction are, therefore, important research topics, due to the widespread nature of this phenomenon and its specific challenges in different contexts and student demographics (Lorenzo-Quiles et al., 2023).

The central decision problem addressed by this research is the systematic identification and prioritization of factors influencing student persistence to improve overall retention efforts (Zaparan-Cardona et al., 2024). Although Multi-Attribute Decision-Making (MADM) methods and Machine Learning (ML) models have become a popular approach for predicting outcomes (Goren et al., 2024), their broad application to identifying, structuring, and prioritizing the complex factors influencing university dropout remains insufficiently explored within a pre-decision framework (Alshamsi et al., 2023). Researchers in educational psychology and computer science generally agree that a systematic approach to dropout prevention is superior to an isolated, one-dimensional approach (Vaarma and Li, 2024). However, there is still no exact consensus on what should be included in an all-inclusive and efficient student support and retention system, even though MADM methods have become one of the most popular topics in the decision literature (Laguna-Sánchez et al., 2021). Developing a structured foundation for these methods is therefore essential to ensure their effectiveness in educational settings.

An all-inclusive retention system refers to a set of distinct institutional practices (academic, social, and financial support) that, when used in combination, are reported to enhance students' abilities, motivation, and integration opportunities, thereby improving retention and completion rates (Leal et al., 2022). In the HEI context, understanding and implementing a systematic retention model is essential to foster an environment where students can maximize their engagement and achieve academic success (Drosos et al., 2024). There is clearly a need for a decision support framework that can integrate multiple variables (for example, academic performance, socioeconomic status, factors of a psychological/personal nature, institutional environment) to optimize the allocation of limited resources toward effective retention interventions (Kim et al., 2023). Due to the multidimensional and complex nature of these factors, traditional bivariate statistical models are often insufficient to fully examine and structure the complex interdependencies that lead to student dropout (Martins et al., 2023; Seo et al., 2024). Consequently, more advanced structuring techniques are required to capture the non-linear relationships between these variables.

The objective is to develop a foundational structuring phase for MADM-based framework that can organize the most important factors influencing university dropout, utilizing the capabilities of MADM (and related methods) to address a variety of criteria. This would improve the planning of institutional interventions and, consequently, student retention. However, the absence of an extensive, clearly defined, and consensus-based list of factors determining student persistence success represents an obstacle to the development of this decision support framework.

The primary question motivating the present study is, therefore: “What attributes (of an academic, social, personal, financial, and institutional nature) determine the success of student persistence in higher education and how can they be structured and prioritized for potential integration into a Multi-Criteria Decision-Making (MCDM) framework for the development of optimal retention strategies?.” To address this, the study pursues two secondary research questions: (SQ₁) What is the consensus in recent literature regarding the most frequent factors influencing dropout? and (SQ₂) How can these factors be grouped into clusters that reflect their interdependencies to reduce the complexity of future decision-making models? These attributes must be considered in all research and policy-making activities, whether qualitative or quantitative. The potential results of this research phase include a structured list of key dropout factors and a conceptual framework for the smart prioritization and allocation of institutional support resources in HEIs, serving as a robust foundation for subsequent MCDM implementations.

To answer this question, a systematic and broad literature review was conducted to extract the factors, which were then be classified into clusters. This classification is specifically designed to facilitate MADM/MCDM methods that require such a structure (e.g., the Analytical Network Process-ANP), including articles from the fields of Educational Psychology, Higher Education Management, and Data Mining/Machine Learning studies that offer insights into student persistence and dropout. Although the topic of student persistence and dropout has been extensively explored in the literature, existing reviews often do not specifically focus on the pre-MCDM phase of structuring and prioritizing multidimensional factors using a robust decision framework such as MCDM/MADMs. For example, some reviews may focus on a single category of factors (for example, of an academic or financial nature), but do not delve into the nuances and interdependence of a vast set of attributes, or they may discuss prediction models without specifically providing the structured criteria hierarchy necessary for institutional decision-makers. The novelty of this approach lies in bridging the gap between qualitative factor identification and the structural requirements of network-based decision models.

This research is distinguished by the systematic identification and classification of a vast set of factors of an academic, social, personal, financial, and institutional nature from the existing literature. In doing so, it provides a structured foundation for the subsequent development MADM/MCDM models aimed at optimizing retention efforts in HEIs.

To identify the key factors influencing student persistence and dropout, a systematic literature review was conducted. This involved reviewing the scientific articles that study student dropout in higher education. The systematic literature review adhered to the established guidelines for conducting rigorous reviews, relying on PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) principles (Figure 1), to ensure the transparency and reproducibility of the search process. The selection process focused on identifying primary research and review articles that explicitly discuss causal or descriptive factors of dropout to provide a consistent basis for the subsequent clustering analysis.

The database from which the articles were retrieved, Web of Science (WoS) Core Collection, together with the publication period (the last 5 years), were intentionally designed to ensure the quality and contemporaneity of the results. While it is acknowledged that using a single database may introduce a selection bias, the selection of WoS provides access to a high-quality and rigorously peer-reviewed body of literature, ensuring that the factors identified for the MCDM framework originate from validated research. Furthermore, by limiting the search to the last 5 years, the research focused on the most recent and relevant factors, recognizing that the dynamics of the student population and institutional support practices are subject to rapid change. This approach allows us to construct a conceptual framework that is extremely pertinent to the current challenges related to retention and decision-making in higher education.

Regarding the temporal limitation, although this strategy excluded seminal foundational works published before 2021 (such as the classic models of student integration), it is argued that the core dimensions of these theories are inherently captured and updated within the more recent literature. This deliberate trade-off was made to ensure the contemporary relevance and applicability of the framework to the post-pandemic educational landscape, where student dynamics have significantly shifted.

Following the application of the initial query (Formula 1), 80 results were returned. Applying the filter to include only articles published in the last 5 years led to the exclusion of 37 articles, leaving a total of 43 documents. The subsequent filter for document type, strictly limited to “Article,” further refined the corpus, resulting in 32 articles that were ultimately included in the final analysis and extraction of factors. Gray literature, including dissertations, conference abstracts, and non-peer-reviewed reports, was intentionally excluded to maintain a high level of methodological stringency and ensure data comparability.

This formula was employed to enhance the accuracy of searches (Table 1).

The concept of “student persistence and dropout” in this research is broadly defined to include a range of factors that influence a student's continued enrollment and eventual completion of studies, encompassing academic success, social integration, financial stability, and personal/institutional support. The selection of search terms was thus validated by their direct relevance to these persistence dimensions and their frequent appearance in the literature discussing higher education attrition and retention evaluation.

The inclusion criteria for the selection of articles in this systematic review were strictly defined to ensure the quality and contemporary relevance of the corpus. The primary inclusion criteria were:

Conversely, the following were established as exclusion criteria:

Each article was analyzed to identify the main factors contributing to student persistence and dropout (Table 2). Each literature source was assigned a code ranging from A to AF.

The value in the last column of Table 2, labeled “No. addressed elements characteristic of student persistence/dropout,” represents the total count of unique determinants associated with student retention or attrition that were identified and analyzed within each respective publication.

The process used to identify these factors involved a methodical content analysis of each paper. After obtaining the full text of the 32 articles, the coding process was conducted by two independent researchers with expertise in educational sciences and data analysis to minimize individual bias. They extracted all references, both explicit and implicit, to those elements, variables, or conditions that the authors presented as being influences, contributions, or determinants of student success or failure. The identified elements were subsequently classified and aggregated into a final list. To ensure the reliability of the extraction, any discrepancies arising between the findings of the two researchers were resolved through a formal reconciliation process involving discussion and mutual agreement. This approach ensured a rigorous and uniform identification process, aligning with established standards for inter-rater reliability in systematic reviews.

Table 3 provides a comprehensive summary of the factors deemed influential in evaluating student persistence and dropout within higher education. The complete set of factors, identified through the systematic review, is presented to ensure full transparency and to establish a foundational basis for future research.

The inclusion of the frequency of each factor is a deliberate choice to illustrate the extent of scholarly and institutional attention, and to potentially highlight biases toward elements that are more readily quantifiable or easily managed. Providing this detailed and transparent data is considered useful for researchers who wish to build upon this work, as it allows for the freedom to re-interpret the results and ensures a transparent research process, while preventing the duplication of effort in literature traversal.

Using a quantitative comparative approach, the factors were identified and analyzed based on their frequency of occurrence across the examined body of literature. This offers a broad view not only of the factors invoked in assessing student success, but also of the dimensions perceived as most essential in current institutional discourse and research.

It is necessary to point out that the frequency with which a particular factor is mentioned does not inherently reflect its objective value or intrinsic importance in determining student persistence. Instead, this frequency primarily indicates the degree of attention given to these factors by researchers and practitioners. In other words, the data show what is taken into account more often, not necessarily what is decisively impactful in all situations.

For instance, factors such as progress percentage (31 mentions) and student integration (31 mentions) are among the most frequently cited. Their high visibility is likely due to their prominence in established theoretical models and their direct observability. Similarly, internal policies (29 mentions) and financial situation (27 mentions) are highly cited, suggesting they represent institutional levers that are relatively clear to conceptualize, measure, and manage in the context of administrative decision-making.

In contrast, factors that may have a disproportionately high systemic impact, such as reputation of the institution (2 mentions), technological barriers (12 mentions), and evaluation and recognition of merit (12 mentions), are mentioned less frequently. This suggests an asymmetry between operational perception (what is easy to measure) and systemic impact (what fundamentally shapes a student's trajectory) that must be handled with caution in decision modeling.

To construct a pre-decision framework for multi-criteria decision-making in student retention, it is essential to gain a detailed understanding of each of the identified factors. The numerous factors extracted from the literature, which traverse academic, social, personal, financial, and institutional dimensions, outline a broad spectrum of influences that can significantly affect a student's outcomes, engagement, and ability to complete their studies.

Student success is not achieved in isolation but within a complex organizational ecosystem, where interactions between different variables determine not only retention rates but also the viability of the entire educational process. Analyzing these extracted factors enables us to map this territory, where persistence is not the result of a single dominant parameter but rather a network of multiple conditioning factors in a state of constant tension and balance.

At the core of this ecosystem is the student's capacity to navigate academic demands alongside personal challenges. Having current academic performance (23 mentions) as a key measure acts as an immediate indicator of a student's success and provides benchmarks for activity and progress. However, persistence is not solely about grades; it also requires the intangible elements of student integration (31 mentions) and strong tutor engagement (22 mentions). This is precisely where less quantifiable variables come into play, such as the quality of the student's social environment and the responsiveness of institutional support, which are associated with behaviors, decisions, and ultimately, success. Psychological difficulties and health (23 mentions) and work-study conflict (14 mentions) are especially important in situations of mounting pressure, where the issue is not just about institutional processes, but also about the student's willingness to persevere and the support available for joint efforts.

The systematic literature review provided a comprehensive set of 17 distinct factors influencing student persistence and dropout in higher education, extracted from 32 peer-reviewed articles published between 2021 and 2025 in the WoS Core Collection.

These factors were aggregated through a content analysis process conducted by independent researchers, ensuring a robust and reproducible extraction process and minimizing individual bias. Table 3 summarizes the factors, their frequency of occurrence across the corpus (ranging from 2 to 31 mentions), and their distribution. In this analysis, frequency serves as an indicator of scholarly attention rather than a measure of absolute causal weight, highlighting priority research areas while underscoring potential gaps in understudied elements. The most frequently cited factors, progress percentage (31 mentions) and student integration (31 mentions), were found to be dominant, appearing in over 96% of the reviewed studies. Internal policies (29 mentions) and financial situation (27 mentions) followed closely, reflecting institutional and economic levers that are readily measurable and administratively actionable. In contrast, reputation of the institution (2 mentions) was visibly underrepresented, suggesting a research inclination toward proximal, easily observable factors at the expense of more distant systemic influences.

The results of this investigation delineate 17 pivotal factors contributing to student dropout, offering a framework that balances academic, institutional, and personal elements. These determinants—ranging from progress metrics to institutional reputation—collectively underscore the complex interplay of influences precipitating attrition (Quincho Apumayta et al., 2024). This comprehensive structure is suggested to highlight the multifaceted nature of dropout while revealing substantial convergences with extant literature, illuminating nuanced divergences shaped by institutional type and regional demographics (Paura et al., 2025). By juxtaposing empirical outcomes with established scholarly insights from 2023-2025, it becomes evident that while core academic and economic pressures remain steadfast, emerging elements like technological barriers introduce timely adaptations to evolving educational paradigms (Paura et al., 2025).

Regarding progress percentage, low completion rates (below 60%) in initial semesters emerge as a harbinger of dropout, often manifesting through missed deadlines (Nurmalitasari et al., 2023). This observation appears to dovetail with scholarly consensus where stagnant progress leads to dropout probabilities 30-45% above progressing counterparts (González-Morales et al., 2025). While literature typically focuses on isolated GPAs, the current findings emphasize the rhythmic accumulation of credits as essential for momentum (Hanson, 2025). Such granular tracking extends beyond traditional static snapshots, addressing the 26% of dropouts linked to academic stagnation (Rahmani et al., 2024). This refinement aligns with predictive analytics advocating for early warning systems to preempt full attrition (Paura et al., 2025).

Student integration correlates with a 25% uptick in dropout among isolated students, as disaffection breeds a sense of institutional irrelevance (Yaghi and Alabed, 2025). This resonates with contemporary literature where disengaged students report 20-35% lower persistence due to diminished support (Quincho Apumayta et al., 2024). In hybrid environments, virtual disconnection further amplifies dropout by 15-20% (Rahmani et al., 2024). While some studies prioritize cultural factors, this research underscores a universal applicability, implying that mandatory cohort-building could reduce isolation-driven exits by up to 18% (Paura et al., 2025). The results further suggest that integrated environments foster resilience against academic stressors when reinforced by social ties (Nurmalitasari et al., 2023).

Internal policies exert a pervasive influence, with rigid protocols linked to 18% higher attrition (Avilés-González et al., 2025). This factor intersects with literature noting that flexible policy environments correlate with 22% improved completion (González-Morales et al., 2025). By portraying policies as both enablers and barriers, the study echoes reviews on administrative inertia while specifically foregrounding merit evaluation (Quincho Apumayta et al., 2024). Transparent frameworks help reputed entities retain 12-15% more students, highlighting the need for accessibility-driven reform (Yaghi and Alabed, 2025). Furthermore, streamlining the 30% of cases where policy bottlenecks exacerbate financial strain could serve as a low-cost retention lever (Bouchrika, 2022).

The financial situation remains an unequivocal juggernaut, with resource scarcity precipitating dropout in over 40% of cases (Hanson, 2025). This mirrors the universal thread where economic pressures drive 35-50% of departures in low-income brackets (Quincho Apumayta et al., 2024). Congruence is found in surveys where 41% of dropouts cite financial exigencies as the primary cause (Bouchrika, 2022). However, by disaggregating costs, this research reveals that peripheral expenses like transportation contribute to 22% of economic dropouts, a detail critical for rural cohorts (Paura et al., 2025). Holistic support packages could thus attenuate dropout by 25-30%, especially when addressing the immediate gaps identified in micro-financing models (Nurmalitasari et al., 2023).

Geographical and social origin entrench vulnerabilities, with rural students exhibiting 2.5-fold elevated risks (González-Morales et al., 2025). Literature corroborates this, as underprivileged students face 20-40% higher attrition due to support deficits (Quincho Apumayta et al., 2024). Unlike generalized overviews, the current findings highlight how geography exacerbates digital divides, adding 15% risk for non-metropolitan entrants (Paura et al., 2025). This suggests that origin-informed bridging programs could halve the dropout premium (Yaghi and Alabed, 2025). In rural contexts, gendered disparities further enrich literature's focus by underscoring intersectional vulnerabilities where conservative origins amplify barriers for female students.

Current academic performance dominates the academic cluster, with subpar grades signaling a 28% surge in dropout likelihood (Rahmani et al., 2024). This echoes meta-analyses linking low performance to 25-35% of total attrition (Nurmalitasari et al., 2023). When integrated with teaching quality, results show that suboptimal instruction inflates failure rates by 18%, converging with findings that attribute 26% of exits to pedagogical shortcomings (Hanson, 2025). Divergences appear in the bidirectional link between performance and health; mental strain may double the impact of poor grades (Quincho Apumayta et al., 2024). Thus, continuous assessment serves as a safeguard, preempting declines more effectively than end-of-term evaluations (Paura et al., 2025).

Psychological difficulties are implicated in 22% of dropouts through anxiety and burnout (Bouchrika, 2022). This aligns with post-2020 literature where mental health drives 22-30% of exits (Rahmani et al., 2024). Systematic reviews reinforce that untreated stressors elevate risks by 20% (Quincho Apumayta et al., 2024). Unlike earlier emphases on overt disorders, this research captures subclinical stress from work-study imbalances, advocating for wellness integration (Yaghi and Alabed, 2025). Embedding mental health protocols is suggested to avert psychogenic dropouts, especially where psychological factors intersect with high activity volumes (Paura et al., 2025).

Tutor engagement buffers multiple risks, with active mentorship correlating to 20% lower dropout (Avilés-González et al., 2025). Literature affirms that faculty-student interactions reduce attrition by 18-25% (Nurmalitasari et al., 2023). This study's emphasis on career vision development converges with studies linking advising to 30% persistence gains (Quincho Apumayta et al., 2024). Engaged tutors also bridge digital gaps in technological contexts (González-Morales et al., 2025). Similarly, quality of teaching inversely relates to dropout, with deficient quality causing disengagement in 22% of cases (Rahmani et al., 2024). Interactive methods boost retention by 15-20%, while feedback delays contribute to 12% of quality-related exits (Yaghi and Alabed, 2025; Hanson, 2025; Paura et al., 2025).

Volume and duration of activities contribute to 22% of burnout-induced dropouts (González-Morales et al., 2025). Literature supports this, indicating that high loads elevate attrition by 20-25%, while programs exceeding 4 years increase risks by 15% (Quincho Apumayta et al., 2024). Intensive STEM workloads amplify these effects by 10-15% (Yaghi and Alabed, 2025). The research extends prior models by showing that excessive activities precipitate mental health declines in 18% of cases (Paura et al., 2025). While earlier works focus on semester-specific overloads, the current findings emphasize long-term erosion, advocating for modular redesigns to reduce burnout-related attrition by 25% (Rahmani et al., 2024; Nurmalitasari et al., 2023; Avilés-González et al., 2025).

Age and gender modulate risks, with older males showing 15-20% higher vulnerability (Bouchrika, 2022). Non-traditional ages amplify dropout by 15-18%, often due to career interruptions (Yaghi and Alabed, 2025). Older students experience 20% higher work-study conflicts, while male attrition in engineering spikes by 12% (Quincho Apumayta et al., 2024). Divergences in weighting reflect contextual norms, with regional models showing 10-15% additional risks for older rural males (Paura et al., 2025). Tailored supports could equalize outcomes, lowering demographic-driven dropout by 20% (González-Morales et al., 2025). Older students' unclear trajectories further amplify risks by 25%, suggesting life-coaching as a remedial measure (Avilés-González et al., 2025).

Previous academic performance sets the trajectory, as weak high school records predict 28% higher risks (Rahmani et al., 2024). High school metrics forecast 25-35% of university attrition variance (Quincho Apumayta et al., 2024). Comparative studies find that prior records outperform socioeconomic indicators in early forecasts (Nurmalitasari et al., 2023). Integration with origin factors highlights how preparatory gaps inflate risks by 15-20% for first-generation students (Paura et al., 2025). While young entrants‘ risks tie to recent grades, older students' risks include professional gaps (Hanson, 2025). Remedial bootcamps tied to these prior assessments could mitigate 30% of entry risks (Yaghi and Alabed, 2025).

Career vision absence is linked to 25% demotivation-driven dropouts (Avilés-González et al., 2025). Clear goals extend persistence by 30-35%, framing vision as a psychological buffer (Quincho Apumayta et al., 2024). Vision deficits contribute to 20% of mid-program exits (Yaghi and Alabed, 2025). Institutional branding aids vision by signaling employability, with reputed programs boosting goal alignment by 18% (Paura et al., 2025). Intersections with age amplify risks by 15% when vision conflicts with life realities (González-Morales et al., 2025). Embedded career mapping is suggested to avert demotivation through longitudinal guidance (Nurmalitasari et al., 2023).

Work-study conflict affects 37% of employed students, precipitating exhaustion (Hanson, 2025). Employment exceeding 20 h weekly doubles dropout odds, acting as a disruptor in 37-40% of cases (Bouchrika, 2022). Flexible scheduling reduces conflict-related attrition by 25% (Paura et al., 2025). The research extends this by linking conflict to 20% of burnout instances (Quincho Apumayta et al., 2024). Older and rural students face 15% higher impacts due to commute and family roles (Avilés-González et al., 2025). Subsidized on-campus jobs could alleviate 30% of these pressures (Nurmalitasari et al., 2023).

Technological barriers hinder 20% of students through access inequities (Rahmani et al., 2024). Digital divides cause 12-15% risk increases, contributing to 18% of online attrition (Avilés-González et al., 2025). Rural linkage reveals that non-urban students face 25% higher disruptions (Paura et al., 2025). Tech hurdles increase dropout by 10% in under-supported groups (Quincho Apumayta et al., 2024). Unlike pre-digital literature, the factor's recency is critical, suggesting that provisioning devices could mitigate 22% of these risks (Yaghi and Alabed, 2025; González-Morales et al., 2025).

Merit evaluation reduces dropout by 15-20% through validated achievements (Nurmalitasari et al., 2023). Recognition incentives are linked to 18% higher retention (Quincho Apumayta et al., 2024). Bureaucratic delays in recognition erode trust, elevating attrition by 12% (González-Morales et al., 2025). Personalized feedback amplifies merit's impact by 10% (Paura et al., 2025). While biases against non-traditional students exist, automated evaluation tools could reduce related dropout by 25% (Avilés-González et al., 2025; Yaghi and Alabed, 2025).

Reputation of the institution lowers attrition by 15% via perceived prestige (Yaghi and Alabed, 2025). Reputed institutions retain 12-18% more students due to employability perceptions (Avilés-González et al., 2025). Reputation clarifies pathways and reduces demotivation by 20% (Paura et al., 2025). High reputation draws diverse cohorts but can challenge integration for marginalized groups, increasing risks by 10% (Quincho Apumayta et al., 2024). Branding investments thus amplify retention when paired with merit recognition (Nurmalitasari et al., 2023; González-Morales et al., 2025).

The primary originality and contribution of this work lie in the structural and mathematical transition it facilitates between qualitative synthesis and quantitative decision modeling. While existing literature extensively identifies dropout factors through narrative or systematic reviews, this research distinguishes itself by executing a rigorous pre-structuring of 17 variables specifically tailored for MCDM applications.

By transforming a vast body of scholarly discourse into a functional architectural bridge, this study provides a validated foundation that allows future researchers to bypass the exhaustive phase of manual literature traversal. The noutate of this approach is not merely in the identification of factors, but in their quantifiable organization. By mapping the frequency of these determinants, the research offers a direct reflection of the current “managerial gaze” and scholarly consensus within HEIs. This provides a ready-to-use framework for complex decision environments where interdependencies between academic, institutional, and personal variables must be calculated. Ultimately, this work serves as a “structural permit” for advanced modeling, ensuring that subsequent MCDM stages are built upon a data-driven, scholarly-validated baseline rather than arbitrary criteria selection.

A transparent acknowledgment of the boundaries of this research is essential to contextualize the findings and demonstrate a deliberate methodological design. A primary limitation is the strategic decision to rely exclusively on the WoS database. This was a conscious choice intended to serve as a quality filter, prioritizing scientifically rigorous and high-impact peer-reviewed research over the exhaustive but unverified breadth of gray literature or lower-tier databases. By selecting only high-caliber sources, the structural integrity of the resulting MCDM framework is ensured, providing a more reliable foundation for subsequent mathematical modeling.

Furthermore, the temporal focus was intentionally restricted to literature from 2021 onwards. This compromise was made to ensure the framework captures the current state of the art, reflecting the significant shifts in higher education paradigms following global disruptions. As this pre-structured framework is designed to be utilized and cited by future researchers, the decision to focus on the most recent data was a forward-looking strategy. Recognizing that the gap between publication and practical application can be significant, the use of contemporary data ensures that by the time this framework is fully integrated into institutional decision-making processes, it maintains its relevance and applicability. It is argued that providing an actual and updated baseline is more useful for the research community than a longitudinal overview that includes outdated educational contexts.

Additionally, a critical conceptual distinction is maintained regarding the use of frequency: visibility does not equate to causal weight. While the most frequently occurring factors (e.g., financial situation, progress percentage) indicate where current managerial and scholarly attention is concentrated, this frequency mapping is provided as a strategic asset. It allows future users of this study to bypass the redundant phase of literature traversal, offering a “ready-to-use” map of documented consensus. It is admitted that “invisible factors” may exist (variables with high individual impact but low reporting frequency). However, the current framework is positioned as a dynamic starting point, specifically designed to be refined through expert judgment in future MCDM stages, thereby bridging the gap between current visibility and long-term importance.

The shift from simply identifying factors to actively designing effective retention strategies necessitates the use of a structured decision-making methodology. Given the inherent complexity and interdependence of student persistence factors, an MCDM approach is suggested as essential to move beyond single-metric analysis. However, for complex techniques like the ANP and other hybrid MCDM methods, it is crucial to first establish hierarchical clusters of the identified determinants. This clustering process is necessary to both manage the analytical complexity, allowing for structured comparison of factors within thematic groups, and to mitigate the aforementioned risk of expert fatigue during the sensitive pairwise comparison phase. Therefore, the goal of this section is to present the theoretical framework by structuring the 17 extracted factors into logical clusters, ensuring the methodological feasibility and efficacy of any subsequent MCDM application for optimizing HEI resource allocation.

To move beyond simple frequency and identify the relational structure among these factors for MCDM integration, a comparative quantitative approach was employed. This involved classifying factors by frequency and, concurrently, analyzing their interdependencies through a correlation analysis across the articles. To move beyond simple frequency and identify the relational structure for MCDM integration, a comparative quantitative approach was employed. The analysis used the Phi Coefficient ϕ to measure the association between the co-occurrence of any two pairs of factors across the 32 articles. This coefficient is the mathematically correct choice for binary data (presence/absence). To calculate ϕ, a 2 x 2 contingency table was constructed for each pair of factors (Factor X and Factor Y), where:

The formula used for calculating this association is:

The results of these exhaustive calculations are systematized in Table 4, which serves as a comprehensive “menu” of scholarly visibility and interdependency for the scientific community. By providing these metrics, the study ensures that the framework remains aggregatable with future research from 2026 onwards, allowing other researchers to utilize this relational map without duplicating the systematic review effort.

The composition of the clusters in Table 5 is determined exclusively by the strength of the Phi Coefficient found in Table 4, identifying functional groupings based on statistical proximity.

The Academic-Learning Cluster is validated by the relational bond between progress percentage and current academic performance (ϕ = 0.23) and the association between the quality of the teaching process and volume and duration of academic activities (ϕ = 0.41). These metrics confirm that these variables form a distinct functional network representing curriculum interaction.

The Socio-Personal and Psychosocial Cluster is anchored by the statistical link between student integration and psychological health (ϕ = 0.23), alongside the relationship between career vision and work-study conflict (ϕ = 0.28). This demonstrates that personal development and social embedding variables act as a distinct mediating layer in the persistence network.

Finally, the Institutional and Structural Cluster is defined by the high structural synergies found in the dataset. Specifically, the strong association between financial situation and social origin (ϕ = 0.69) and their joint relationship with Internal policies (ϕ = 0.61) indicates that these factors form a tightly coupled foundation within the analyzed environment.

In conclusion, this structured framework transforms an unmanageable decision layer into a logical network. By grouping factors based on both the statistical associations derived from the Phi Coefficient analysis, covering academic relationships, psychosocial links, and structural synergies; and validated theoretical models, this study ensures that any future MCDM weights will accurately reflect the multi-layered reality of student persistence while maintaining mathematical consistency and practical feasibility.

This study successfully addressed the intricate challenge of student retention management by developing a rigorous theoretical framework grounded in MCDM principles. By systematically analyzing contemporary research from the 2021–2025 period, the study provides explicit answers to the research questions formulated in the introduction.

Regarding the first secondary research question (SQ₁), the initial frequency analysis identified a consensus on 17 key factors, highlighting the dominance of highly-cited, proximal factors like Progress percentage and Student integration, thus confirming their central roles in established retention theories. However, the critical methodological step was the use of the Phi Coefficient analysis to map the statistical interdependencies among all 17 factors, directly addressing the second research question (SQ₂). This quantitative analysis definitively confirmed that student persistence must be treated not merely as a sum of variables, but as a complex, interconnected system, thereby substantiating the strategic imperative for a structured MCDM approach. The integration of these findings provides a direct answer to the central research question: factors of a social, personal, financial, and institutional nature determine persistence through a web of statistical co-occurrences that can be formally structured for MCDM integration.

The originality and core theoretical contribution of this research lie in providing the necessary structural foundation for making complex MCDM models applicable to retention strategy. This was achieved by moving beyond simple frequency counts to a relational model, formally organizing the 17 factors into three distinct network clusters: Academic-Learning, Socio-Personal & Psychosocial, and Institutional & Structural. This three-cluster architecture is vital because it directly solves the inherent methodological constraint of expert fatigue prevalent in network-based decision-making techniques. By transforming the complex set of 17 factors into a streamlined, three-tiered structure, the framework drastically reduces the number of initial required comparisons, ensuring the methodological feasibility, consistency, and integrity of any subsequent advanced decision model used by administrators for strategic resource allocation.

Despite its rigorous methodology, the study has several limitations. The defined temporal scope (2021–2025), while ensuring contemporary relevance, intentionally excludes seminal pre-2021 foundational works. Furthermore, the reliance solely on the WoS database may introduce a database bias by potentially overlooking valuable research from non-indexed sources or specific geographical contexts. A theoretical limitation is the reliance on factor frequency as a proxy for scholarly attention, which may inadvertently overstate the operational significance of easily measurable metrics (e.g., academic results) while understating the systemic impact of distal, hard-to-quantify factors, such as institutional reputation or deep-seated socio-economic inequalities, thereby creating a potential mismatch between research focus and actual causal influence.

Based on these foundational findings, future research directions should focus on validating and expanding the utility of the proposed framework. The immediate next step is the empirical application and validation of the three-cluster structure using an appropriate network-based decision-making technique. This is necessary for deriving objective factor weights and transitioning the theoretical classification into a practical strategic planning tool that guides optimal resource allocation for retention efforts. Moreover, future studies should prioritize contextual and geographical expansion, systematically incorporating primary data from diverse and underrepresented HEI environments (such as specific regions in Eastern Europe, Africa, or Asia) to rigorously test the generalizability of the cluster structure and refine factor weighting based on specific local social and economic contexts.