This review was conducted in strict accordance with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [70]; also see Multimedia Appendices 1 and 2 [71]. We registered the study in PROSPERO (Prospective Register of Systematic Reviews; CRD42023486090). We used the PICOS (Population, Intervention, Comparison, Outcomes, and Study Design) framework [72] to structure study inclusion and exclusion criteria (Table 1). Our population of interest had at least 1 indicator indicating cognitive ability, behavioral ability, or emotion-related ability, and their age was ≥60 years. We sought studies describing interventions with digital SGs implemented among older adults with cognitive impairment in an experimental group (severe game group [SGG]), including exergames or video games, cognitive training games, computerized CBT-based games, and biofeedback games. There were no restrictions regarding the time of intervention. For comparison, we had a control group (CG) that received a different intervention, such as routine health care, traditional Tai Chi, or music. Our outcomes of interest included cognitive ability, daily behavioral capacity, and depression in older adults with cognitive impairment. We excluded studies that presented reviews, non–randomized controlled trials (RCTs), and papers for which the full text was unavailable. All included studies were published in Chinese or English. Studies in which the results were interpreted from the perspective of older adults with cognitive impairment were deemed eligible.

The following databases were searched to identify the relevant literature: PubMed, Embase, Web of Science, Scopus, and Cochrane Library. The following search terms were used for the search: (“Cognitive Ability” OR “Daily Behavior Ability” OR “Mental Health” OR “Depression”) AND (“Serious Games Based” OR “Serious Games” OR “Web-Based or Mobile”) AND (“Old Adults in Cognitive Impairment” OR “Old Adults” OR “Older People”) AND (“A pilot study” OR “Randomized Controlled Trial” OR “RCT”). Detailed search strategies for each database are given in Multimedia Appendix 3. The keywords were used to search for papers published from January 1, 2017, to December 31, 2023. The snowball method was adopted to search the reference lists of the included studies. Additionally, the reference lists of the included studies were manually searched to identify eligible papers. Unpublished academic literature was not considered eligible. The reference lists and relevant reviews were then screened to identify any pertinent studies. To identify additional relevant publications, we also retrieved gray literature [73].

Four researchers (authors XZ, YT, YC, and ZZ) screened all the literature for inclusion. After removing duplicates using EndNote X9, all studies were independently screened based on titles and abstracts. Next, we carefully read the full texts of the remaining papers in accordance with the inclusion and exclusion criteria. Finally, data were extracted from the included studies. The co-first author (XZ) used the modified version of the data extraction table in the Systematic Review Manual of Cochrane Interventions to extract data. A modified version of the intervention description and replication was using following the template for intervention description and replication (TIDieR) list and guidelines [74] to extract data of the intervention, design, and delivery features, as defined in the introduction. First, data were extracted by the co-first author (XZ). Next, the second author (ZZ) checked the accuracy of the process. Finally, the co-first author (YC), through thematic analysis [75], extracted the following data from the list of 14 studies [34-47]:

Using the Cochrane Collaboration risk-of-bias tool [76], we independently assessed the risk of bias. Additionally, we assessed the quality of the 14 studies using the Jadad scale [77]. The risk-of-bias tool assesses 7 domains: (1) generation of a random sequence, (2) allocation concealment, (3) subject and experimenter blinding, (4) outcome assessor blinding, (5) resulting data integration, (6) selective reports, and (7) other bias risk. Each study was categorized as having a low, high, or unclear risk of bias. We also performed Begg and Egger tests to evaluate the degree of publication bias [78].

Stata 15.1 and Review Manager 5.3 were used to conduct data analysis. Forest plots were constructed to intuitively illustrate the results. In the included literature, the outcomes were measured as continuous variables, and the same indicator was assessed with different tools. These indicators were presented as standard mean differences (SMDs). α=.05 was considered statistically significant. Heterogeneity was assessed with I² statistics, which were classified as high (>75%), moderate (50%-75%), or low (<50%) [79]. In cases of high heterogeneity, sensitivity analysis was performed using the leave-1-out method to identify the sources of heterogeneity. In the analysis, numerous weeks were compared as subpoints to check the results. The Begg test [80] and Egger test [81] were used to assess publication bias. P<.05 indicated significant results. When a meta-analysis included at least 10 studies, publication bias was evaluated with a funnel plot [82,83]. In this study, 95% CI and the SMD were examined [84]. Overall, P<.05 indicated statistically significant effects. Subgroup analyses were performed based on the country of intervention (outside Europe or not) and the length of intervention (weeks).

Figure 1 shows the study selection process. After searching 13 databases, 4260 relevant records were identified. For duplicate removal, we imported all the studies into EndNote X8 [85]. After removing 1893 (44.4%) duplicates and eliminating 2352 (55.2%) papers via a strict screening process, 14 (0.3%) RCTs [34-47] from before December 2023 involving 714 participants were finally included. Studies were excluded because they did not report SDs [26,86-88], were review papers [89,90], or were not within the scope of this meta-analysis [91-94]. All the included studies reported that digital SGs have a positive impact on cognitive ability, daily behavioral capacity, and depression. The primary outcomes of interest were an increase in cognitive ability scores (assessed with the MMSE and the MoCA), an increase in daily behavioral capacity indicators (assessed with the Barthel Index, the Katz Index, the Lawton and Brody Index, the Korean version of the Modified Barthel Index [K-MBI], and the Activities of Daily Living [ADL] scale), and a reduction in depression (assessed with the Goldberg Anxiety and Depression Scale [EADG], the 15-item Geriatric Depression Scale [EDG-15], and the Cornell Scale for Depression in Dementia [CSDD]).

Table 2 and Multimedia Appendix 4 [34-47] show the overall characteristics of the included studies. All 14 studies were published before 2023. The sample sizes ranged from 17 to 112, and 714 older adults aged >60 years were enrolled in the included studies, including 340 (47.6%) subjects in the SGG and 374 (52.4%) subjects in the CG. All participants were older adults with cognitive impairment, but they had not been diagnosed with psychiatric disorders. All interventions were based on digital SGs, and their durations ranged from 4 weeks to 12 weeks. The interventions lasted from 45 minutes to 5 hours each week. Both group training and individual training methods were used [34-47]. All 14 studies [34-47] were subdivided into digital SG training methods (exergames or video games, cognitive training games, computerized CBT games, and biofeedback games).

Figures 2 and 3 show the results of the risk-of-bias assessment. All 14 papers [34-47] described the randomization methods in detail. The blinding method was detailed in 7 (50%) studies: 4 (57%) of these studies were double-blind trials [40,44,45,47], while 3 (43%) were single-blind trials [36,37,46]. The average Jadad score across all included studies was 4.7, indicating fair-to-mild quality.

Subgroup analyses of daily behavioral capacity and depression scores were performed based on the country of intervention (outside or within Europe) and the duration of intervention (weeks). For daily behavioral capacity, significant differences were found in the SMD between the 2 subgroups: within Europe (P=.85) [43,46,47] and outside Europe (P<.001) [37,38,40]. SGs had a significant effect on the outcome in the outside-Europe group (SMD 6.68, 95% CI 5.08-8.28; P<.001), as shown in Figure 8. For depression, significant differences were found in the SMD between the 2 subgroups: within Europe (P<.001) [43,47] and outside Europe (P=.02) [34,40]. SGs had a significant effect on the outcome in the within-Europe group (SMD –2.08, 95% CI –2.94 to –1.22; P<.001), as shown in Figure 9. Of the 6 (43%) studies, 4 (67%) [38,40,46,47] reported a pooled effect of an intervention duration of ≥8 weeks (SMD 0.67, 95% CI 0.19-1.16; P=.007). For the remaining 2 (33%) studies [37,43], the pooled effect within the intervention period was as follows: SMD −1.32, 95% CI −3.05 to −0.42; P=.14. The effects of SGs on daily behavioral capacity during the 2 periods of intervention were significantly different from those in the CG (P<.001), as shown in Figure 10. For depression, no significant differences were found in the SMD between the 2 groups after >8 weeks (P=.005) [40,47] or <8 weeks (P=.008) [34,43], as shown in Figure 11. Among the 4 (29%) studies on depression indicators [34,40,43,47], 2 (50%) [40,47] had a combined effect in an intervention duration of ≥8 weeks (SMD −2.57, 95% CI −4.00 to −1.13; P=.005). For 2 (50%) studies [34,43] in which the intervention duration was <8 weeks, the combined effect within the intervention period was as follows: SMD −4.95, 95% CI −7.84 to −2.07; P=.008. An intervention period of 8 weeks or more had a significant effect on reducing cognitive impairment in older adults (Figure 11). Compared to with the CG, the effects of SGs on depression during the 2 periods of intervention were significantly different, and a significant difference between the 2 groups was found (P<.001), as shown in Figure 11.

In the sensitivity analysis (Table 3), we excluded the study by Yang et al [38] and observed a significant change in heterogeneity from 91% to 0%. We hypothesized that the outcome of daily behavioral capacity indicators may have been the possible source of heterogeneity in this study. Additional possible reasons included the inclusion of patients, specific treatment modalities, and inconsistencies in clinical indicators between domestic and foreign countries. The patients included and specific treatment modalities, as well as the clinical indicators applied, in this study were as follows: Yang et al [38], Fujian (China), age 61-83 years; older adults with cognitive impairment, participated in somatosensory interactive games, and completed the MMSE and the ADL scale. All these different factors were possible sources of heterogeneity.

We found that the funnel plots for the MMSE (Figure 12a), MoCA (Figure 12b), daily behavioral capacity indicators (Figure 12c), and depression indicators (Figure 12d) were all symmetrical, indicating the absence of publication bias. P>.05 indicated the absence of obvious publication bias. The Begg test (P=.03) and Egger regression (P=.026) also indicated a lack of publication bias.

This study systematically evaluated the efficacy of training older adults with cognitive impairment using digital SGs, drawing upon data from 14 studies with a total of 714 participants. Our findings revealed that digital SGs significantly alleviate depression and yield large improvements in cognitive ability and daily behavioral capacity. This suggests that SGs offer a highly effective alternative for older adults with cognitive impairment. Due to statistical heterogeneity, these results should be interpreted with caution. In this study, sensitivity analysis was used to explore the major sources of statistical heterogeneity to determine the impact of a single study on overall risk. The type of intervention (various digital SGs), sample size (range 17-112), intervention duration (range 4-12 weeks), weekly intervention time (range 45-300 minutes/week), type of CG (traditional Taijiquan group, routine health care group, music group), cultural background, measurement tools, or other confounding factors may have caused heterogeneity. To the best of our knowledge, this is the first meta-analysis and systematic review evaluating the effectiveness of the 3 outcome indicators of serious gameplay in older adults with cognitive impairment. In this study, SGs were found to significantly increase cognitive ability and daily behavioral capacity and decrease depression.

Previous studies have explored the effects of digital SGs on cognitive ability [57], daily behavioral capacity [95], and depressive emotion problems [96] in patients with MCI or AD. However, previous studies have reported inconsistent findings regarding the effects of digital SG therapy on cognitive function [21,60-62], daily behavioral capacity [52,63], and depression [21,50,64-66] in older adults with MCI or AD. This systematic review included the recent literature (2017-2023) and examined 3 indicators of the effects of SGs on cognitive ability, daily behavioral capacity, and depression among older adults with MCI or AD. Consistent with previous findings, our results indicated positive effects of SGs on MMSE [97] and CSDD [50,98] scores and negative effects on depression on EDG-15, EADG, and GDS scores. In addition, there was a significant increase in cognitive ability (measured with MoCA) and daily behavioral capacity (measured with the ADL scale, the K-MBI, the Instrumental Activities of Daily Living (IADL) scale, the Katz Index, the Barthel Index, and the Lawton and Brody Index). This systematic review is the first to simultaneously test the effects of SGs on the 3 indicators of cognitive ability, daily behavioral capacity, and depression. This systematic review also divided the subgroup analysis of weeks and continental plates, which was not available in previous systematic reviews.

SGs are defined as games that are not for entertainment purposes and are mainly used for prevention, education, diagnosis, screening, and rehabilitation [99,100]. The word first appeared in 1968 when Abt [101] named his new book on SGs. According to Clark’s [101] proposed treatment, games can exist in a variety of forms, including (1) exergames or video games that require physical exercise as part of playing the games [38,39]; (2) cognitive training games, which aim to maintain or improve users’ cognitive abilities (eg, executive function, memory, learning) [34-37]; (3) computerized CBT games, which are video games that provide CBT to the users; and (4) biofeedback games, which are video games that use electrical sensors attached to the player to receive information about the player’s body state (eg, electrocardiogram sensors) and seek to influence some of the player’s body functions (eg, heart rate). Special care units (SCUs), also known as dementia care units, began to appear in the United States in the 1980s for older adults with dementia [102]. With an increase in the number of older adults with dementia and the diversification of the demand for dementia care in the market, these SCUs provide not only services to older adults in institutions but also family counseling, care support, and community therapy services to older adults with MCI [103]. At the policy level, the United States also launched the National Plan to Address Alzheimer’s Disease in 2012, with the goal of preventing and effectively treating AD by 2025 [104]. The latest SG therapy has also had a positive impact on easing the stress of dementia among older adults in US society [66]. However, the attention to older patients with MCI and AD in the West is much earlier than in China. Model-based projection data put China’s older adults with dementia at 6.73-15.7 million in 2025; the number of older adults with dementia will increase to 10.018-23.3754 million in 2035 [105]. China’s Circular on the Promotion of Dementia Prevention and Treatment Action (2023-2025) [106] formally focuses on groups such as older adults with MCI and AD at the national policy level. Ren et al [11] found that the 2015 annual treatment cost of patients with AD in China was US $167.74 billion, with ever-rising treatment costs expected to reach US $1.8 trillion by 2050. Older care facilities, social work stations, and counseling services in China may use SGs to carry out nonpharmacological treatments for patients with MCI and AD. This will provide both physical and psychological health improvements to older adults and at the same time reduce the financial burden on China. In traditional Chinese society, positive family and social relationships contribute to physical and mental health [107]. Due to the “Xiao Dao” culture, this also constitutes a unique social security system for older adults [108,109]. However, rapid industrialization and urbanization have changed these traditional arrangements [110]. Today, older adults are less likely to live with and interact with younger generations, which has resulted in older adults being isolated [111]. This has a direct impact on access to social care and financial security and even affects the physical and mental health of older adults [112]. SGs can provide opportunities for older adults to interact with others [113] and reduce the negative emotional impact on them [114]. In addition, the sports elements included in SGs [115] can exercise older adults’ cognitive ability [67] and daily behavioral capacity [116] and may even promote intergenerational relationships [117]. Our findings also show that SGs can have a positive impact on cognitive ability and daily behavioral capacity and decrease depression in older adults. With this in mind, we suggest that nursing home caregivers, social workers, and psychologists consider using SGs in their work and collaborate with SG professionals to bring benefits to older adults with MCI or AD. This will slow down the cognitive loss in older adults with MCI or AD to a certain extent.

Compared with previous studies [90,118,119], this review is the only one to specifically evaluate the effects of SGs on older adults and related types of games. Because the study was conducted in strict accordance with PRISMA guidelines, it was a high-quality, transparent review. Only RCTs were included; this is the most rigorous method of studying causal relationships [120,121]. Therefore, the conclusions of this review are more credible. In this review, there was minimal publication bias. We searched 12 worldwide databases and included research from the United Kingdom, China, Belgium, the Netherlands, Portugal, Spain, and Korea. We also performed forward and backward checks of reference lists, and an all-around search strategy was applied. We searched a database of gray literature without restricting the search to a specific comparator, population, setting, or country, which made the samples more representative. Since all the processes (ie, data extraction, research selection, evidence quality assessment, and bias risk assessment) were independently conducted by 4 reviewers, there was no risk of bias in this review. In addition, because we used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) method to evaluate the quality of evidence, this review will enable readers to draw more accurate conclusions. Where possible, our statistics integrate data, which will increase not only the power of the research but also estimates of the effects of digital SG therapy on cognitive ability, daily behavioral capacity, and depression. Moreover, due to the large sample size, we performed a subgroup analysis and a sensitivity analysis. In addition, this review used pre- and postintervention outcome data from each group to evaluate the effectiveness of each meta-analysis. All previous studies have reported the SD and average overall cognitive changes pre- and postintervention in each group. In addition, in these studies, there were significant differences in overall cognition between groups at baseline.

The limitations of this study should also be considered. First, due to the design of the systematic review, a causal relationship with a clear structure could not be inferred. Second, the outcome indicators of anxiety (EADG scores) and quality of life (Dementia Quality of Life (DQoL)) were also related to the mental health of older adults with cognitive impairment. However, with only 1 study reporting these findings, a comparison could not be made, so we did not include these indicators. Third, because of the limitations of the designs of the included studies, the randomization and blinding methods used were seldom described in detail. Only 11 studies described the randomization method. In contrast, randomization was only mentioned in other studies, but no explanation of the method used was provided. Blinding was only implemented in 7 studies. Fourth, since many potentially eligible studies were pilot RCTs or quasiexperiments published before 2017 or in languages other than English, we excluded them. As a result, some related research may have been missed. The reason for excluding these studies is that the internal validity of pilot RCTs and quasiexperiments is lower than that of RCTs [122,123]. In addition, we were also unable to translate all the non-English research literature. In our previous review, we found some studies published before 2017. However, serious progress has been made in the field of gaming in the past 7 years, so we did not include RCTs published before 2017. Fifth, this review focused on the short-term effects of SGs. This meta-analysis of pre- and postintervention data did not include follow-up data. We found that follow-up data were reported in only 4 studies, and there was no consistency between the follow-up periods. Therefore, it was impossible to evaluate the long-term effects of SGs on the cognitive ability, mental health, and daily behavioral capacity of older adults. The quality of the evidence in our meta-analysis was average, which may have affected the internal validity of the findings. Sixth, in this review, interest intervention was limited to SGs used as therapeutic interventions on digital platforms, so it would not have been possible to comment on and check the effectiveness of SGs and nondigital SGs used for other purposes, such as monitoring, screening, or diagnosis.

Since the review concentrated on the effectiveness of SGs in improving cognitive ability, daily behavioral capacity, and depression among older adults with cognitive impairment, researchers should evaluate the effectiveness of SGs by performing further reviews of their positive effects on cognitive abilities (eg, processing speed, executive function, learning, memory, speed), daily behavioral capacity (eg, executive function, capacity for reaction, upper limb activity), and mental health (eg, anxiety, stress, sleep quality, quality of life) among people in different age groups. We included some studies from middle-income countries, so the generalizability of the findings of this review may be limited to these countries. As the cultural, social, and economic conditions of middle-income countries are different, additional relevant research should be carried out. To date, only a few studies have assessed the effectiveness of SGs in improving cognitive ability. Therefore, additional research is needed to help draw clearer conclusions about the effectiveness of SGs. Due to the lack of follow-up data, the long-term effects of SGs were not evaluated in this review. To assess the long-term cognitive effects of SGs in a worldwide context, researchers should follow up with participants. A few studies did not report the SD or average global cognitive changes in each group before and after the intervention. In the future, relevant studies need to report this information to evaluate the effect more accurately. Because most of the existing studies have problems reporting results and choosing the randomization process, 8 studies had a low risk of overall bias. Therefore, when developing and reporting on RCTs, researchers should use recommended tools (eg, RoB-2 [124]) and follow recommended guidelines to avoid the aforementioned biases. Many studies have examined the worldwide impact of digital SGs on the cognition of healthy older adults [57], but this review revealed that only 14 studies have investigated the effects of digital SGs on older adults with cognitive impairment. We hope that researchers can perform additional research in this area to fill this gap.

According to the results of this review, SGs, especially digital SGs, exhibit greater effectiveness than a lack of interventions. These findings have meaningful implications for clinical practice since the use of SGs can improve cognitive ability and daily behavioral capacity and decrease depression in older adults with cognitive impairment. However, these findings should be interpreted with caution because the quality of the evidence is low in most meta-analyses. Moreover, most of the studies included in this review had some problems in terms of overall bias. Except for 7 studies [34,39,41-43,45,47], in all the other included studies, heterogeneity was high and the total effect was not accurate enough. Therefore, before more strong evidence is obtained, psychiatrists and psychologists should consider SGs as a complement to existing interventions rather than a direct substitute. In addition, the number of older persons worldwide will grow exponentially in the next few years, so we can try to use SGs to lighten the burden on the health care system. We can use SGs to improve the physical, psychological, social, and motor functions of older persons with cognitive impairment, thereby improving their quality of life [54,125]. Currently, many old people live alone and have limited social interactions, which can often lead to increased morbidity and mortality [126]. Older adults can play SGs in a relatively cozy environment (ie, at home), thus promoting their connections with friends and family [51]. In all the studies, the platforms for SGs were mobile devices (tablets and smartphones). Mobile devices are more popular than game consoles, cheaper than computers, and easier to access; thus, they are particularly popular. As of 2021, there were more than 7.1 billion mobile device users and 15 billion mobile devices worldwide [127]. Game and app developers can work together to develop SGs aimed at improving the mental health, daily behavior capacity, and cognitive ability of older people. These SGs can be operated and played using mobile devices. A survey of a small number of studies conducted in middle-income countries revealed that these countries seem to prefer SGs more than high-income countries do [128]. Middle-income countries lack many psychological professionals (currently 1 for every 10 million people). Therefore, more SGs should be developed in middle-income countries to improve the cognitive ability, mental health, and daily behavioral capacity of older persons in those countries.

Overall, SGs, particularly digital SGs, may be associated with increased cognitive function and daily behavioral capacity, as well as decreased depression, among older adults with cognitive impairment. However, we are still unable to draw definitive conclusions on the safety and effectiveness of SGs in improving the cognition, behavioral capacity, and mental health of older people with cognitive impairment. This is due to the poor quality of existing meta-analysis evidence, which is mainly caused by concerns about the bias of most studies, the accuracy of the total effect, and the high degree of evidence heterogeneity. Therefore, until more reliable evidence is obtained, psychiatrists, psychologists, patients, and social workers should all regard SGs as a complement to existing interventions rather than a direct substitute. Before SGs are used, their effectiveness and safety should be further reviewed and evaluated, mainly in terms of improving daily behavioral capacity (eg, reaction ability, executive function, upper limb activity) of people with or without cognitive impairment in different age groups, cognitive ability (eg, processing speed, executive function, learning, memory), and mental health aspects (eg, stress, anxiety, sleep quality, quality of life). Further research is needed to assess the safety of SGs, the effects of sports games, and their long-term effects. In the future, further research on methods for increasing the motivation of participants, reducing the rate of missing data, and maintaining the effects of SGs is needed. Second, due to study limitations, it was impossible to perform subgroup analyses for intervention types. Consequently, in-depth and stratified discussions and comparisons of different intervention types were performed. In the coming years, it will be necessary for researchers to perform higher-quality studies with larger samples to verify the effectiveness of SG for older adults with cognitive impairment.