Effectiveness of real-time classroom interactive competition on academic performance: a systematic review and meta-analysis

In recent years, different tools have been introduced into the educational landscape to promote active participation and interaction between students and teachers through personal response systems. The evolution of this methodology has allowed students to participate in real-time by answering questions posed. Previous reviews on the effectiveness of real-time classroom interactive competition (RCIC) on academic performance have been performed; however, this research was based only on Kahoot, without considering other RCIC tools or programs. In addition, the RCIC effectiveness at different educational levels and its effect according to the duration of the intervention has not been meta-analytically analyzed until to date. The aim of this meta-analysis was to analyze the RCIC effectiveness in improving academic performance. A search focused on studies from the educational field published from 2010 until September 2022 was performed. Experimental studies with objective and valid data (scores based on tests or exams) were included. From a total of 397 studies considered potentially eligible, 23 studies met the inclusion criteria. The sample was n = 1,877 for the experimental group and n = 1,765 for the control group with an academic improvement in favor to experimental group (MD 7.34; CI [5.31–9.43]; p < 0.001). There was also significant improvement in academic performance when analyzing different educational levels and different tools. In addition, both short-term interventions (two weeks or less in duration) and long-term (from two weeks to one year in duration) were effective. Therefore, RCIC interventions seem to be an effective strategy to improve academic performance.


INTRODUCTION
In recent decades, electronic devices, such as smartphones and tablets, are increasingly present in the market, as a sign of the great technological development of society, which, in addition to allowing a permanent connection to the internet, stimulate the participation and interest of students at all educational stages. One of the current challenges in teaching is optimizing class time and increasing student motivation towards the teaching-learning process. The proper use of this type of device makes them a methodological tool that could be useful for this purpose, complementing the traditional classes (Heflin, Shewmaker & Nguyen, 2017). The traditional class teaching method seems to cause a loss of concentration and passivity in the students (Miller, 2014). To overcome this problem, teachers should try to increase the level of participation of students during the development of the class, thus preventing loss of concentration in long duration sessions.
There are different strategies to making learning more interactive, such as: using audience responses, videos, audiovisual aids, written material. Moreover, strategies such as: dividing the class into groups, presenting cases where students can work, questioning the audience, organize guest talks, debates, using simulations and role plays (Snell, 1999). Nevertheless, at the beginning of the 21st century, student response systems, initially called ''clickers'', burst onto the educational scene. These are systems equipped with an electronic command by means of which questions and questions are answered in real time. These systems made it possible to ask collective questions to a group-class and collect the individual responses issued by the students and show the statistical graphics of the results (Caldwell, 2007). Nowadays, these systems have evolved, being available in various formats with intuitive interfaces for any portable electronic devices, regardless of the operating system used. As a result of this evolution, and the mobile phone penetration in the population (Langford et al., 2019), many of the inconveniences that the first systems presented have been solved: the use of special equipment, the time invested in training by teachers and students, and the economic cost involved in purchasing the number of models necessary for use in a class. The game-based learning is an advancement in learning technology, it could motivate and involve players in such a way that they learn without being aware of it (Gee, 2003). Serious games can be beneficial for classroom dynamics, motivation and academic performance (Sharples, 2000). Several student response systems have introduced game features to increase student participation, such as the Space Race games in Socrative (Dervan, 2014) and Quizlet (Chien, 2015).
With the development of real-time response system applications for smartphones, there is now the ability to quickly assess student knowledge. Real-time classroom interactive competition (RCIC) is carried out with the use of platforms web-based student immediate response systems, which seems to help students learn through play, through an environment of communication and interaction that encourages feedback and reinforcement between teachers and students, and between students themselves with their peers. This initiative can incorporate a fun learning method for students (Chin, 2014). These types of platforms are generally known as mobile learning (ML). ML is widely used in the educational context, the students have access to the questions posed by the teacher through the ''room or quiz number''. The administrator can generate or import multiple choice, short answer questions or true/false. These ML programs provide statistics on the answers given that can be evaluated in real time. Both the version for computer and mobile devices (smartphones and tablets) are free (although there are paid versions) and the only requirement for its use is a previous registration by the teacher. ML programs can facilitate the construction of knowledge, problem solving and the development of diverse skills and abilities in an autonomous and ubiquitous way, thanks to the mediation of portable devices (mobile phone, laptop, tablet, etc.) and internet (Georgiev, Georgieva & Smrikarov, 2004). Normally, the ML interface is designed with many interactive features (including music in some cases), where students via mobile devices can join games and answer questions, and view the answers to their choices. The applications software has been designed to involve users in activities motivated by their personal preferences (Licorish et al., 2018). This low-cost technology has been recognized as being of great value in enhancing the educational experience for teachers and students alike (Collins, 2007;Guerrero et al., 2016;Rodriguez, Ortiz & Aguilar, 2018). However, some studies consider that excessive contact with technology or use of ML could present several issues in class, such as a negative impact on teaching (Tondeur et al., 2017) and the students (Chen, Pedersen & Murphy, 2012), and there is reluctance among some teachers to use mobile devices regularly in class, claiming that they impair human communication and socialization and are potential distractors (O'Bannon & Thomas, 2015). Moreover, the use of ML could present problems related to network connectivity (Wang & Tahir, 2020). However, despite these drawbacks, this traditionalist point of view is challenged by the results of recent studies that highlight the benefit that mobile technology can bring in different aspects of the teaching-learning process (Wang & Tahir, 2020).
Previously, reviews have been carried out on gamification based on ML in the educational field (Crompton & Burke, 2018;Baptista & Oliveira, 2019;Huang et al., 2020;Noroozi, Dehghanzadeh & Talaee, 2020;Sailer & Homner, 2020;Wang & Tahir, 2020;Zhang & Yu, 2021). In relation with the use of RCIC tools or programs, a recent literature review concluded that Kahoot can positively affect learning outcomes from the perspective of interaction (Zhang & Yu, 2021). This study concluded that future research could be conducted to determine the effectiveness of Kahoot in different contexts such as primary education, higher education, and occupational training. Another review of the literature that focused solely on Kahoot concluded that it can have a positive effect on learning performance, classroom dynamics, student and teacher attitudes, and student anxiety. However, the researchers noted that there are also studies in which Kahoot has little or no effect (Wang & Tahir, 2020). Therefore, objective and specific reviews seem to be needed. In addition to extending the research to other programs or applications besides Kahoot. Many of these reviews cited above have not focused exclusively on the promotion of academic performance (getting required marks or academic scores according to the standards set) through the RCIC. Therefore, there is currently no consensus on RCIC contribution to improving academic performance, despite its recent popularity.
This article provides a systematic review and meta-analysis of the literature on the effectiveness of RCIC through mobile applications or web-based student immediate response systems platforms on academic performance. Due to the fact that there are different types or classifications of RCIC without a consensus on their effectiveness in improving academic performance, this article has the aim of shedding light and establish a consensus on the benefits of RCIC in the academic performance of students, and that these considerations can be taken into account by education professionals. Therefore, the present systematic review and meta-analysis can be of great benefit to education professionals (different academic levels) who wish to apply new methods based on the information and communication technologies and ML to improve the academic performance of their students, obtaining after reading a broad view of the effectiveness of these RCIC tools as a learning strategy. In addition, it could provide a base for researchers or developers who wish to carry out new work or create new tools or technological applications, since it seems that the applications designed by educators or researchers themselves could be effective in improving academic performance.
To ensure broad coverage, inclusion criteria based on the PICO (S) question (Moher et al., 2009) were used, performing a search in the following databases: Wiley InterScience, Science Direct, PubMed, Web of Science and ERIC. The search was based on the following terms: 'academic performance', 'technological learning', 'classroom dynamics' and 'gamification'. These aspects are detailed in the manuscript.
Therefore, the present meta-analysis aimed to contribute in terms of its design to resolve the current gaps in the literature. It examined studies carried out in all academic cycles with different intervention duration (short-time and long-time). The studies were based on comparing RCIC interventions to improve academic performance with respect to control groups (CG) that did not receive a RCIC intervention, and academic performance was objectively measured. The study also aimed to contribute on which RCIC tools are the most effective for improving student performance, including popular known tools, such as Kahoot to self-designed tools by the researchers themselves, with the aim of shedding light and establish a consensus on the benefits of RCIC in the academic performance of students, and that these considerations can be taken into account by education professionals.

MATERIALS & METHODS
A systematic review and meta-analysis was carried out ac-cording to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) (Moher et al., 2009;Page et al., 2021) (Supplementary Material S1). The review protocol was registered at the PROSPERO (International Prospective Register of Systematic Reviews), with the registration number CRD42020212646.

Inclusion criteria of selected studies
The inclusion criteria of this systematic review and meta-analysis were established according to the PICO (S) outline: P (population): Students of all levels and types of education (including children and adults); I (intervention) and C (comparison): Experimental group (EG) that received intervention or evaluation with RCIC versus a CG; O (outcome): Improvement in academic performance; (S) (type of study): experimental studies. Exclusion criteria were: (i) the article is not accessible in any way through university services or memberships; (ii) studies that were not reported with measurable and objective data (numerical values or scores based on a test or exam) on academic performance; (iii) articles in other languages apart from English.

Protocol for electronic searching
The databases Wiley InterScience, Science Direct, PubMed, Web of Science and ERIC were used to execute the search query between 2010 and September 2022. These databases were considered main sources to retrieve highly quality related papers. The search query was composed in English. ML based in RCIC was queried using terms such as 'academic performance', 'technological learning', 'classroom dynamics' and 'gamification'.

Study selection and data collection
The search and analysis of the studies was conducted by two researchers (J.M.J-C. and J.B-P.). The reference manager RefWorks was used to code the articles found (Hendrix, 2004), the discrepancies regarding the interpretation of the extracted data were discussed by both investigators. Moreover, the articles were filtered manually using the inclusion criteria, an internal code for researchers was associated with each selected article.
Next, an individual analysis was conducted for each study separately (Petticrew & Roberts, 2006). The following information about each study was extracted: (a) general information (country, year of publication), (b) method (participants, educational level, measure of educational performance), (c) kind of intervention, and (d) general results and conclusions.

Risk of bias in individual studies
According to the Cochrane Collaboration, the risk of bias was analyzed for each study (Sterne et al., 2019). Seven domains (sequence generation, allocation concealment, blinding of participants and staff, blinding of outcome assessment, incomplete outcome data, selective outcome reporting, and other aspects considered) were assessed, with a rating of high, low or unclear risk of bias.

Statistical analysis
To perform this meta-analysis, the effect of RCIC educational interventions was examined on improvement of academic performance, comparing EG with a CG. Data were obtained using the sample size, the mean and standard deviation of the evaluation data (numerical values) presented after the intervention or a pos t -test. The numerical evaluation values presented in the selected studies were subjected to a multiplier to present the data over a value of 0-100 in order to standardize the results for subsequent analysis and reduce heterogeneity.
The results of the present meta-analysis were presented as forest plots with mean differences (MD) and 95% confidence interval (CI). Heterogeneity was also presented, which was calculated by measuring its scope by the I 2 index. To analyze publication bias, Egger test (Egger et al., 1997) was used. A sensitivity analysis was also performed to assess that the effect size did not significantly vary after eliminating each study from the analysis. The authors examined the value of P for this statistic, warning of the presence of heterogeneity when P < 0.05, which compromises the validity of the pooled estimates (Takkouche, Cadarso-Suárez & Spiegelman, 1999). Moreover, the I 2 index considered a low heterogeneity (0% to 40%); moderate (30%-60%); considerable (50%-90%); or substantial (75%-100%) (Shamseer et al., 2015). Given that heterogeneity is presumed in this study, a random-effects model method was used to measure the effect of the included studies (Berlin et al., 1989;Lipsey & Wilson, 2001;Shadish & Haddock, 1994). Moreover, to identify the possible sources of variance in overall effect and heterogeneity, as two categorical moderators (school level and program type) were categorized in selected studies, subgroup analyses were applied. Moreover, a subgroup analyses were applied to difference the effectiveness between short-term (≤ two weeks) and long-term (≥ two weeks) interventions.
To carry out the meta-analysis, the Review Manager program was used (RevMan, computer program) version 5.4.1. was used (Cochrane, 2008). There was considered a statistical significance in all analyses with a value of P < 0.05. Results are shown in MD followed by CI.

Studies selected
A total of 5,846 articles were identified from the included databases. 277 duplicates were eliminated, thus leaving a result after the search of 5,569 publications. According to the title and abstract, 397 publications that were potentially relevant were identified, from which the full text was extracted. After applying the inclusion and exclusion criteria, 347 studies were excluded. Therefore, 23 articles were finally selected for the present meta-analysis ( Fig. 1).
Information on experimental results and academic performance evaluation between CG and GA in each of the selected studies is provided in Table 1.

Effects of real-time classroom interactive competition interventions on academic performance
Academic performance was evaluated in the twenty-three studies selected. The total sample (students) of the meta-analysis was 3,642 (n = 1,877 for the EG and n = 1,765 for the CG). The meta-analysis showed an academic performance improvement in favor to EG (MD 7.37; CI [5.31-9.43]; P < 0.001; I 2 = 91%) (Fig. 2). A greater learning was observed in students who used RCIC in their courses compared to those who did not.
The weight of the studies was distributed, with none of them standing out substantially above the other. The lower limit was found at 1.4% (Asmali, 2018) and the upper one at 6.6% (Wichadee & Pattanapichet, 2018), oscillating the weight of the other studies between these values. Only one study (Sanchez, Langer & Kaur, 2020) showed an effect in favor to CG (MD −1.35; CI [−3.9 to 1.2]).
To reduce the presence of heterogeneity (I 2 = 91%), several moderators with random effects were used to explore possible causes. In this sense, three main categories of variables

Effect of school level
The included studied were divided in subgroups according to student's level school: (i) University; (ii) Secondary school; (iii) Primary school; (iv) Other. The test for subgroups differences showed a valor of I 2 = 57.1% displayed a moderate heterogeneity. A greater weight (72.7%) was observed for university level subgroup, followed by secondary school level (12.7%), subgroup described as ''other'' (8.1%), and finally primary school level (6.5%) composed only by one study (Pozo-Rico & Sandoval, 2019) (Fig. 3).

Effect of mobile learning program based on real-time interactive competition
Another subgroup classification was realized, in this case the studied were divided in subgroups according to program used: (i) Kahoot; (ii) Other; (iii) Designed by educators or researchers themselves. The test for subgroups differences showed a valor of I 2 = 45.8% displayed a moderate heterogeneity. A greater weight (49.3%) was observed for Kahoot subgroup, followed by other programs (subgroup described as ''other'') (39.7%), and finally ML designed by researchers (11%). In this subgroup classification, heterogeneity was only substantially reduced in the group ''designed by researchers'', where was observed a I 2 index = 0%. (Fig. 4).

DISCUSSION
The present meta-analysis based on quantitative studies measured by objective test scores comparing CG vs EG, provided evidence that academic RCIC interventions carried out were effective to improve student academic performance. Based on 23 RCIC interventions and a total of 3,642 students, our study found that students in the EG scored an average of 7.37% higher on exams than those in the CG. The EG obtained a higher score (8.4%) when the interventions were carried out at the university, compared to the rest of academic cycles, a higher score (10.02%) when the interventions were carried out with ML tools designed by the researchers or educators themselves, and with better results (8.42%) when these were carried in a short term of less than weeks. Despite the popularity of ML based on RCIC, there is no consensus on whether it contributes to improving academic performance or not. According to a recent metaanalysis in the pre-publication phase based on RCIC interventions with Kahoot (Yu, 2021), has been shown to significantly improve the academic performance of students. The present meta-analysis examined whether the effect of RCIC with different tools, beyond that Kahoot, improved the academic performance in the students.
This meta-analysis found that the RCIC interventions did not differ across grade levels on academic performance. This finding shows that this type of intervention could be a design applicable from elementary school to university level. In this context, three studies are at the secondary education level and only one study is at primary education, so more studies are required to clearly establish the effect of gamification to improve academic performance in primary and secondary schools. Despite the different subgroups, the results of our meta-analysis of subgroups by academic levels showed a greater effect in improving academic performance for gamification studies carried out with university students, in addition to being the most numerous in our selection of studies, and if there is a wide literature on the matter, in different areas or subjects (Andreu, 2020;Domínguez & Bezanilla, 2019). In addition to academic performance, positive results have been found in the motivation of university students with the implementation of gamification experiences in the classroom (Andreu, 2020).
Based on the inclusion criteria of our study, other interventions were found that did not belong to these academic levels, however they were based on education and training for the performance of their employment. This subgroup classified as ''others'' was made up of the Lai et al. (2020) study, which relied on gamification in ultrasound training for beginning physicians, and by Linganna et al. (2020), also carried out with young doctors based on an asynchronous educational application based on smartphones for the knowledge of transesophageal echocardiography compared with traditional modalities alone. These ''others'' interventions did not have a significant effect.
Different programs or tools were described to carry out RCIC interventions, although the most common one so far in the literature, as well as in the studies included in our work, has been Kahoot. According to the results of our meta-analysis, the RCIC tools self-designed by the researchers themselves were more effective with a greater difference in means in the score of the EG group on the CG, followed by Kahoot, and the studies that included other programs in their interventions, despite this all tools had a significant improvement in improving academic performance. The success of self-designed tools compared to the others is surely due to the fact that the interface and the gamification environment were completely adapted to the specific needs of the students. Among the most widely accepted mechanics in the reviewed literature, the scoring systems, fictitious or real gifts or rewards, ranking, achievements, avatars, stamps (badges), unlocks of new skills, levels, missions or challenges stand out team or alone.
Study interventions selected a duration ranging from less than two weeks to 1 academic year. Both short-term and long-term interventions showed a significant effect. The fact that short-term interventions tend to have a greater MD score (EG vs CG) compared to long-term, 8.42 vs 7.03, respectively, raises the problem of the 'novelty effect'.
An important aspect to consider is that the novelty effect seems to be a conditioning factor in the success of short-term interventions, because the participants are subjected to something new to them (Clark, 1983). In this regard, gamification used in the shortterm could lead to high participation by students, and may provide a greater academic achievement, therefore, the novelty effect could be a confusing variable on the student academic achievement in gamification interventions (Jeno et al., 2019). In relation to long-term intervention, an aspect to take into account related to the Cowan study (Cowan, 2010) that argued that long-term interventions can lead to superficial understanding and learning, because the ability to memorizing of each student is highly variable. Lai et al. (2020) study, selected in our meta-analysis, carried out an evaluation after the intervention and at two months post-intervention, without finding differences in the evaluation after the test with those of the 2 months after the training, which suggests that there was good retention of knowledge and skills.
Students, especially university students, regularly use their smartphones in the educational center on many occasions as a communication and entertainment tool. This can be detrimental to student learning in class. However, students feel comfortable using smartphones as a pedagogical tool.
Another important aspect to consider is that the integration of ML elements does not guarantee its effectiveness. Efficacy depends on different factors such as how to implement gamification (Hamari, Koivisto & Sarsa, 2014), as well as learner characteristics and student's context (Buckley & Doyle, 2017). It requires significant effort and good planning in order to match gaming elements with instructional objectives. The objectives of gamification must be clear, making sure that the proposals respond to the needs raised. Students must understand the scoring system. The profile and interests of the student body, the class size, and the teaching and learning approaches that are being implemented to accommodate different types of students from different disciplines should be taken into account (Andreu, 2020).
According to the results of this meta-analysis, the use of RCIC can help increase academic performance and the effectiveness of learning in the classroom. Based on previous research, the use of certain RCIC applications, such as Kahoot it is very effective and helps the learning process in the classroom The Kahoot app can be used effectively for gamification lessons. The application of gamification with this medium can have an impact on students that makes them more ambitious and motivated to learn (Lin, Ganapathy & Kaur, 2018). The use of the RCIC in the learning process can enrich the quality of student learning in the classroom, with the greatest influences reporting on class dynamics, participation, motivation and the improvement of learning experiences (Bicen & Kocakoyun, 2018). Another study (Andreu, 2020) has suggested that the learning styles and the dynamics and mechanics used during gamification may influence differently depending on the type of students. One of the keys to integrating gamification is the context of learning and play on which it is based. So, gamification appears to have a positive effect on learning compared to traditional approaches.
The present study has some limitations. First, the level of heterogeneity is high, which has been controlled with different analysis of categorical possible moderators. In addition, in terms of subgroup analysis, there were subgroups with a reduced number of articles per category, this was due to the limited literature that met the inclusion criteria, for example, in interventions carried out in the primary cycle education or tools of self-designed ML based on RCIC, so the results in the present meta-analysis should be interpreted with caution.
Second, the studies come from different countries, with their own education system characteristics and this factor should be taken into account. Future research should evaluate, if the combination of different ML based on RCIC tools can increase the improvement in academic performance. In this sense, there is a gap in the literature

CONCLUSIONS
This meta-analysis supports the notion that RCIC interventions seem to be successful in academic performance. There was a significant improvement in academic performance in the different educational cycles, as well as with the use of different tools. In addition, both short-term interventions and long-term were found to be effective.
Education professionals and teachers can consider the results obtained in this metaanalysis as practical application. Firstly, they should know that interventions based on RCIC seem to be effective in improving students' academic performance. Secondly, they should consider that the interventions have proven to be effective in all educational cycles, especially in university students, with the exception of instruction on young workers related to specific topics of medical sciences. Third, the tools self-designed by the educators seem to be effective for improving academic performance. Self-designed ML based on RCIC allows adapting to the specific educational needs of students, both in student motivation and educational learning. Finally, both short-term (less than weeks) and long-term (up to one academic year) interventions seem to be effective for improving academic performance, although considering that, in short-term interventions, the students have obtained greater success on the final score.
Finally, we believe that this study contributes to an objective meta-analytic understanding of the RCIC effectiveness to improve student academic performance. The consolidation of communities is beneficial to student retention and academic engagement. Consistent technological advancements have facilitated the design of innovative tools and approaches with potential applications in education. For instance, virtual reality devices, augmented reality, and massively multiplayer online systems represent such tools. The integration of these technologies with relevant learning theories incorporating an RCIC system within the classroom environment can promote students' educational outcomes and academic performance across diverse educational levels, surpassing current methods, as the tools used by the studies included in our meta-analysis. Consequently, it is recommended that future research investigate the feasibility of these new technological tools within an RCIC based system to establish their efficacy in enhancing academic performance. Additionally, future research could expand to examine more components such as cognitive loads, satisfaction, or anxiety with the use of RCIC.