In 2017, the global population of adults aged ≥60 years was 962 million, more than twice the number of older adults in 1980 [1]. By 2050, it is expected that the number of older adults will double, reaching nearly 2.1 billion [1]. As the population ages, delaying the onset of illness and disability and retaining physical function are top public health priorities [2]. Physical activity (PA), defined as any bodily movement produced by skeletal muscles that requires energy expenditure [3], is one way to achieve this. However, evidence suggests that 31% of the global population does not meet the recommended levels of PA [4,5], and inactivity has been identified as a leading risk factor for mortality, accounting for >5 million global deaths annually [6]. A recent umbrella review including 24 systematic reviews reported that older adults that are physically active have a lower risk of all-cause and cardiovascular mortality, breast and prostate cancer, fractures, disabilities with activities of daily living, functional limitations, risk of falling, cognitive decline, dementia, Alzheimer disease, and depression [7]. In 2018, the World Health Organization released their global action plan on PA to combat inactivity and improve health over the next decade [8].

In late 2019, SARS-CoV-2, which causes COVID-19 disease, emerged and quickly became an international health crisis, and in March 2020, the World Health Organization declared COVID-19 a global pandemic [9]. Since then, many countries have established strict public health measures to curb the spread of the disease, including social distancing and isolation. Although these measures have the benefit of minimizing viral transmission, which is critical for older adults who are at a higher risk for more severe illness [10,11], they have also exacerbated levels of physical inactivity. A systematic review including 66 studies with nearly 87,000 participants from 26 countries reported significant declines in PA during the lockdown owing to the COVID-19 pandemic [12]. Unfortunately, older adults are also at a higher risk for consequences of inactivity, such as frailty, sarcopenia, and chronic diseases, compared with their younger counterparts [13]. These data highlight an urgent need to evaluate alternative methods of improving PA levels. Fortunately, with advancements in technology, smart technology has become an increasingly relevant and studied tool for achieving health objectives [14,15]. Smart technology interventions may represent an ideal alternative to traditional face-to-face programs as they have the potential to overcome service delivery barriers such as limited access; inconvenience of travel; absenteeism from work and family [16]; and, now importantly, minimizing unnecessary exposure to COVID-19 for those who are most at risk. The role of smart technology in improving PA in older adults warrants further evaluation both now and for informing future directions of health care delivery.

Smart technology capitalizes on communication and information technologies (eg, internet and video calls) [14] and uses different mediums, such as computers and tablets, or mobile health, which includes smartphones, wearables (eg, FitBit), and mobile apps (eg, My Fitness Pal, Samsung Health, and Apple Health) [17]. Systematic reviews have demonstrated that smart technology interventions can improve PA levels, specifically steps per day and minutes per day of moderate to vigorous PA in generally healthy older adult populations (mean age ≥55 years) [18-21]. However, there are still unanswered questions and a need for more and better evidence. For example, existing reviews have examined only specific types of smart technology [20] or included only digital PA estimates without considering participant-oriented outcomes [19]. Importantly, existing systematic reviews have not compared smart technology PA interventions with more traditional modes of PA intervention delivery (ie, face-to-face) [15,20-23]. This comparison is essential for determining whether interventions that include smart technology components are more, less, or as effective as face-to-face alone interventions. Therefore, the purpose of this review is to determine the effects of PA interventions that use smart technology compared with face-to-face PA interventions on PA and physical function in community-dwelling older adults.

This systematic review will answer whether the PA interventions that use smart technology are more, less, or as effective as face-to-face alone interventions for increasing PA and function in older adults. The secondary questions were as follows: (1) What are the effects of smart technology PA interventions on secondary outcomes, including health-related quality of life (HRQoL), anxiety, and depression? and (2) Does the effectiveness of smart technology interventions differ by type of PA or by the type of smart technology used (eg, wearable vs mobile app)?

This systematic review was conducted in accordance with a peer-reviewed protocol [24] registered in PROSPERO (CRD42020135232) and followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) guidelines [25]. The full protocol has been published elsewhere [24].

A comprehensive search of the MEDLINE, Embase, CINAHL, and AMED databases from inception to February 2021 was conducted after consultation with a health research librarian [26]. The full search strategies are available in Multimedia Appendix 1 [27-45]; common Medical Subject Headings across databases included age, technology, physical fitness, with keywords to capture all types of PA and smart technology. Reference lists of included studies were hand searched to identify additional relevant studies.

Data from the included studies were extracted independently and in duplicate using a standardized data collection form [24]. Two reviewers independently assessed studies using the Cochrane risk of bias tool [47] and the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) system [48].

Meta-analyses for primary and secondary outcomes were conducted using random effects models with standardized mean difference (SMD) and mean difference (MD) where appropriate in Review Manager (RevMan; version 5.4, The Cochrane Collaboration, 2020) [24]. According to the Cochrane handbook, when necessary, we converted scales to correct for the difference in direction [49], and median and IQR were converted following the methods of Wan et al [50]. When possible, we performed sensitivity analyses by removing studies with an overall rating of a high risk of bias for each outcome. This deviates from the initial protocol, wherein we planned to remove only studies with a high risk of bias in ≥3 domains. Where appropriate, we completed subgroup analyses for our secondary questions. Where possible, we completed analyses for interventions that used smart technology alone compared with face-to-face alone.

We identified 12,245 records from our search; reviewers screened 9434 titles and abstracts after duplicate removal, and 19 RCTs were eligible for inclusion (Figure 1). The reasons for full-text exclusion are provided in Multimedia Appendix 1. Reviewers attempted to contact 7 corresponding authors for missing information, and 1 author responded. All the studies were combined in a narrative synthesis, evaluating a total of 3455 participants. A total of 18 RCTs with 3405 participants randomized to either a smart technology or face-to-face intervention were included in the quantitative analysis.

Characteristics of the studies are shown in Tables 1-2. Of the 3455 participants, 1874 (54.24%) were female, and the mean age ranged from 60 to 72 years. Studies were conducted at outpatient or community practices in the following countries: United States (6/19, 32%); Belgium (2/19, 11%); Netherlands (2/19, 11%); United Kingdom (2/19, 11%); and one each in Australia, Chile, Denmark, Finland, Hong Kong, New Zealand, and Spain. Participants were community-dwelling older adults, with 14 studies focused on specific clinical populations, including people with chronic obstructive pulmonary disease (COPD) [27-31], cardiovascular disease [32-35], diabetes [36,37], knee arthritis [38,39], obesity [40], and cognitive impairment plus physical frailty [41].

Detailed descriptions of the smart technology interventions in each study are provided in Multimedia Appendix 1. A total of 16 studies included a single smart technology component, including smartphone apps [36,41], wearable activity trackers (eg, Fitbit) [35,38], telerehabilitation (eg, video conferencing, virtual advisor, or health buddy device) [28,32,42], and pedometers (ie, only provides step counts) [27,29,30,39,40,43,44], whereas the remaining 3 had multiple components, including video conference and pedometer [37]; website and SMS text messaging [34]; and website plus pedometer, SMS text messaging, and email [33]. A total of 15 studies included smart technology and a face-to-face component in the intervention group [27,29-36,38-41,43,44]. Furthermore, 4 studies evaluated smart technology alone versus face-to-face alone [28,37,42,45]. The length of interventions ranged from 10 weeks [27] to 1 year [29], and follow-ups ranged from 6 weeks [32] to 2 years [37].

Overall, the risk of bias was a concern for studies included in this review. On the basis of the outcome with the highest risk of bias (ie, if we assessed risk of bias for 3 outcomes in a study and 1 of those was rated a high risk of bias and the other 2 were some concerns, we rated the study as high risk of bias), 14 studies were judged to be high risk [27,29,31,32,34,36,37,39-45], 4 studies had some concerns [28,30,35,38], and 1 study had a low risk of bias [33]. The areas of greatest concerns were missing outcome data (10/19, 53% high risk) and risk of bias related to measurement of the outcome (eg, awareness of intervention and influence of knowledge of intervention on the assessment; 9/19, 47% high risk). A summary figure is available in Multimedia Appendix 1.

A total of 10 studies did not report adherence rates [30,32-34,37,38,40,43,45]. Two studies reported overall adherence to the intervention [28,42], and 7 studies reported adherence to the components of the intervention [27,29,31,35,39,41,44]. Overall, across the available studies, adherence rates to smart technology interventions ranged from 64.8% [42] to 95.5% [27]. Adherence to specific smart technology interventions was reported as follows: pedometers ranged from 76% [39] to 95.5% [27], and telerehabilitation ranged from 64.8% [32] to 85% [28]. The adherence to face-to-face interventions ranged from 64% [28] to 100% [41].

Three studies measured depression using different scales, including the Hospital Anxiety and Depression Scale [28], Beck Depression Inventory [27], and Care Depression Instrument [37]. We pooled data from 2 RCTs for a random effects meta-analysis, which showed no significant difference between interventions that included smart technology and face-to-face alone (SMD 0.09, 95% CI −0.23 to 0.41) [27,28]. Furthermore, 6 studies measured HRQoL using the Clinical COPD Questionnaire [28,31], St George’s Respiratory Questionnaire [27,30], HeartQoL global score [33], and Chronic Respiratory Disease Questionnaire [29]. A meta-analysis of 6 studies demonstrated a significant SMD (0.31, 95% CI 0.11-0.51) in favor of smart technology interventions. For HRQoL, we completed a subgroup analysis by smart technology component; multicomponent interventions (n=2; SMD 0.35, 95% CI 0.05-0.66) showed significant improvements compared with face-to-face alone, but pedometers did not (n=3; SMD 0.04, 95% CI −0.49 to 0.57). After performing a sensitivity analysis excluding studies with a high risk of bias, our results did not change the significance or direction of results. Only 1 study examined anxiety using the Hospital Anxiety and Depression Scale and demonstrated an improvement in favor of smart technology at the first follow-up [28]. Meta-analyses for secondary outcomes are available in Multimedia Appendix 1.

PA is integral to reducing age-related illness and disability. This review of 19 studies involving 3455 patients found that interventions that include smart technology may improve steps per day, total PA, walking, and HRQoL in older adults. Although the overall quality and certainty of the evidence were judged to be very low and more precise estimates will need to be obtained, our results may have important implications for research and practice on PA promotion, especially in the context of the COVID-19 pandemic and public health restrictions.

Among the included studies, we found that most (n=15) smart technology interventions used multiple components with intervention groups also receiving a face-to-face component. One of the challenges of multicomponent interventions is that it limits our understanding of the effectiveness of smart technology alone. For example, intervention and control groups in several studies received usual care, with the only difference between groups being the addition of a pedometer or wearable activity tracker [29,38-40]. Of the 4 studies examining smart technology alone, 2 examined the effect on steps per day; our subgroup analysis found a nonsignificant difference of 260 steps per day favoring smart technology [28,42]. Notably, the results from these 2 studies were contradictory, with Hansen et al [28] reporting that face-to-face interventions were more effective. However, they evaluated patients with severe COPD and reported a decline in PA for both groups over the course of the study. This may be caused, in part, by the progressive nature of COPD or by the short length of the intervention at only 10 weeks [28]. In line with our meta-analysis for steps per day, the 3 RCTs examining the effect of smart technology alone on total PA showed a small, nonsignificant effect favoring smart technology [37,42,45]. It will be important for future work to determine whether smart technology on its own is more, less, or as effective as face-to-face interventions. This is especially important during the COVID-19 pandemic, as in-person contacts should be limited to minimize the risk of transmission. Conversely, if smart technology interventions alone are less effective, it will be important to revert to including some type of face-to-face component as soon as it is safe and feasible.

Patient-important outcomes were not well represented in our included studies, which highlights an important gap in the design and reporting of smart technology interventions. Among 19 studies, only 7 (37%) unique studies reported on patient-important outcomes: 6 reported on HRQoL [27-31,33], 3 on depression [27,28,37], and 1 on anxiety [28]. Evidence has demonstrated associations between PA levels and mental health and quality of life in older adults [51]. Therefore, it is vital that research includes these measures to further our understanding of the magnitude and direction of these relationships. This is timely as we have seen the prevalence of depression and anxiety increasing over the last 2 years with the pandemic and public health restrictions [52]. Older adults have been shown to be at an increased risk for anxiety [52]. In addition to the lack of patient-important outcomes, we found that data on intervention adherence were largely missing or inconsistently reported. This leads to several challenges, including understanding the acceptability of these interventions for older adults and interpreting results of the primary studies. For example, if adherence to the intervention was poor, it would be difficult to appreciate if the differences (or lack thereof) between groups were because of intervention failure or implementation failure. The issue of acceptability is also crucial, particularly considering that the current literature on the usability of technology in this population is limited.

Our results suggest that the type of smart technology used for PA interventions may influence the effectiveness of the intervention. We found that studies that used smartphone apps led to significant improvements in steps per day and total PA scores [53]. Other systematic reviews examining the effectiveness of mobile phone interventions in adults aged >18 years and >50 years and older adults (>65 years) have shown mixed evidence for improving PA [53-55]. Potential reasons for these discordant results may be the diversity of interventions within control groups, small sample sizes, and moderate to high levels of heterogeneity across studies [53,54]. We also found that wearable activity trackers significantly improved the number of steps per day. Researchers should consider the type of smart technology in conjunction with the PA goals and the population to achieve the best outcome. Further research is warranted to determine the optimal types of smart technology interventions for older adults.

The existing literature has several limitations that warrant future research. Our findings are based on very low quality of evidence, as per the GRADE criteria. All included studies had some concerns or a high risk of bias, most commonly because of missingness of outcomes, lack of intention-to-treat analyses, inadequate allocation concealment, and lack of prespecified statistical analyses or protocols. The variability in populations, smart technology interventions, control group interventions, and outcomes among studies are major contributing factors to the large degree of statistical heterogeneity. Furthermore, most studies used multiple components, making it difficult to assess which parts may have contributed to differences between groups or, conversely, if additive components may have diluted a potentially effective intervention. Importantly, 79% (15/19) of the studies included an element of face-to-face interaction in the intervention group, making it difficult to evaluate the effectiveness of smart technology alone. Due to the limited number of studies and small sample sizes, there is a need to group studies and smart technologies broadly to have sufficient sample sizes to conduct meta-analyses. Given that different smart technologies can be used for different intervention components, it may be difficult to apply our findings in practice when designing interventions. Although we conducted subgroup analyses where there were sufficient data, there is a need for additional research comparing different types of smart technologies for supporting specific PA interventions in older adults [18,19,23]. Finally, our results may have been influenced by the high risk of performance bias caused by the impracticability of blinding therapists and participants owing to the nature of the interventions. Future research should focus on minimizing the risk of bias, evaluating individual smart technology components and smart technology alone, and including standardized control groups. Improved reporting of control group interventions may also assist with interpretation of results.

This review also has some limitations. We did not include gray literature owing to a lack of central sources to identify and retrieve these citations [56]. In addition, we excluded studies published in languages other than English because of the feasibility of the review. Therefore, our cohort of studies may not represent the entirety of the literature. This review has several important strengths. To the best of our knowledge, this is the first study to attempt to compare smart technology interventions specifically with face-to-face interventions, which is critical for determining their effectiveness compared with traditional modes of delivery. In addition, we published our peer-reviewed protocol, and we developed and conducted our search in collaboration with a health research librarian.

In the context of substantial heterogeneity and very low quality of evidence, our results suggest that PA interventions that include smart technology components may significantly improve steps per day and total PA in community-dwelling older adults. Subgroup analyses showed that smartphone apps and wearable activity trackers seem to be the most effective smart technology, which may be helpful for health care practitioners when determining appropriate methods of remote PA promotion. When comparing smart technology alone with face-to-face alone, there were few studies with discordant results and no significant differences between groups. The results should be interpreted with caution given the challenges with the existing literature cited in the discussion.