Since the mid-20th century, with steady improvements in living standards and the growing affluence of global societies, greater longevity has led to great concern over the decline in performance and cognitive abilities that accompanies normal and neuropathological aging [1]. The decline in performance includes balance function, physical function, and fear of falling and depression. General cognitive function decline is described in terms of processing speed, working memory, attention, and executive functioning [2-5]. These demographic changes and neurological aging compromise the sustainability of health care systems, with more health care resources needed to care for the aging population [6]. Therefore, it is essential to propose feasible and acceptable interventions to promote active aging, intended to preserve and optimize opportunities for health, participation, and security in order to enhance the quality of life as people age.

With the rapid development of computer technology and the video game industry, game types and game experience have greatly improved. A video game is any game played on a digital device, encompassing a wide range of interfaces, including web-based programs and apps for mobile devices [7]. Exergames are video games that require physical activity and movement when played [8]. Video games are promising and adaptable cognitive training tools in the active aging process that can help stave off the negative effects of aging in performance and cognitive function. Moreover, video games are inexpensive and interesting and can be employed in hospitals as well as in the community [9]. Furthermore, older adults are now more digitally connected than ever, and most older people aged 65 years or more have a computer with an internet connection [10]. Therefore, the use and implementation of video game interventions are high.

Although off-the-shelf video games were not developed for serious purposes or these specific interventions, many studies have determined that video game interventions may benefit the elderly in terms of serious purposes, including cognitive function, fall prevention, and other benefits [1,11,12]. No previous systematic review has provided a comprehensive overview with meta-regression and Bayesian network meta-analysis evaluating which type of video game intervention has the best effect on performance and cognitive function.

We conducted systematic literature searches of PubMed, EMBASE, and the Cochrane Library from their inception to June 30, 2020. The MeSH search terms were as follows in the full-text search: “active video game,” “active game,” “older adults,” and “elderly.” We included both randomized controlled trials (RCTs) and case-controlled trials (CCTs) that met the following criteria: participants were older adults (aged ≥60 years) without dementia; the interventions were all kinds of active video games (such as those on Xbox 360 or Nintendo Wii, computer-based games, virtual reality–based games), which means that the participants perform the video game intervention in an active state rather than a static state, and the whole body needs to be involved in the video games; and controls included exercise, puzzle games, visual stimulation, and no game play, which means that even patients who play games are still in a static state. The above declared the difference between the intervention group and the control group in order to study the effect of active video games in the intervention group. Comparisons of interventions with video games versus nonvideo game control were made with reported outcomes of performance and cognitive function.

We assessed the effectiveness of 6 performance and cognitive outcomes by comparing the intervention group and the control group at the final point. The first outcome was balance function, tested using the Berg Balance Scale (BBS), balance test time (s), and balance test speed (m/s). The second outcome was executive function, tested using performance on the trail-making test B (TMT-B, s), delta (s), Stroop word (s), attention, working memory, and the Corsi block test. The third outcome was related to general cognition, tested using the Mini-Mental State Exam (MMSE; score) and the Montreal Cognitive Assessment (MoCA; score). Physical function was the fourth outcome, tested using everyday function and function tests (s and cm). The fifth outcome was processing speed tested using the TMT-A (s) and processing speed (number). The final outcomes were the fear of falling and depression, tested using the Efficacy Scale International (score) and the Geriatric Depression Scale (score). All these outcomes were continuous data.

Two investigators (CY and XH) independently conducted the electronic literature search. The reference lists of relevant publications were also checked, and no language restrictions were set. These 2 researchers evaluated eligible titles, abstracts, and full texts, and disagreements between them were resolved by discussion with a third researcher (author YZ [Yingshi Zhang] or PC). A preset table was designed to extract details of potentially relevant papers, including the first author, publishing year, study type, region, sample size, gender, age, body mass index (BMI), education year, MMSE score, video game type, frequency, period, follow-up, control type, other associated diseases, and community or hospital. Two investigators (authors CY and EJ) extracted all continuous data independently onto a Microsoft Excel spreadsheet. The quality of the included RCTs was evaluated according to Cochrane Collaboration’s tool for assessing the risk of bias [16], and the quality of the included CCTs was assessed using the Newcastle-Ottawa Scale (NOS) score [17]. We also used the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) scale [18] to evaluate the quality of outcomes. Disagreements were resolved by a third researcher (YZ [Yingshi Zhang] or PC).

We applied standardized mean differences (SMDs) and 95% CIs to summarize the 6 outcomes from pairwise meta-analysis. To determine the efficacy of video game intervention versus nonvideo game control, we performed subgroup analysis and meta-regression of various intervention types (ie, Nintendo Wii, Xbox 360, and other video games), types of activity control activities (eg, physical activity, visual stimulation, puzzle games, and no game play), and periods (0-4 weeks, 4-8 weeks, 8-12 weeks, and more than 12 weeks). To determine the heterogeneity among our included studies, P≤.05 or I²>50% indicated heterogeneity in the outcome. P<.10 revealed that a grouping method was a source of heterogeneity. The random effects model was used to ensure the accuracy of the summarized data. Publication bias was assessed using the Begg and Egger tests, where P≤.05 indicated the existence of publication bias [19,20].

We performed a Bayesian random effects network meta-analysis composed of 4 chains with 100,000 iterations after an initial burnin of 10,000 and a thinning of 2.5 in order to determine the most suitable video game intervention. We calculated the SMDs and corresponding 95% credible intervals (CrIs), and the mean rank and surface under the cumulative ranking curve (SUCRA) values were produced from network meta-analysis estimates with a consistent model. We also produced comparison-adjusted funnel plots to explore publication bias from network meta-analysis. All the aforementioned analyses were performed using StataMP, version 14.0 (StataCorp) and WinBUGS, version 1.4.3 (MRC Biostatistics Unit and Imperial College School of Medicine).

Figure 1 shows details of the selection process. The search strategy generated 820 citations in total. After duplication removal and preliminary screening, 122 potentially publications were scrutinized for eligibility. Finally, we identified 47 original studies [21-66] that met the inclusion criteria. Only 1 (2.1%) of them was a CCT, and the remaining 46 (97.9%) were RCTs. Overall, 1651 of 3244 (50.9%) participants were assigned to the intervention group and the remaining 1593 (49.1%) were assigned to the control group. The sample sizes ranged from 12 to 977 (Table 1 and Multimedia Appendix 1). Baseline characteristics were balanced except that the intervention group was notably older (Table 2). Risk-of-bias assessment was performed for each RCT and CCT, and all included studies had acceptable quality (Multimedia Appendices 2 and 3).

For the outcomes of balance function, we first compared balance scores (BBS, etc) pre- and postintervention. No significant differences were found overall and in subgroups by intervention and control types. Significant differences were only found at 8-12 weeks. Substantial heterogeneity was found overall and in all subgroups. The source of heterogeneity was not determined from meta-regression, and publication bias was occasionally found. For the balance test time (s), no significant differences were found overall and in all subgroup meta-analyses, with frequent substantial heterogeneity and no publication bias. In balance test speed (m/s) assessment, significant differences were found only for no game play as the control type (odds ratio [OR] 0.611, 95% CI 0.048-1.175). These findings suggest that video game interventions may improve balance in the elderly (Table 3); however, the most suitable intervention has not yet been determined.

For assessment of executive function, we first evaluated the results of the TMT-B (s). No significant differences were found in all subgroups, with frequent substantial heterogeneity. Second, no significant difference was found in the outcome of delta (s). Similarly, no significant differences were found in the subgroup and overall outcomes of the Stroop word test (s), with substantial heterogeneity. When we evaluated the attention in video games compared with nonvideo games, significant differences were found only in the period of more than 12 weeks, with only 1 original study. In the evaluation of working memory, significant differences were detected for overall outcomes (OR=1.034, 95% CI 0.305-1.763), other video games as the intervention type (OR=1.076, 95% CI 0.295-1.858), no game play as the control type (OR 1.023, 95% CI 0.133-1.914), and 4-8 weeks as the intervention period (OR 1.401, 95% CI 0.559-2.243), with substantial moderate-grade heterogeneity in these subgroups. For the Corsi block test, significant differences were found for no game play and intervention periods of 0-4 and 8-12 weeks. Only the outcome for 8-12 weeks in 4 studies showed no heterogeneity (OR 0.429, 95% CI 0.080-0.778, P=.42, I²=0.0%) with high quality (Table 4). In summary, video game interventions had little effect on executive function, except for memory-related functions.

Regarding general cognition, the first item assessed was the MMSE score. A significant difference was found with Xbox 360 as the intervention type and 4-8 weeks as the intervention period. The second item evaluated was the MoCA score. A significant difference was only found for activity control as the control type (OR 1.826, 95% CI 0.043-3.609), with substantial heterogeneity (Table 5). With respect to physical function, first, in the everyday function outcome, a significant difference was found only in the intervention period of 4-8 weeks (OR –1.045, 95% CI –1.866 to –0.223), with substantial heterogeneity. A significant difference was found in the function test (cm) outcome (OR 0.725, 95% CI 0.235-1.214), with low heterogeneity (P=.26, I²=25.5%). No significant difference was found for the function test (s) outcome (Table 6). In summary, video game intervention may have an effect on improving general cognitive function, but it had little effect on physical function.

We considered processing speed after video game intervention versus nonvideo game control. First, when we evaluated the TMT-A (s), significant differences were found for overall outcomes (OR –0.833, 95% CI –1.463 to –0.204), other video games (OR –0.874, 95% CI –1.558 to –0.190), activity control (OR –1.033, 95% CI –1.830 to –0.235), and for more than 12 weeks (OR –2.395, 95% CI –4.272 to –0.519), with substantial heterogeneity. In terms of processing speed (number), significant differences were found overall and in subgroups, while substantial heterogeneity was found overall (OR 1.084, 95% CI 0.765-1.402, P<.001, I²=95.5%) and for activity control (Table 7). In Falls Efficacy Scale assessment, no significant differences were found overall and in all subgroups. In the Geriatric Depression Scale assessment, significant differences were found for Xbox 360 and other video games (OR –0.651, 95% CI –1.164 to –0.138) as the intervention types and 0-4 and 8-12 weeks as the intervention periods (OR –1.800, 95% CI –2.745 to –0.854), with substantial heterogeneity (Table 8). In summary, video game intervention had a significant advantage in terms of processing speed and had a tendency to reduce depression scores.

Using pairwise meta-analysis, we observed that the video game intervention improved clinical performance and cognitive function, especially processing speed and depression scores, in the elderly. However, the most suitable video game intervention has not been determined. Moreover, many subgroups included fewer studies, which may have affected the accuracy of the results. Therefore, we selected the outcomes of the included studies to conduct Bayesian network meta-analysis among the 6 indicators to identify the most suitable type of video game intervention for the elderly. Figure 2 provides network plots of balance time with the intervention period (Figure 2A) and without the intervention period (Figure 2B), because most studies were included in this outcome.

For the indicator of balance function, first, in terms of balance time (s), compared with no game play as the control group, Xbox 360 as the intervention type ranked first, with a significant difference (SMD –3.34, 95% CrI –5.54 to –2.56), followed by other video games (SMD –1.15, 95% CrI –2.69 to –0.64), Nintendo Wii, and activity control. Significant differences could also be found in Xbox 360 versus other video games as the intervention type (SMD –1.62, 95% CrI –4.89 to –1.03), Nintendo Wii (SMD –4.14, 95% CrI –9.64 to –0.28), and activity control. For the BBS, significant differences were found only in other video games compared with Nintendo Wii as the intervention type (SMD 0.38, 95% CrI 0.03-1.79), with some publication bias (Multimedia Appendix 4). In summary, the intervention typical of Xbox 360 and other video games may be the best intervention method to maintain balance function over time (Figure 2A).

For general cognition, we combined the scores of MMSE and MoCA. Compared with no game play as the control type, which ranked the lowest, other video games as the intervention type ranked the highest, with a significant difference (SMD 1.23, 95% CrI 0.82-1.86), followed by Xbox 360 (SMD 1.13, 95% CrI 0.74-1.72), Nintendo Wii (SMD 0.90, 95% CrI 0.59–1.37), and activity control (SMD 1.44, 95% CrI 0.94-2.20). Significant differences were found in all comparisons, which suggests that the ordering from our Bayesian network meta-analysis was reasonable. No significant differences were found among the network outcomes from the TMT-B (s) to determine executive function (Figure 2B). In summary, video game intervention improved the cognitive function of elderly patients but had no effect on executive function.

In the assessment of processing speed, the first item examined was the TMT-A (s). Compared with no game play, other video games ranked first, with a significant difference (SMD –0.29, 95% CrI –0.49 to –0.08), followed by activity control (SMD –0.36, 95% CrI –0.57 to –0.15), virtual reality–based games, and Xbox 360. Significant differences were also found in comparisons of other video games and activity control versus virtual reality–based games and Xbox 360. The second item assessed was processing speed (number). Other video games as the intervention type also ranked first, with a significant difference compared with no game play as the control (SMD 0.72, 95% CrI 0.36-1.09), followed by virtual reality–based games (SMD 0.60, 95% CrI 0.26-0.94) and activity control (SMD 0.42, 95% CrI 0.06-0.77). In general, video game intervention improved processing speed (Figure 3A). For the evaluation of depression scales, compared with no game play, Xbox 360 ranked the highest, followed by other video games, activity control, and Nintendo Wii, with no significant difference. In terms of the Falls Efficacy Scale, there was no significant difference (Figure 3B).

In conclusion, based on Bayesian network meta-analysis, we determined that video game intervention improves balance function, cognitive function, and processing speed, which were similar results to those obtained using pairwise meta-analysis.

Our Bayesian network meta-analysis quantified the comparative effectiveness of video games based on 47 studies including 3244 elderly participants, with acceptable quality. We comprehensively summarized the comparative efficacy of video games in improving performance and cognitive function in 6 domains: balance function, executive function, general cognition, physical function, processing speed, and fear of falling and depression. The results suggested that, first, on pairwise meta-analysis, video game interventions are beneficial for cognition scores, processing speed, and depression scores. There are tendencies toward benefits for balance function, executive function, and physical function. Second, on Bayesian network meta-analysis, interventions with video games may improve balance function, cognitive scores, and processing speed in the elderly, which was similar to the results of pairwise meta-analysis. Third, from the ranking of the Bayesian network meta-analysis, Xbox 360 and other video games always ranked first, while Nintendo Wii always ranked last of all interventions. This was accompanied by having the most outcomes of moderate GRADE with low publication bias in both pairwise and Bayesian network meta-analyses.

In this systematic review, we used a comprehensive search with clear inclusion and exclusion criteria and carefully examined the efficacy of video game interventions in improving performance and cognitive function in 6 domains and 18 outcomes. Generally, the consistency of Bayesian network meta-analysis was similar to that of pairwise meta-analysis. On Bayesian network meta-analysis, compared with no game play as the control type, significant differences were found in balance time (s), cognitive scores, processing speed (TMT-A), and processing speed (number). For expert balance time (s), significant differences were also found in terms of cognitive scores, processing speed (TMT-A), and processing speed (number) on pairwise meta-analysis (Table 2; Figures 3 and 4). Based on the results of our study, video game interventions had the most obvious benefit for cognitive scores and processing speed. Processing speed is defined as the time spent completing mental tasks. It relates to the patient’s speed of understanding the information they obtain, whether it is visual (letters and numbers), auditory (language), or mobile. Similar results were found by Ozdogar et al [67], suggesting that video-based exergaming is almost as effective as conventional rehabilitation with respect to improving walking, upper- and lower-extremity functions, cognitive function, fatigue, depression, and health-related quality of life. Shin et al [68] found that participants who frequently played video games showed enhanced processing speed, which could be an effect of game practice. Mansor et al [2] determined limited effects of video games on cognitive function, and another valuable research published by Wang et al [69] also proved that game-based brain training can be considered a supplementary intervention for improving cognitive function in community-dwelling older adults. Moreover, Vázquez et al’s [70] research indicates that video game–based interventions may assist adults in active aging processes and prevent secondary aging. The above valuable studies all support our results.

In our study, we found that other video games and Xbox 360 are more effective than Nintendo Wii. Other types of video games were defined as exergames, video games, computer-based games, and virtual reality–based games. The main reason for these results may be that Xbox 360 and other video game screens require equipment to play, which have good platforms, a strong visual sense, and good interactivity. However, they are not easy to carry because of the requirement of external equipment.

A previous meta-analysis did not distinguish between different video games; therefore, this is an innovation of our study. A previous study suggested that after Nintendo Wii therapy, patients experienced motor learning retention, achieving a sustained benefit using the technique [71]. We observed that Nintendo Wii only benefited patients in general; therefore, when we choose video game interventions for elderly participants, we should choose better-intervention video game types with better interaction and visual stimulation for older adults. Video games could also improve mental health among older adults [72].

The mechanisms underlying changes following video game interventions remain unclear, although they might be related to tonic/phasic activation or inhibition of affected brain regions during video game playing [73] that make the participants feel satisfied. Specifically, psychological satisfaction and pleasure might be related to the various feedback mechanisms provided to the player by the active video game. This elaborate reinforcement and reward schedule has the potential to maximize motivation [74]. Video games are able to maintain flexibility of striatal responses to reward, a mechanism that might be extremely important to keep motivation high and therefore might be of critical value for many different applications, including cognitive training and therapeutic possibilities [75]. In the studies we included, only a few participants felt fatigue or leg muscle soreness; however, the players could tolerate and relieve themselves, suggesting that video game interventions are safe. Dankbaar et al [76] showed that video lectures from a serious game are effective for specific topics, such as patient safety.

This study had several limitations. First, we predesigned the inclusion criteria of video game intervention versus nonvideo game control. The definition of “video game” was broad, including Xbox 360, Nintendo Wii, virtual reality–based games, and computer-based games. However, the degree of stimulation, interaction, and pleasure of different video games for participants differed, resulting in clinical heterogeneity. Second, substantial heterogeneity was frequently determined on pairwise meta-analysis. One reason is the existence of clinical heterogeneity (many subgroups included less than 3 studies), and the other is the different types of combined scales for outcomes, leading to methodological heterogeneity. Third, some publication bias was detected in both pairwise and Bayesian network meta-analyses, which may be due to the difficulty in publishing negative outcomes. Lastly, to partially maintain the slight stability of the results under the Bayesian framework, we only chose outcomes that included more studies. Therefore, although the Bayesian network meta-analysis was not comprehensive, the results were more accurate.

In summary, impaired cognitive function is a highly prevalent condition that can profoundly influence the quality of life and accounts for major health care expenditures among the elderly. Our comprehensive Bayesian network meta-analysis provided evidence that video game interventions could be considered for the elderly to improve their performance and cognitive function, especially general cognitive scores and processing speed. Video games with better interactivity and visual stimulation have better curative effects. Based on the available evidence, we recommend video game interventions for the elderly. Future studies should be designed as multicenter RCTs, involving more subjects and providing more detailed description of the types of video games, in order to determine the most appropriate type of video game for older adults.