Effectiveness of Mobile Serious Games in Nursing Education: A Systematic Review and Meta-Analysis of Randomized Controlled Trials

Search strategy and eligibility criteria

Studies were selected based on the following inclusion criteria: 1) Population: nursing students; 2) Intervention: mobile serious games; 3) Comparison: a comparison group receiving standard intervention or no intervention; 4) Outcome: learning outcomes, namely knowledge and skills; 5) Study type: the study design was randomized controlled trials, considered the gold standard for assessing an intervention’s effectiveness.¹⁴ As for the exclusion criteria, studies were excluded if they were published before 2016, were not published in English, and computer-based serious games were excluded to ensure a specific focus on mobile platforms, which provide unique affordances such as portability and ubiquitous access that may differently influence learning outcomes.

The literature search was conducted by five independent reviewers across nine electronic databases: Scopus, ScienceDirect, IEEE, SAGE Journals, PubMed, Taylor & Francis, EBSCO, CINAHL, and the Cochrane Library (CENTRAL) during the period from January 1, 2026, to February 15, 2026. The main search strategy used a combination of keywords and Boolean operators (AND/OR). The keywords used were ((“nursing student*” OR “student nurse*” OR “pupil nurse*”)) AND ((“serious game*” OR “game-based learning” OR “digital game*” OR “simulation game*” OR “mobile game”)). All identified records were imported into Rayyan AI, and duplicate records were removed prior to screening. The inter-rater reliability was assessed using Cohen’s kappa coefficient (κ = 0.90), indicating almost perfect agreement. Any disagreements were resolved through discussion until consensus was reached.

Data extraction and management

Data extraction was conducted independently by five researchers using a previously developed extraction form. The information collected included: authors and year of publication, study objectives, sample size, participant characteristics, type of intervention (game name), game format, total playing time, platforms, control group, time of measurement, study outcomes, and main findings. The extraction results were compared between the researchers to ensure the accuracy and completeness of the data.

Data synthesis

The meta-analysis was conducted using Comprehensive Meta-Analysis (CMA) software version 2 (Biostat, Englewood, NJ). Data from similar outcomes were pooled to obtain a combined effect estimate. Heterogeneity across studies was evaluated using the I2 statistic; I2 values>50% with p-values <0.05 indicated substantial heterogeneity. The effect estimate was calculated using a random-effects model with the standardized mean difference (SMD) as the effect size and a 95% confidence interval (CI). Subgroup analysis was conducted to identify factors affecting variation in results, with groups defined by academic year, platform, total playing time, and game format.

Sensitivity analysis was conducted using a leave-one-out approach to assess the robustness of the pooled results. Publication bias was assessed using funnel plot visualization and Egger’s regression test. The synthesis of results involved pooling effects, interpreting findings, and presenting results in tables and forest plot graphs. The entire analysis was conducted by two researchers and reviewed by two additional researchers to ensure consistency and accuracy of the results.

Study risk of bias assessment

The methodological quality assessment was conducted to identify potential risks of bias in all included studies. Two independent researchers assessed each article using the Cochrane Risk of Bias 2 (RoB 2) tool for randomized controlled trials (RCTs) (Sterne et al., 2019). The assessment covered five domains: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. Each domain was classified as low risk of bias, some concerns, or high risk of bias, and the overall judgment was determined based on the results across all domains. Any discrepancies in the assessments between the researchers were resolved through discussion until consensus was reached. The results of the risk of bias assessment are presented in Table 1 (Study Risk of Bias Assessment Using ROB 2 Tool).

Table 1. Study risk of bias assessment using ROB 2 Tool.

No	Study ID	D1	D2	D3	D4	D5	Overall	Judgement:
1	Bayram & Caliskan (2019)								Low risk
2	Demirel & Kaya (2026)								Some concerns
3	Farsi et al. (2021)								High risk
4	Fijačko et al. (2024)
5	Genç et al. (2025)							Domains:
6	Gu et al. (2022)							D1	Randomisation process
7	HajiAbad et al. (2026)							D2	Deviations from the intended interventions
8	Kulakaç & Çilingir (2024)							D3	Missing outcome data
9	Nasirzade et al. (2024)							D4	Measurement of the outcome
10	Tang et al. (2023)							D5	Selection of the reported result

Selection of studies

A total of 544 studies were identified from nine databases. After removing 121 duplicates, 423 studies were screened based on title and abstract, of which 386 studies were excluded. A total of 37 articles were assessed through full-text review, and 27 articles were excluded for not meeting the inclusion criteria. Ultimately, 10 studies were included in the systematic review and met the criteria to proceed to the meta-analysis. The completed PRISMA flow diagram is available as extended data in Zenodo.¹³

Study characteristics

A total of 10 RCT studies published between 2019 and 2026 involved nursing students at various academic levels (from first-year to clinical practice) as participants. The interventions used consisted of serious games with varying playing durations, ranging from short sessions of approximately ±2 minutes to repeated use over 7 days up to 8 weeks. Outcome measurements were conducted at various time points, generally immediately after the intervention, and several studies performed follow-up evaluations up to 2–4 weeks to assess learning retention. The intervention platforms were predominantly smartphones (Android/iOS) and web-based multi-device applications. The control groups generally received conventional learning, such as lectures, demonstrations, or no additional intervention. The intervention topics encompassed various nursing clinical skills, with the primary outcomes comprising knowledge and clinical skills, along with additional variables such as anxiety, satisfaction, and learning curve. The complete study characteristics are presented in Table 2: Study Characteristics.

Effects of mobile serious games intervention

The effects of the use of mobile serious games on the learning outcomes of nursing students are presented in Figures 1 and 2. Compared with conventional learning methods, mobile serious games showed statistically significant improvements in both knowledge and clinical skills. The meta-analysis results showed a moderate effect on knowledge (SMD = 0.72 (0.34–1.1); p < 0.001) and a large effect on skills (SMD = 1.01 (0.55–1.46); p < 0.001), indicating that mobile serious games–based learning was more effective than traditional approaches in improving cognitive and psychomotor learning outcomes in nursing education. These findings suggest that mobile serious games can provide meaningful improvements in clinical competence, particularly by enhancing students’ ability to apply knowledge and perform clinical procedures in simulated or practice-based settings. Sensitivity analysis using the leave-one-out approach showed that no single study significantly altered the pooled effect size, indicating the robustness and stability of the results.

Figure 1. Forest plot of the effect of mobile serious games on knowledge outcomes.

This figure presents the pooled standardized mean difference comparing mobile serious games with control or conventional learning approaches for knowledge outcomes among nursing students.

Figure 2. Forest plot of the effect of mobile serious games on clinical skills outcomes.

This figure presents the pooled standardized mean difference comparing mobile serious games with control or conventional learning approaches for clinical skills outcomes among nursing students.

Test of heterogeneity and subgroup analysis

Significant heterogeneity was observed for knowledge outcomes (I² = 75.21%) and skills outcomes (I² = 85.22%) (p < .05) (Figures 1 and 2). These results indicate substantial to considerable variability among the included studies. This heterogeneity may be explained by students’ academic level, intervention platform, total usage time (total play time), and game format ( Table 3: Subgroup Analysis). Based on academic level, there were significant differences in effect on knowledge outcomes (p < 0.001) and skills outcomes (p = 0.022). Final-year students (years 3–4) showed a greater effect on knowledge (SMD = 1.353) compared with early-year students (years 1–2; SMD = 0.378). A similar pattern was also observed for skills, with final-year students showing a greater effect (SMD = 1.482) than early-year students (SMD = 1.139).

Based on the platform, no significant difference was found in knowledge outcomes (p = 0.439); however, a significant difference was found in skills outcomes (p = 0.044). Multi-platform–based interventions showed a greater effect on skills (SMD = 1.464) compared with interventions that used mobile phones only (SMD = 0.550). Based on total usage time, there was a significant difference in knowledge outcomes (p < 0.001), but not in skills outcomes (p = 0.782). The effect on knowledge increased with the duration of use, with the greatest effect observed in interventions lasting more than three weeks (SMD = 1.434), followed by two weeks (SMD = 0.604) and one week (SMD = 0.248). Based on game format, there was a significant difference in skills outcomes (p = 0.012), but not in knowledge outcomes (p = 0.925), with the greatest effect observed in scenario-based games (SMD = 1.861).

Publication bias

Funnel plot inspection revealed an asymmetrical distribution, with most studies clustered on the positive effect side and a relative absence of studies on the negative side, suggesting potential publication bias. The findings are limited by the small number of included studies and the variability in study designs, which may affect the generalizability of the results.

Limitations

Several limitations need to be considered when applying the results of our study. There was the limited number of included randomized controlled trials (n = 10), relatively high heterogeneity across studies (I² > 75%), reflecting variations in intervention design, duration of use, participant characteristics, technology platforms, as well as outcome measurement methods. Although subgroup and sensitivity analyses were conducted, the exact sources of variability could not be fully identified. Additionally, the relatively small number of included RCTs limits statistical power, particularly for subgroup analyses, which may affect the stability, reliability of subgroup findings, and generalizability of the results. Variations in outcome measures, intervention fidelity, and reporting quality may further introduce bias and limit comparability, while the inability to perform advanced analyses such as meta-regression restricts deeper exploration of moderating factors. Moreover, most studies reported only short-term outcomes, limiting conclusions about long-term knowledge retention and skill sustainability. Finally, although publication bias was assessed using funnel plots, the possibility of selective reporting cannot be entirely excluded.