Undergraduate Dropout in Colombia: A Systematic Literature Review of Causes and Solutions

ABSTRACT Higher education dropout rates in Colombia are the second highest in Latin-America. Almost 50% of students who start an undergraduate program in Colombia drop out. In this paper, we present a systematic literature review that surveys publications related to university dropout in Colombia between 2000 and 2021. This review followed the Kitchenham guidelines. Databases such as Publindex, Scielo, Wos, Scopus were reviewed. To create cause and solution taxonomies, we identified causes and/or solutions reported by researchers in each revised article. Each cause/solution was then grouped using the university dropout taxonomy proposed by Castaño. 107 papers, 66 different causes, and 62 proposed solutions related to university dropout were reported in the papers analyzed. The results suggest there is an increasing interest in understanding (i) the undergraduate dropout phenomenon, and (ii) the use of data science to solve the problem. These studies also evince a lack of integration between stakeholders for developing crosscutting solutions. The information related to some of the reported solutions is not sufficiently developed to enable a better classification, or they lacked information on implementation, results, or impact. This makes it difficult to make progress with designing new strategies based on previous studies.


Introduction
College dropout is a common phenomenon across different countries, and is caused by similar factors including study conditions, academic integration, social integration, personal effort and motivation, student sociodemographic background, and external conditions, among others Kehm et al. (2020).Dropping out from college harms student futures, limits society's evolution, and affects university income (Ramírez et al., 2020).According to the OECD (Organization for Economic Co-operation and Development), countries like Spain, Germany, and Denmark have dropout rates greater than 40%; and the average per-country dropout rate in over 25 OECD members during 2010 was 39% (OECD, 2010).The situation is worse in Latin America, with dropout rates of more than 55% in countries like Mexico.
According to the World Bank, between 30% and 50% of students in Colombia drop out of undergraduate programs at some point in their academic career (Ferreyra et al., 2017).Reacting to this situation, the Colombian government, local universities, and other organizations have permanently researched this phenomenon, finding various common causes and individual, academic, institutional, and socioeconomic triggers.These four main categories of causes for university dropout have been used by the Colombian government and the universities to tackle the problem (Burgos Mantilla et al., 2009).However, other factors like the COVID-19 pandemic have aggravated this problem, leading to increased dropout rates in recent years (Michael Onyema et al., 2020).
To address the university dropout problem in Colombia, various aspects of this phenomenon must be understood, including population size, main actors, and, above all, the causes of academic dropout.In Colombia, about 1,552,078 university students were enrolled as of ( 2019) and there were 290 universities distributed throughout different regions of the country; 209 (70%) of these are private higher education institutions (Melo-Becerra, Ramos-Forero, & Oswaldo Hernández-Santamaría, 2017).Concerning Government institutions, the Colombian National Ministry of Education is the organization responsible for controlling the country's universities.The local government has developed several initiatives to reduce the university dropout rate.These include ICETEX (Spanish Acronym), the government entity in charge of providing financing to students without economic resources, and SPADIES (Spanish Acronym), the organization in charge of measuring and reporting dropout indicators in higher education.However, despite the different efforts to reduce the problem, Colombia is 2 nd country in Latin America with the highest university dropout rates (Ferreyra et al., 2017).
Between 2000 and 2021 the Colombian government, organizations, and universities have proposed a significant number of solutions to reduce undergraduate dropout rates.However, although dropout rates decreased over the last 10 years in Colombia, desertions rates are still quite high (Lyche, 2010).We performed this study by looking for relevant documents on national, regional, and international databases including the Web of Science Group WOS (Clarivate, 2021), Scopus (Elsevier, 2021), Publindex (Gobierno Nacional de Colombia, 2021), and Scielo (FAPESP, 2021).In our work, we identified and organized the causes and proposed solutions for university dropout rates between 2000 and 2021.University dropouts that can be grouped according to the model proposed by Castaño et al. (Castaño et al., 2006) where they suggest that there are four main causes for university dropout: individual, socioeconomic, academic, and institutional.
• Individual determinants: These causes are of a personal nature, related to the context in which students are immersed, for example student life situations, interests, beliefs, motives, personality, and temperament (Bäulke et al., 2018).Different situations can influence a student's intention to drop out, including anxiety, stress, depression, and procrastination, among others (Rodríguez-Arce et al., 2020).• Socioeconomic determinants: These include social factors like family, friends, culture, and neighborhood, among others, that influence the decision to drop out (Rodríguez-Arce et al., 2020).Economic determinants include a lack of resources, lack of employment, incompatibility with work, and debts (Yair et al., 2020).• Academic determinants: These factors include course failures, course repetition, low performance, poor understanding, inability to perform academically, lack of learning with teaching methodologies, among other problems (Acevedo Calamet, 2020).• Institutional determinants: These factors are related to the university institution, and can include regulations, policies, teaching methods, tuition, location, schedules, and curricular flexibility, among others (Tungkunanan, 2020).
For the SLR reported in this paper, we also used the model by Castaño et al. to collect, group and analyze existing publications.

Methods and protocol
This SLR was developed following guidelines proposed by Kitchenham (2004) Two authors of this paper defined the databases to be queried and developed an instrument to collect the information (i.e., Excel spreadsheet).Weekly work sessions were carried out to analyze the documents collected.Details of each step in the analysis pipeline are detailed below.

Research questions
Knowledge of previously published academic works is important when studying the dropout phenomenon in Colombia Organizing and aggregating this information allows new researchers to access information on university dropouts in Colombia in a more agile manner.Knowing of the amount of research and the years in which it was done can provide valuable information to readers that intend to make evidence-based decisions.The following research questions are addressed by this study: • RQ1.What papers were published on university dropout in Colombia from 2000 to 2021?
• RQ2.What causes for university dropout in Colombia were identified in/reported by the literature?• RQ3.What strategies/solutions were proposed and/or implemented to reduce university dropout rates in Colombia, as reported by existing literature?
Regarding RQ1, we identified the total number and types of documents (i.e., article, book, conference proceedings, others) related to university dropout in Colombia published between 2000 and 2021.To answer RQ1 we also calculated the frequency of publications during the analyzed period.For RQ2, we identified different causes of university dropout by reading the selected documents.These results were organized in a taxonomy that included the frequency of publication in each category included in the taxonomy.We used the four categories proposed by Castaño et al. (Castaño et al., 2006) as a starting point for the taxonomy.For RQ3, and similarly to RQ2, after reading the selected documents, we organized planned/reported solutions in a taxonomy.

Quality assessment
After executing the queries and selecting an initial list of 132 papers, we defined the papers to be included in the study using a process based on the quality assessment scale proposed by the Center for Review and Dissemination (CDR) of the University of York (Tacconelli, 2010), which is also used by Kitchenham (Kitchenham et al., 2009).
In our case, to evaluate the quality of the information in a given paper "p" we selected two criteria: (i) Q 1 (p), related to the research method used by the publication, and (ii) Q 2 (p), the type of publication.Thus, a given paper p was assigned a value of Q(p), where was assigned to each document based on the following rules: 3 points were assigned to papers that used Randomized Controlled Tests; 2 points to papers that used Quasi-experimental studies; 1 point to publications with observational studies; and 0 points to the rest.Values for Q 2 (p) were assigned as follows: 3 points for peer-reviewed articles; 2 points for papers in conferences proceedings and books; and 0 points for the rest.The documents that obtained a Q(p) greater than or equal to 2 (seeking to ensure a minimum of quality, depending on the methodology or peer review) and at least one point in each of the evaluated criteria were included in the SLR.At this point the list was reduced to 110.Finally, we discarded duplicate papers, ending up with a total of 107 papers.The list of initial papers and their quality points can be found in our online appendix [Link].

Method of analysis
During the article review, we first checked whether the article title was related to our research.We then reviewed the abstract to see whether it contained information that would contribute to our study.Finally, we reviewed the documents by reading the results, discussion, and conclusion sections, while looking for information on the causes of and solutions for college dropout.All the information was recorded in a textual instrument, i.e., an Excel table that can be found in our online appendix [Link].One of the authors (hereinafter referred as A1) was in charge of information collection and a second author (hereinafter referred as A2) reviewed the paper selection process.Both A1 and A2 implemented teamwork sessions to group and analyze the collected articles.We extracted the following information from each article selected for the SLR: publisher (journal, conference, website), authors, year of publication, reported causes of dropout, proposed solutions, and implemented solutions.A third author -expert in pedagogy -carried out a second review to confirm the taxonomic categorization of the causes and solutions found.Finally, all three authors discussed the results and drafted the discussion and conclusions together.
To create cause and solution taxonomies, we identified causes and/or solutions reported by researchers in each revised article.Each cause/solution was then grouped using the university dropout taxonomy proposed by Castaño et al. (Castaño et al., 2003).During the the first phases of the research process, we found multi-causality and a diversity of solutions proposed in most of the articles reviewed.For this reason, a new category not existing in previous taxonomies had to be created called the multi-variate category.The multi-variate category groups causes/solutions that belong simultaneously to different categories of the Castaño model (e.g., institutional and socioeconomic).

RQ1. What papers were published on university dropout in Colombia from 2000 to 2021?
After applying the quality criteria to an initial list of 132 documents, 107 publications were finally included for this article: 86 peer-reviewed articles, 14 conference papers, 5 literature reviews, and 2 book chapters.39 of these publications were in English and 68 in Spanish.The number of papers published per year can be found in Figure 1.It is worth noting that interest in studying university dropout in Colombia has grown in recent years.Only 15% of studies reported were published between 2000 and 2010, while 85% of the published documents were reported for 2010 and 2021.In addition, Table 1 lists the papers retrieved with each query after screening.There are more studies published on undergraduate dropout in Colombia in Spanish than in English.While 64% of the publications were found with Spanish keywords, only 36% of the articles were retrieved using English keywords.82% of them were published in Spanish.
Regarding the types of publication, over 80% of the documents are peer-reviewed.Of these 86 articles, most of them were published in academic journals of education, social science, computer science, engineering, psychology, economics, and health, among others.Finally, we found increased academic publications in engineering journals.Digitized data generated by universities are being used by data scientists to learn about potential dropouts.Data science is currently widely used for academic purposes.Of the 107 analyzed documents, 65 report causes for university dropout.These causes are included in our taxonomy with 5 top-level categories (socieconomic, individual, institutional, academic, multivariate).Most of the documents analyzed contained more than one type of dropout cause under the Castaño Model.For this reason, as mentioned before, we decided to include the multi-variate category in our taxonomy.In the case of this multi-variate category, the taxonomy uses 1 second-level categories that combine two, three, or the four main categories.The taxonomy can be found in Figure 2. The multi-variate category is the largest category for dropout causes with 37 documents, followed by the academic category with 11 documents, socioeconomic with 8, individual with 5, and institutional with 4.

RQ3. What strategies/solutions were proposed and/or implemented to reduce university dropout in Colombia, as reported by the existing literature?
Of the 107 documents analyzed, 62 reported solutions for university dropout, while 45 documents did not report/propose any solution.The strategies or solutions in most of the documents retrieved were not very descriptive, i.e., they did not contain objectives, methodologies, indicators, or any other detailed information.Even so, they were included in our taxonomy.Reported/proposed solutions were also categorized without "questioning" whether they were effective or not.To classify the strategies, we reviewed the literature from two perspectives.With the first of these, we wanted to identify who is responsible or where the responsibility lies for dealing with the dropout problem according to each strategy According to Castaño Model, the proposed strategies intended to attack or solve (e.g., if their goal was to improve students' literacy, they were categorized as academic, but if they intended to change the curriculum, they were categorized as institutional).The main result of the first approach (see Figure 3) was that institutions were reported as responsible for dealing with university dropout solutions in 75% of the cases, while the remaining 25% indicated that the responsibility lay with someone, or something related to the other four categories (individual, socioeconomic, academic, and multi-variate).We emphasize that private companies, those that benefit the most from professional graduates, do not provide solutions for reducing dropout according to the reviewed literature.No evidence of specific actions by communities, not-for-profit organizations, social movements, among others, to reduce university dropouts was reported.

Discussion
The results of the SLR suggest that an interest in understanding the phenomenon of undergraduate dropout in Colombia has grown between 2000 and 2021.According to our analysis of academic publications issued each year (Figure 1), the growth trend shows that university dropout is increasingly being researched by the social sciences.Recently, the engineering field has shown an interest in dropout, applying predictive mathematical models, big data (Amaya-Amaya et al., 2020;Viloria et al., 2019) virtual environments, e-learning (Sternig et al., 2018) and artificial intelligence (Viloria et al., 2018), among others, to the problem.
A comparison of the causes and solutions of undergraduate dropout in Colombia reported in this article with those reported in other countries around the world returned similar causes and solutions (Table 1).In agreement with other published SLRs (Aina et al., 2021;Estévez Ceballos et al., 2015;Kehm et al., 2020;Orellana et al., 2016;Sobral & Oliveira, 2021), and compared with our findings, the nature of undergraduate dropout is complex and multi-causal.The motivations for dropping out are influenced by individual, social, economic, and academic factors.Our SLR shows that, in most cases, combined causes of dropout increase the risk of dropout.For example, most than half (37/66) of the   reported causes of dropout show that the cause of university dropouts is multi-variate, with the socioeconomic variable the one that as the most influence when combined with another cause of dropout.
As a single category, academic causes are those most reported by the researchers included in this SLR.Inadequate primary and secondary education affects college performance (Orlandoni Merli et al., 2017) specifically in areas like literacy (Olave- Arias et al., 2013), basic sciences (Cadavid & Gómez, 2015), and mathematics (Orlandoni Merli et al., 2017).In this sense, strengthening basic education could have an impact for reducing dropout.This is not an easy task, as in some cases low student performance is due to a lack of quality of the high school itself (Gama et al., 2015).High school quality measurement tools, like the Saber 11 tests, have shown a direct correlation between performance on the exam and subsequent university dropout (Guerrero, 2018).It therefore serves as an instrument to analyze the probability students dropping out from the moment they start their university studies.
Multiple solutions are proposed/implemented to reduce undergraduate dropout for academic reasons.These include academic tutorials (Orlandoni Merli et al., 2017), literacy skills courses (Olave- Arias et al., 2013) (Arias et al., 2013), and remedial/leveling up courses (Suárez-Montes & Díaz-Subieta, 2015b).As for institutional solutions, many of those reported have an academic focus.A highlight includes using technology to provide early warning for students with academic failings (Amaya-Amaya et al., 2020;Silva et al., 2019).In addition, an application of academic performance predictions (Acero et al., 2019) is helping universities to provide better academic support for their students (Alban & Mauricio, 2019).Unfortunately, no efforts between high schools and universities or between groups of universities are reported in the reviewed literature.In general terms, it seems that universities solve their academic dropout problems individually.
The second category with the most reported causes for dropout (excluding the multivariate category) was the socioeconomic category.Two articles included in this SLR state that undergraduate dropout rates are due to structural issues in Colombia (Bravo Castillo Armando Mejía Giraldo, 2010;Zárate Rueda & Mantilla Pinilla, 2015).Issues including low student income (Acevedo et al., 2015), a lack of job opportunities (Paula et al., 2018), and family (Ariza Gasca & Marin Arias, 2009)￼ have a considerable influence on the reasons for dropping out of an undergraduate program.Gender issues also affect both men and women.While women are feminized for bringing up children and maintenance of the home ￼ (González-Bedoya & Molina-Osorio, 2020)￼ men, in turn, are responsible for working to support the family.Demographic issues like student age or region of origin affect their probability of dropping out ￼Lopera (2008)￼ Finally, some research indicates that the current student loan system is not effective enough ￼Duran Cabieles et al. (2019)￼ Socioeconomic solutions, that should be led by the government are reported as insufficient for dealing with the phenomenon of university dropout ￼Guerrero and Soto Arango (2019)￼ Classmates also directly impact a student's educational experience, influencing dropout when there is a lack of proper social adaptation ￼Andres Rocha- Ruiz et al. (2018)￼ With regard to the institutional category, in certain cases, university regulations may affect students who cannot adapt to university rules and culture.Pedagogical strategies and curricular factors also influence the decision to drop out (Ramiréz, 2012).Some authors point out that certain private universities allow underqualified students to enroll, motivated by an interest in increasing profits and meeting institutional goals (Santos, 2006).The high cost of university tuition may also be a cause of undergraduate dropout (Acero et al., 2019).Finally, aspects like the social responsibility of higher education institutions (El-Kassar et al., 2019) or the perception of the university brand (Gutiérrez Torres & Moreno Hernández, 2019) also affect student permanence.Universities with prestige, a good reputation, and community involvement attract more students.
Although the institutional causes are the least reported, it is found that institution-led solutions are the most widely reported in the papers reviewed.Similarly, the second most reported solutions, after those that address multi-causal situations, are those that address institutional causes.Solution design has probably focused on educational institutions, since it can be viewed as the stakeholder with the greatest possibility for action, planning, and monitoring.
In this category, the most common solutions are assistance plans.These are reported by many universities as solutions for dropout (Kadar et al., 2018;Orellana et al., 2016;Suárez-Montes & Díaz-Subieta, 2015a;Timaran Pereira & Caicedo Zambrano, 2017a;Valencia Vargas & López Palacio, 2020).Other solutions related to curricular aspects include course and timetable flexibility (Yepes et al., 2007); doctoral education for teachers and training in new teaching methods (Pineda-Báez et al., 2011); university welfare services including physical and mental health services (Becart, 2017); incorporation of new teaching and learning methodologies like gamification (Zabala-Vargas et al., 2021); and onboarding processes (Salazar et al., 2011).No solutions from private companies, social movements, or the government were reported.
Concerning individual causes, physical or mental health is closely related to college dropout (Vera Cala et al., 2020).Factors like stress (Suárez-Montes & Díaz-Subieta, 2015a), anxiety, and depression (Caballero-Dominguez et al., 2018) have also been reported as causes.Elsewhere, in certain cases, a professional career is not important for a person's life project (Valdés-Henao, 2018), or families can directly affect a student's beliefs, expectations, and aspirations, with the students of single-parent families the most affected (Rueda Ramírez et al., 2020).Finally, an incorrect choice of undergraduate program by the student can lead to desertion because their expectations are not fulfilled (Rojas et al., 2008).

Conclusions
• Universities approach the dropout problem individually: In most of the documents, the student sample used in the experiments come from a single university (each university studies its own students or even students from a specific program); the exclusion of samples from multiple universities could lead to errors in the results and limit research findings of research.A lack of a more global point of view (local, regional, country) could result in bias or an incomplete understanding of dropout.• Lack of description of solution strategies and impact measurements: The information related to some of the reported solutions is not sufficiently developed to enable a better classification, or they lacked information on implementation, results, or impact.This makes it difficult to make progress with designing new strategies based on previous studies.• Low integration across stakeholders: A common factor of the proposed solutions is the lack of interaction between researchers and institutions.In general, the articles reviewed focused on the institutions the researchers belonged to.A lack of research with samples from multiple institutions that could provide a more global perspective of the problem of undergraduate dropout is evident.It should be mention that there is evidence of a disconnect between high schools and universities.

Disclosure statement
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Figure 1 .
Figure 1.Distribution of publications on university dropout in Colombia per year.

Figure 2 .
Figure 2. Reported causes of university dropout in Colombia from 2000-2021.The numbers to the right of each category indicate the number of papers in each category.

Figure 3 .
Figure 3. Reported solutions for university dropout in Colombia between 2000 and 2021 (responsibility perspective.

Table 1 .
Papers and keywords.