Globally, there is an aging population, that is, people are living longer, largely because of improved health care services, greater understanding of nutrition, better standard of living, and advanced biomedical research [1]. The World Health Organization estimates that the percentage of the world’s population of people aged >60 years will almost double from 12% to 22% between 2015 and 2050 [2,3]. Moreover, it is projected that the number of people aged ≥80 years will triple between 2020 and 2050, rising to 426 million [2]. However, this dramatic increase in life expectancy has not been accompanied by a proportional improvement in quality of life for older adults [4]. Generally, increased life expectancy has led to an increased risk of age-related illnesses and disabilities.

An aging population therefore grapples with various challenges, including a substantial prevalence of cognitive impairment [5], which is a decline in cognitive abilities and functions such as memory, attention, processing speed, problem-solving, language, and learning [6]. Among the top chronic diseases affecting the cognitive abilities of older adults are mild cognitive impairment (MCI) and Alzheimer disease (AD). In 2022, the Alzheimer’s Association estimated that 12% to 18% of people aged ≥60 years are living with MCI, and 10% to 15% of individuals living with MCI develop dementia each year [7]. Moreover, MCI often progresses to more severe forms of dementia such as AD [8]. Statistics have shown that 1 out of every 9 individuals in the world aged >65 years has AD [9], and it is estimated that 14 million individuals will have AD by 2050 [7]. Thus, managing and preventing age-related cognitive impairments and functions are important public health concerns.

Learning disabilities are among the major cognitive impairments in older adults. A learning disability is a condition that affects an individual’s ability to understand or use written or spoken language, perform mathematical reasoning, concentrate, or coordinate movements [10]. Learning disabilities can generally be grouped into 2 types: verbal and nonverbal. Verbal learning disorder is a learning disability that causes difficulty with basic speaking, reading, and simple social skills [11,12]. By contrast, nonverbal learning disorder is characterized by visual and spatial complications, motor problems, and difficulty understanding nonverbal information (such as body language, facial expressions, and tone of voice) [13]. Successful aging requires the ability to learn new information to be able to accomplish complex tasks. Interestingly, some types of learning seem to be relatively unaffected by normal aging, whereas others decline dramatically, but the factors that determine the extent to which learning is affected by age have not yet been fully identified. Although learning disabilities cannot be cured, they can be ameliorated through a variety of therapeutic interventions and accommodations that can make it much easier for older adults to live with learning disabilities.

Technological advances have created opportunities for the use of computerized nonpharmacological interventions, including serious games, to improve learning functions. Serious games refer to participative electronic games that are designed for goals other than entertainment, such as therapeutic rehabilitation; education; prevention of, for example, cognitive impairments; and training. Serious games have also been used to improve several cognitive functions such as memory [14], processing speed [15], executive functions [16], language processing [17], visuospatial skills [18], and global cognition [19]. Common types of serious games are exergames and cognitive training games. Exergames include physical exercises as part of the intended gameplay. By contrast, cognitive training games are video games that aim to stimulate cognitive functions such as memory, processing speed, and attention. Serious games are an emerging field of study and have recently received increased attention from researchers and practitioners as a low-cost, low-risk, enjoyable, and effective alternative to cognitive interventions [20]. Therefore, it is envisioned that serious games will have a significant potential to improve the quality of life of the aging population by supporting their learning functions, cognitive abilities, physical activity, and mental health [21,22].

Many studies have assessed the effectiveness of serious games in promoting learning among older adults. The evidence from these studies has been synthesized in 2 systematic reviews; however, these reviews are undermined by the following limitations: (1) they included older adults without cognitive impairments [23], (2) they did not exclude pilot randomized controlled trials (RCTs) [23,24], (3) they did not appraise the quality of evidence [23,24], (4) they considered only particular types of serious games such as exergames [23] or cognitive training games [24], (5) they have been outdated because they were published in 2017 [23,24], and (6) they did not analyze the results of the studies according to the type of comparator (eg, no intervention, conventional exercises, or conventional cognitive training) [23,24]. To address the aforementioned limitations, this review aimed to assess the effectiveness of serious games on verbal and nonverbal learning among older adults with cognitive impairment. Our review focused on older adults with cognitive impairment, included only RCTs, assessed the quality of evidence, included all types of serious games, and analyzed results of studies according to the type of comparator. Our research question for this review was as follows: what is the effectiveness of serious games on verbal and nonverbal learning among older adults with cognitive impairment in comparison with that of other interventions such as conventional exercises, conventional cognitive training, and sham interventions?

Only RCTs that examined how effectively serious games can help older adults with cognitive impairment to improve their learning skills were included in this review. We specifically focused on studies that included serious games played on PCs, video game consoles (such as the Xbox and PlayStation), mobile phones, Nintendo consoles, tablet devices, portable devices, or any other sort of digital device. The primary element of intervention must be the game, which had to be played only for therapeutic purposes. Studies that combined serious games with other interventions were included, as long as the control group received the same adjacent intervention. Board games, card games, and other nondigital games, as well as serious games used for research, screening, diagnosis, or monitoring, were not considered in this review.

The population of the study had to be older persons (aged >60 years) with any type of cognitive impairment or disorder, as assessed by comparing baseline test results with established diagnostic criteria (eg, Mini-Mental State Examination). This review did not include studies on older people without cognitive impairment, health experts, or care providers. However, there were no restrictions on the sex and ethnicity of the participants.

The review focused on learning abilities as an outcome. Regarding the outcome measures, no restrictions were applied. Studies that simply examined acceptability, feasibility, satisfaction, or cognitive abilities other than learning were not included in this study. This review focuses on postintervention data (ie, outcome data collected immediately after the intervention). The review does not focus on follow-up data (ie, outcome data collected after a certain period had elapsed after the intervention ended) because follow-up periods were different among studies, and some studies did not follow up with participants.

This review included all types of RCTs (eg, parallel, cluster, crossover, and factorial), whereas pilot RCTs, observational studies, quasi-experiments, and reviews were not considered. Journal articles, conference proceedings, and dissertations were included. However, abstracts, conference posters, commentaries, proposals, and editorials were excluded. Studies published after 2010 in English were considered for this review. There were no restrictions on the country of publication, comparator, and research settings.

We used the following procedure to find relevant studies. First, all retrieved studies were originally imported into EndNote X9 (Clarivate) to identify and eliminate duplicates. Second, 2 reviewers independently examined the titles and abstracts of all retrieved studies (the first and second authors). Third and last, both reviewers separately read the full texts of the studies that were included during the previous phase. The 2 reviewers discussed any disagreements to resolve them.

Using Microsoft Excel, 2 reviewers independently extracted data from the included studies. Before extracting the data, we pilot-tested the data extraction form using 2 (18%) of the included 11 studies. Discussions between the 2 reviewers resolved all disagreements. The data extraction form used to collect data from the included studies is shown in Multimedia Appendix 3. If the published study did not include metrics such as mean, SD, and sample size, the first and corresponding authors were contacted to obtain the data.

Using the Cochrane Risk-of-Bias 2 tool, 2 reviewers (the first and second authors) independently assessed the risk of bias in the included studies [26]. The randomization procedure, deviation from the intended intervention, missing outcome data, assessment of the outcome, and selection of the reported results are the 5 areas of RCTs where this tool assesses the risk of bias [26]. Disagreements between the reviewers regarding risk-of-bias judgments were resolved through discussion.

The data obtained from the included studies were summarized using both narrative and statistical techniques. To be more precise, narrative synthesis was used to describe the study metadata, intervention features, participants, comparisons, and outcome measures using text and tables. The results of the included studies were divided into 2 groups based on the measured outcomes (ie, verbal learning and nonverbal learning). For each outcome, the results were assembled and grouped according to the comparator, which included no or sham intervention, conventional exercises, conventional cognitive training, and other serious games. Meta-analyses were conducted using Review Manager (version 5.4; The Cochrane Collaboration) when ≥2 studies from the same comparator contributed sufficient data (ie, number of participants in each intervention group, mean, and SD). The standardized mean difference (SMD; Cohen d) was used to determine the overall effect of each study because the outcome of interest (learning) was based on continuous data, and the included studies used various instruments to measure the outcome. Because of the significant clinical heterogeneity of the meta-analyzed studies in terms of serious game attributes (such as type, duration, frequency, and period), population characteristics (such as sample size, mean age, and health condition), and outcome measures (such as tools and follow-up periods), we decided to use the random effects model in the analysis.

We calculated I² and a chi-square P value to evaluate the level of heterogeneity and statistical significance of heterogeneity in the meta-analyzed studies, respectively. A chi-square P value of ≤.05 indicates heterogeneous meta-analyzed studies [27]. The degree of heterogeneity was considered insignificant, moderate, substantial, or considerable when I² varied from 0% to 40%, 30% to 60%, 50% to 90%, and 75% to 100%, respectively [27].

When a statistically significant difference between groups was found in a meta-analysis, we checked whether this difference was clinically significant. The term minimal clinically important difference (MCID) refers to the smallest change in a measured outcome that a patient would consider worthwhile and significant enough to merit a change in treatment. The MCID bounds were computed as ±0.5 times the SMD of the meta-analyzed studies.

To determine the overall quality of the evidence derived from the meta-analyses, we used the Grading of Recommendations Assessment, Development, and Evaluation approach [28], which evaluates the quality of the evidence based on 5 aspects: publication bias, indirectness, imprecision, inconsistency (ie, heterogeneity), and risk of bias [28]. Two reviewers independently evaluated the quality of the meta-analyzed data (the first and second authors). Any disagreements between the 2 reviewers were resolved by discussion. The reviewers’ interrater agreement (Cohen κ) was 0.89.

A total of 559 records were retrieved after searching the predefined databases (Figure 1). Using EndNote X9, from these 559 records, we eliminated 104 (18.6%) duplicates. The screening of the titles and abstracts of the remaining 455 records led to the elimination of 368 (80.9%) citations. After examining the full texts of the remaining 87 publications, we excluded 76 (87%) for various reasons (Figure 1). It is worth mentioning that interrater agreements (Cohen κ) for screening titles and abstracts and reading full texts were 0.93 and 0.95, respectively. Backward and forward reference list checking revealed no further studies. Therefore, of the 559 studies identified initially, we included 11 (2%) in this review [29-39]. The meta-analysis included 8 (73%) of these 11 studies [29-34,38,39].

The selected studies were published between 2012 and 2022, but no studies were published in 2013, 2016, and 2021 (Table 1). The included studies were conducted in 7 different countries, and the country that published the most number of studies was France (3/11, 27%). Of the 11 included studies, 10 (91%) were journal articles, and 1 (9%) was a book chapter. The majority of the included studies (10/11, 91%) used parallel RCTs. The sample size in the included studies ranged from 20 to 114 participants, with an average of 56.7 (SD 33.9) participants. The mean age of the participants in the included studies was 74 (SD 4.21; range 66-81.2) years. The percentage of men in the included studies ranged from 30% to 71%, with an average of 49.2% (SD 14.5%). The mean Mini-Mental State Examination score was reported in 8 (73%) of the 11 studies, with a range of 17.9 to 28.1 and an average of 24.3 (SD 3.2). MCI was the most common disorder among participants in the included studies (8/11, 73%). Of the 11 studies, 9 (82%) recruited participants from clinical settings, whereas the remaining 2 (18%) studies recruited participants from the community.

Eleven different serious games were identified in the RCTs (Table 2). Two types of serious games were identified based on the therapeutic modality they provide: cognitive training games (10/11, 91%) and exergames (1/11, 9%). In 10 (91%) of the 11 studies, games were created with a serious specific goal from the outset (designed serious games). By contrast, a game in 1 (9%) of the 11 studies was not intended to be a serious game from the beginning, but it was modified to be used for a serious purpose (modified serious game). Computers were the most common platform used for playing games in the included studies (9/11, 82%). In most of the studies (8/11, 73%), serious games were played under the supervision of health professionals or caregivers. The game duration in the RCTs ranged from 7 to 90 minutes. The frequency of playing the games ranged from 2 to 5 times a week, whereas in more than half of the RCTs (6/11, 55%), it was 2 times a week. The intervention period varied from 2 to 25 weeks, but it was ≤12 weeks in 73% (8/11) of the studies.

The comparison groups received no or sham interventions in 7 (64%) of the 11 studies, whereas the groups received active interventions in 8 (73%) of the 11 studies (such as conventional exercises and other serious games; Table 3). The numbers do not add up because both active and passive comparators were used in 4 (36%) of the 11 studies. The active comparators lasted between 7 and 100 minutes. The active comparators were used 2 to 5 times a week. The active comparator period varied from 2 to 25 weeks. Of the 18 different tools used to evaluate the outcome of interest (ie, learning), the Rey-Osterrieth Complex Figure Test was most used by the included studies (4/11, 36%). The measured outcome was verbal learning in 9 (82%) of the 11 studies, whereas it was nonverbal learning in 8 (73%) of the 11 studies. The numbers do not add up because the outcomes in 6 (55%) of the 11 studies were both verbal and nonverbal. Although the outcomes were assessed immediately after the intervention in all included studies, only 4 (36%) of the 11 studies had a follow-up period, which varied from 4 to 74 weeks. Of the 11 studies, 9 (82%) reported participant attrition, which ranged from 0 to 22 (0%-22.9%) participants.

Reviewers’ judgments about each risk-of-bias domain for each included study are presented in Figure 2. A low risk of bias in the randomization process domain was observed in 45% (5/11) of the studies (Figure 3). The majority of the studies were judged to have a low risk of bias in 2 domains: deviations from the intended intervention (10/11, 91%) and missing outcome data (9/11, 82%). The risk of bias in the measurement of the outcome was low in all studies. In 4 (36%) of the 11 studies, the risk of bias in the selection of the reported results was low. According to these judgments, only 3 (27%) of the 11 studies were judged to have a low risk of bias in the last domain (ie, overall bias). The interrater agreement (Cohen κ) between the reviewers in risk-of-bias judgments was 0.87.

This review assessed the effectiveness of serious games on verbal learning among older adults with cognitive impairment. Our meta-analysis revealed that serious games are effective in improving verbal learning among older adults with cognitive impairment in comparison with no or sham interventions. According to our subgroup analysis, serious games do not have a different effect on verbal learning between patients with AD and those with MCI. Our review demonstrated that serious games are as effective as conventional exercises in promoting verbal learning among older adults with cognitive impairment, and this could be attributed to 2 reasons. First, only 2 (18%) of the 11 studies were included in the relevant meta-analysis [30,31]. Second, the serious games in these studies are of different types (cognitive training games [30] and exergames [31]) and genres (designed [30] and modified [31]) and have different platforms (PC [30] and Kinect [31]). In this review, we found that serious games outperformed no or sham interventions and conventional cognitive training in enhancing nonverbal learning among older adults with cognitive impairment. It is worth mentioning that none of the findings are based on high-quality evidence.

A previous review compared the effectiveness of serious games (cognitive training games in particular) with that of any intervention (passive or active intervention) on verbal and nonverbal learning among older adults with MCI and dementia [24]. In contrast to our findings, the study showed that serious games have a different effect between patients with AD and those with dementia [24]. To be more specific, the previous review [24] demonstrated that serious games are more effective than other interventions in improving verbal learning (P=.002) and nonverbal learning (P<.001) among older adults with MCI, whereas they are as effective as other interventions in enhancing verbal learning (P=.14) and nonverbal learning (P=.50). Another review examined the effectiveness of serious games (exergames in particular) on verbal and nonverbal learning among older adults in comparison with that of any intervention (passive or active intervention) [23]. Contrary to our findings, the review showed that serious games are as effective as other interventions in enhancing verbal learning (P=.09) and nonverbal learning (P=.26). The contrary findings between our review and the previous reviews can be attributed to the main differences between them as mentioned in the Research Gap and Aim subsection under the Introduction section.

Although this review demonstrated that serious games have the potential to improve verbal and nonverbal learning among older adults with cognitive impairment, our findings remain inconclusive for 3 reasons. First, the quality of evidence from most of the meta-analyses (5/6, 83%) was very low mainly because of the high risk of bias, low homogeneity, and lack of precision of the estimated total effect sizes. Second, the meta-analyses included a small number of RCTs (2-6). Third and last, most of the meta-analyzed studies (6/8, 75%) had a small sample size. Therefore, until further convincing proof of their effectiveness is offered, serious games should be used to supplement current interventions for verbal and nonverbal learning rather than replace them entirely.

Although PCs were used to deliver serious games in most of the included studies (9/11, 82%), no study used mobile devices (eg, smartphones and tablet devices), and only 1 (9%) of the 11 studies used VR headsets. As a PC allows for more distraction from outside sources, a serious game delivered through a VR headset might be able to hold the user’s attention for longer than a serious game delivered through a PC. In addition, mobile devices are more appealing because they are typically more affordable, widespread, and accessible than traditional platforms such as PCs and game consoles. In 2021, there were approximately 15 billion mobile devices and 7.1 billion mobile users worldwide. By 2025, it is anticipated that these numbers will rise significantly [40]. Therefore, game developers need to collaborate with medical experts and develop serious games that can be played on mobile devices and with VR headsets. This could pose a challenge because commercial game development is a lucrative industry and to convince game developers to develop serious games that can be played on nontraditional platforms would require incentives with appropriate funding sources. The result, however, would be commercial-level user-tested and appealing games designed especially with the aging population with cognitive impairment in mind. Given that none of the serious games in the included studies were designed specifically to improve learning, this approach could be groundbreaking.

Further research is needed to address the following limitations and gaps in the previous studies and to confirm the results with a larger evidence base: (1) lack of studies comparing serious games with other interventions such as conventional cognitive training and conventional exercises; (2) lack of studies comparing different types of serious games (adaptive vs nonadaptive games and exergames vs cognitive training games) and different platforms used to play serious games (PCs vs VR headsets vs mobile devices); (3) limited research on the effectiveness of serious games on learning among older adults with other cognitive disorders such as vascular dementia, Lewy body dementia, and Huntington disease; (4) lack of a follow-up period in most studies; (5) short intervention periods (≤3 months); (6) scarcity of studies conducted in low-income countries; (7) missing information of the mean and SD of pre- and postintervention changes in learning for each group; (8) small sample sizes; (9) issues of high risk of bias mainly in the randomization process or selection of the reported results; (10) missing information on the underlying psychological methodology or intervention of the serious games; and (11) variation in the selection of tools to assess verbal and nonverbal learning.

This review focused on the short-term effect of digital serious games on learning among older adults with cognitive impairment. Therefore, we cannot comment on the effectiveness of other interventions (nondigital serious games used for other purposes), on other outcomes (eg, long-term effect on memory, attention, and processing speed), or on intervention effectiveness among other populations (adolescents, young adults, middle-aged adults, and those without cognitive impairment) because they were beyond the scope of this review.

Another limitation of this review is that the effect sizes might be overestimated or underestimated, given that we used postintervention data rather than the pre- and postintervention change data to estimate the effect size for each study. Postintervention data were used because none of the meta-analyzed studies reported the mean and SD of the pre- and postintervention changes in learning, and there was no statistically significant difference in verbal and nonverbal learning between the groups at baseline in the majority of the meta-analyzed studies (7/8, 88%).

Most likely, some relevant studies were missed in this review, given that we did not include quasi-experiments, pilot RCTs, and studies published in a language other than English before 2010. There may be a concern about the internal validity of our findings because the quality of evidence in most of the meta-analyses (5/6, 83%) was very low.

Serious games can play a significant role in promoting verbal and nonverbal learning among older adults with cognitive impairment. Nevertheless, our findings remain inconclusive for 3 reasons. First, the quality of evidence from most of the meta-analyses (5/6, 83%) was very low mainly because of the high risk of bias, low homogeneity, and lack of precision of the estimated total effect sizes. Second, the meta-analyses included a small number of RCTs (2-6). Third and last, most of the meta-analyzed studies (6/8, 75%) had a small sample size. Hence, until further convincing proof of their effectiveness is offered, serious games should be used to supplement current interventions for verbal and nonverbal learning rather than replace them entirely. Game developers should consider developing serious games that can be played on mobile devices and with VR headsets. Further studies should be conducted to compare (1) serious games with other interventions such as conventional cognitive training and conventional exercises, (2) different types of serious games (adaptive vs nonadaptive games and exergames vs cognitive training games), (3) different platforms used to play serious games (PCs vs VR headsets vs mobile devices), (4) different intervention periods, and (5) short-term effect with long-term effectiveness of serious games.