Study selection
The selection process for the studies included in the final review is depicted in the PRISMA 2020 flow diagram [21]. Initially, 1,237 studies were retrieved from the six databases, from which 401 duplicates were removed. Subsequent screening of the titles and abstracts led to the exclusion of 781 studies. After reviewing the full text of the remaining 55 articles, 44 were excluded, leaving 11 studies for the final review. The final selected studies are shown in Appendix A, and the study selection process is illustrated in Figure 1.
Characteristics of the included studies
The general characteristics of 11 studies analyzed in this review are shown in Table 1. The publication years ranged from 2017 to 2024, with eight of the studies [A3-A11] being published after 2020, indicating that the majority were published within the last 5 years. The sample sizes were as follows: five studies [A2, A3, A5, A8, A11] included 40-100 participants, five studies [A1, A4, A6, A7, A10] involved 100-200 participants, and one study [A9] included more than 200 participants, resulting in a total of 1,358 participants.
Regarding the characteristics of the interventions, seven studies [A1, A2, A4, A6, A8, A10, A11] involved games covering specific nursing skills (blood transfusion, tracheostomy care, basic and advanced life support, venous catheter flushing/locking, neonatal resuscitation training, venous blood specimen collection, and stoma care/colostomy irrigation), while four studies [A3, A5, A7, A9] focused on broad course matter (pathophysiology, health assessment, endocrine course, fundamentals of nursing, and diet/nutrition). Six studies [A2, A4, A6, A9, A10, A11] utilized mobile applications as gaming platforms, whereas five [A1, A3, A5, A7, A8] used websites. The most common intervention period was 1 week, as reported in five studies [A2, A3, A5, A6, A10], and three studies [A7, A9, A11] involved intervention periods longer than 2 weeks. All the studies measured the effects before and after the intervention was implemented. In five studies [A1, A3, A4, A7, A11], additional follow-up surveys were conducted at least 2 weeks and up to 7 weeks later.
The most frequent control group interventions were lectures and laboratory classes (four studies [A1, A2, A5, A11]), followed by video training (three studies [A6, A8, A10]), lectures (two studies [A4, A9]), case discussion [A3], and flipped classroom [A7].
Risk of bias
The quality appraisal results for the 11 selected studies are shown in Figure 2. All included studies were evaluated using RoB 2.0, which assesses bias across five domains: biases arising from the randomization process, biases due to deviations from intended interventions, biases due to missing outcome data, biases in measurement of the outcome, and biases in selection of the reported result. The evaluation revealed that all the studies exhibited a low risk of bias across these domains. Consequently, all the studies were included in the analysis.
Effects of digital serious games
The outcomes of the studies analyzed according to the Kirkpatrick evaluation model are presented in Table 2. The analyzed studies predominantly assessed Levels 1 and 2—Reaction and Learning, respectively, whereas Levels 3 and 4—Behavior and Results—were not evaluated. The assessments included the game experience [A4], usability [A5], and satisfaction [A8]. In the learning domain, knowledge [A1, A2, A3, A4, A7, A8, A9, A11] and performance [A1, A2, A3, A6, A7, A8, A10, A11] were the most frequently evaluated; each outcome was covered in eight studies. In addition, four studies evaluated confidence [A1, A5, A7, A8], whereas anxiety [A5], intensity of preparation [A7], motivation [A7], and perception [A1] were each examined in a single study.
A meta-analysis was conducted using eight studies selected, which provided the mean, standard deviation, and sample size for their outcome variables. We requested original data from the authors of the three studies that did not provide the mean or standard deviation; however, we were unable to obtain this information. Despite the use of diverse outcome variables and measurement instruments across these studies, separate meta-analyses were conducted for common variables, such as knowledge, performance, and confidence, with the findings displayed in forest plots (Figure 3).
A meta-analysis of six studies examining the effects of digital serious games on nursing students’ knowledge revealed a significant large effect size with high heterogeneity (Hedges’ g = 0.74, 95% CI = [0.27, 1.22], p = .002, I² = 90.51%). A meta-analysis of six studies examining the effects of digital serious games on confidence among nursing students revealed a significant large effect size with high heterogeneity (Hedges’ g = 0.73, 95% CI = [0.23, 1.24], p = .005, I² = 82.71%). A meta-analysis of four studies examining the effects of digital serious games on nursing students’ performance revealed a significant medium effect size with moderate heterogeneity (Hedges’ g = 0.49, 95% CI = [0.17, 0.80], p = .003, I² = 56.60%).
On excluding each included study individually from the sensitivity analyses of knowledge, confidence, and performance among nursing students after a digital serious game intervention, the pooled mean scores remained largely unchanged. This finding suggests that the results of the meta-analysis are relatively stable (Appendix B). Additionally, a funnel plot was used to evaluate publication bias. The funnel plot indicated asymmetry in the knowledge, performance, and confidence variables. However, when fewer than ten studies were analyzed, confirming the existence of publication bias was challenging (Appendix C).
Subgroup analyses
Subgroup analyses were conducted based on the duration of the digital serious game intervention (Figure 4). A meta-analysis of three studies investigating the effects of digital serious game interventions spanning ≥2 weeks on knowledge demonstrated a significant medium effect size with low heterogeneity (Hedges’ g = 0.53, 95% CI = [0.32, 0.74], p<.001, I² = 25.41%). Conversely, a meta-analysis of three studies examining the effects of digital serious game interventions spanning <2 weeks on knowledge showed a non-significant effect size with high heterogeneity (p = .128, I² = 95.59%).