According to the United Nations, the number of people aged ≥65 years is projected to be 1.6 billion in 2050, which is double of what it was in 2021 [1], and the number of persons aged ≥80 years is projected to be 143 million in 2050, triple of what it was in 2019 [2]. This demographic shift raises significant concerns for the escalating burden on health care systems and associated financial implications. Advanced age is often associated with a higher prevalence of chronic illnesses and age-related conditions, necessitating increased medical attention and resources [3,4].

Technology can play a role in supporting older adults in health care by allowing quicker information and communication, preventing the development of chronic conditions, and monitoring health conditions [5]. At the same time, they can also reduce caregiver burden, leading to cheaper and better-quality care [6]. However, despite the potential benefits offered by these advancements, their widespread adoption remains limited, primarily due to ambivalence among older adults toward technology acceptance [7]. Therefore, identifying the factors that affect the acceptance of technology for older adults is one of the most important research needs to support older adults’ use of technology [8].

However, a key issue is that the current literature that aims to understand the acceptance of health care technology for older adults exhibits significant heterogeneity, with diverse studies yielding varying effects and strengths of predictors [9]. To explain, the technology acceptance model (TAM) [10] and the unified theory of acceptance and use of technology (UTAUT) [11] are commonly used by scholars to understand technology in the context of health [12] and were applied in research of different types of health care technology and their acceptance by older adults [13-15]. TAM uses perceived usefulness (PU) and perceived ease of use (PEOU), with the subsequent TAM2 adding subjective norm, while UTAUT identifies performance expectancy (PE), effort expectancy (EE), social influence (SI), and facilitating conditions as core determinants to explain behavioral intention (BI). The constructs of TAM and UTAUT can be combined due to their conceptual similarities and complementary nature and previous studies have used the findings of both models in meta-analysis [16]. PU and PE both reflect the belief that using technology will enhance performance, while PEOU and EE indicate the perceived effort required to use the technology. SI, akin to subjective norm in TAM2, addresses the influence of other important factors on technology use. Both models use the constructs to explain BI, the intention to adopt technology.

To elaborate the heterogeneity of the TAM and UTAUT literature, studies have shown that older adults are more likely to accept technology that meets their needs and expectations [17,18]. However, the impact of PU on technology acceptance varies for health care technology. Li et al [19] found that PU had very little impact on the BI to adopt a remote health management service for older adults while Mahmood and Lee [20] reported a high influence of PU for health monitoring wearable technology. For older adults, ease of use is crucial because physical and cognitive abilities affect the acceptance and use of technology [21,22]. However, studies are conflicting. Wu et al [23] identified a high effect for the acceptance of medical self-service terminals, but Khan et al [24] observed a low effect for mobile health (mHealth) services. Similarly, for SI, while it is understood SI can significantly affect older adults’ technology adoption, particularly those from their children, friends, and professional caregivers [25], heterogeneity exists in the current literature. Koo et al [26] detected a high effect of SI for the acceptance of a personalized health care service app while Wong et al [27] determined that SI had no effect on the use of the internet for health information in one of the 2 models used in the study.

Such heterogeneity makes it difficult to interpret results because the inconsistent findings across studies prevent a clear, unified understanding of the effects of PU, PEOU, and SI on BI for health care technology acceptance among older adults. This study aimed to perform a meta-analysis to systematically aggregate and analyze these diverse results, providing a more robust and comprehensive assessment of the factors influencing health care technology adoption in this population and the characteristics of the primary studies that have moderating effects. Similar research, such as the quantitative meta-analysis by Chong et al [16], provided an expansive study into the TAM and UTAUT literature on health care information technologies but did not focus on older adults and was limited to a specific type of health care technology, while Ma et al [28] focused on literature that applied the 2 models for older adults without specifying the type of technology. Therefore, a meta-analysis that specifically targets older adults and their acceptance of different types of health care technology is needed to provide a more defined analysis of the current literature.

To ensure a clearer interpretation of the current literature regarding the acceptance of health care technology for older adults, this study aimed to (1) synthesize the sample size–weighted average of the PU-BI, PEOU-BI, and SI-BI relationships for the current literature that used TAM or UTAUT to examine the acceptance of health care technology for older adults and (2) identify sources of systematic heterogeneity by analyzing the sample and methodology characteristics of the primary studies that moderate the PU-BI, PEOU-BI, and SI-BI relationships.

While traditional systematic review frameworks such as patient, intervention, comparison, and outcome or sample, phenomenon of interest, design, evaluation, research type were considered, these structures did not optimally align with our research focus on technology acceptance behavior, particularly as elements such as interventions or comparisons were not directly applicable to our context. Instead, we developed a focused search strategy specifically designed to capture studies examining TAM, UTAUT, and their associated constructs in the context of older adults’ health care technology acceptance.

The systematic search was conducted in 5 databases (Google Scholar, Web of Science, Scopus, IEEE Xplore, and ProQuest) on March 15, 2024. Google Scholar was included for its ability to capture a broad range of interdisciplinary studies beyond purely clinical research, aligning with the scope of this study. While we initially searched PubMed as it is a prioritized database for clinical studies, it yielded fewer relevant results, leading us to expand our search to additional databases. Testing confirmed that the PubMed results were already captured by the selected databases, affirming the adequacy of our chosen sources.

A combination of 3 groups of word strings was developed using logical operators (AND/OR) as follows: (1) age-related keywords: older adult OR elderly OR ageing OR aging, (2) theoretical framework keywords: unified theory of acceptance and use of technology OR UTAUT OR technology acceptance model OR TAM OR acceptance OR adoption OR intention, and (3) context of use keywords: health OR healthcare OR well-being OR gerontechnology.

The search terms were collaboratively developed by 3 reviewers (HJY, JHL, and WJL) and refined through iterative rapid searches across the selected databases. These rapid searches were conducted to evaluate the effectiveness of the search terms, ensuring they comprehensively captured relevant studies while minimizing irrelevant results. The final search terms were chosen to align with the study’s focus on TAM and UTAUT frameworks for health care technology acceptance among older adults. The full search terms and detailed search strategy are provided in Multimedia Appendices 1 and 2. The number of articles identified, screened, eligible, and included were recorded according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement (Multimedia Appendix 2).

In total, 2 reviewers independently screened the titles and abstracts to identify studies eligible for full-text screening and subsequently conducted the full-text screening of the selected studies. Any disagreements during this process were resolved through a majority vote, with input from a third researcher to ensure consensus. To ensure that the selected studies were directly relevant to the research question, a rigorous inclusion criteria were applied during the full-text screening as described in Textbox 1.

If a study’s sample included 2 or more of the same characteristics (eg, participants from both Asia and Europe and survey regarding multiple types of technology), it was not included. Theses and dissertations were involved as well to reduce the chance of publication bias. If a study met the criteria but did not report the zero-order correlations, an email was sent to the corresponding author to request the information. The inclusion and exclusion criteria are shown in Textbox 1.

The coding procedure for the primary studies’ characteristics was designed to ensure the extraction of the required data. The 2 researchers separately extracted the information of the primary studies into a Microsoft Excel spreadsheet. If any disagreements arose, the third researcher, who is an experienced statistician, was involved for a resolution. The following information was coded:

The mean age was calculated using frequency counts in age-stratified data or directly extracted from the studies if provided. Geographic region was coded based on the location of the sample collection. The gender ratio was determined by dividing the number of male participants by the number of female participants. Technology domains were categorized into 4 distinct groups: mHealth (eg, mHealth apps and mHealth services), wearable (eg, smart clothing and smart watches), online and telemedicine (eg, online health platforms and remote consultations), and home and institutional health hardware (eg, fall monitoring systems and self-service health kiosks). The presence of technology demonstrations in the studies, whether through a video or live demonstration just before the survey or experiment, was recorded as a binary variable.

In total, 2 reviewers independently assessed the methodological quality of studies using the Newcastle-Ottawa Scale. The Newcastle-Ottawa Scale uses a star system to evaluate the methodological quality of studies. The adapted cross-sectional tool assigns up to 8 points across 3 domains: (1) selection of study groups (up to 4 points), (2) comparability of the groups based on age and sex (up to 2 points), and (3) assessment of outcomes (up to 2 points). For this study, it was adapted to properly evaluate TAM- and UTAUT-related studies. The ascertainment of exposure was evaluated based on the use of surveys, with a score of 1 if surveys were used and 0 if no information was given. The comparability domain assessed control for age and sex, with 2 points if both were controlled, 1 point if one was controlled, and 0 points if neither were controlled. The assessment of outcomes was based on the use of validated surveys and the reporting of Cronbach α, with 2 points if both were provided, 1 point if only the validated survey was used, and 0 points if neither were described.

Data were synthesized according to the review objectives.

The literature search process yielded 1167 studies (Figure 1). After the removal of duplicates, 928 studies remained for the full-text screening. After assessing the full-text articles for eligibility, the resulting number of studies included in the meta-analysis was 41. There were 2 studies that met the eligibility criteria after additional information was provided by the corresponding author.

The summary of the characteristics of the studies is included in Table 1. The total sample size of the primary studies was 11,574. Of the 41 studies, 33 (80%) explicitly included the mean age of participants, with an overall mean of 67.58 (SD 4.76) years and a range of 57.86 to 82.1 years. In total, 5 (12%) studies provided an exact age range: ≥60 years [19,24,33], 50 to 65 years [34], and ≥50 years [35]. The remaining 3 (7%) studies [36-38], while not specifying a mean age, used consistent but indefinite terminology, such as elderly or older adults in their title, abstract, introduction, or methodology. Therefore, it was assumed that the sample of the 3 (7%) studies was consistent with the focus of this research and fell within the broader age range represented by the 33 (80%) studies that reported mean ages. In total, 36 (88%) studies reported the gender distribution of their samples, the total ratio being 2.00, with more female participants than male. For each technology type, there were 9 (22%) studies for wearable, 12 (29%) for mHealth, 12 (29%) for online or telemedicine, and 8 (19%) for home or institution hardware. In total, 23 (56%) studies were conducted in Asia (China, Taiwan, and Korea), 7 (17%) in Europe, 4 (10%) in the United States, and 7 (17%) in African-Islamic countries (Bangladesh, Saudi Arabia, and Pakistan). The description of the studies is included in Table 2.

Of the 41 studies, most did not end with just a correlation analysis but conducted more in-depth statistical methods, such as structural equation modeling and multivariate regression. Although all the correlation coefficients were positive, certain studies reported path coefficients that were not statistically significant or did not have the exact pathway in their final model. The rate at which the correlation coefficient represents the final reporting of the studies is given in Table 3. For the PU-BI relationship, 41 correlation coefficients were extracted from a correlation matrix given in the study. Of the 41 correlation coefficients, 31 (76%) had PU-BI path analysis, 27 (66%) of which were significant and positive with a rate of 87% (27/31). For the PEOU-BI relationship, 41 correlation coefficients were extracted. Of the 41 studies, 28 (68%) included the PEOU-BI path analysis, 21 (51%) of which were positive and significant with a rate of 75% (21/28). For the SI-BI relationship, 28 correlation coefficients were extracted. Of the 28 studies, 21 (75%) included the SI-BI path analysis, 16 (57%) of which were positive and significant with a rate of 76% (16/21).

The weighted average of the correlations of the 3 pairwise relationships, PU-BI (r=0.607, 95% CI 0.543-0.665; P<.001), PEOU-BI (r=0.525, 95% CI 0.462-0.583; P<.001), and SI-BI (r=0.551, 95% CI 0.468-0.624; P<.001), was calculated using the random-effects model. The CI of each effect indicated a positive relationship for the intention to accept technology. The heterogeneity was calculated through a Q test and an I² test for PU-BI (Q₄₀=973.77; P<.001; I²=95.9%), PEOU-BI (Q₄₀=626.95; P<.001; I²=93.6%), and SI-BI (Q₂₇=580.59; P<.001; I²=95.3%), all of which showed a high degree of heterogeneity. To test the possibility of publication bias, the funnel plot method was used to detect any asymmetry. If detected, the trim-and-fill method was used to adjust the weighted average of the effect by filling in additional points to maintain symmetry. Among the 3 relationships, SI-BI required 6 additional points as shown in Figure 2, while the PU-BI and SI-BI required none. The weighted average of the effect sizes and the adjustments after the trim-and-fill are reported in Table 5. The table summarizes the results of the 3 pairwise relationships through quantitative synthesis by confirming that each pair has a positive association, indicating that the intention to accept health care technology is dependent on the usefulness, ease of use, and SI.

This meta-analysis aimed to address 2 primary objectives: synthesizing the current evidence on health care technology acceptance among older adults and identifying potential sources of heterogeneity. The findings revealed significant positive correlations for all 3 key relationships: PU-BI (r=0.607), PEOU-BI (r=0.525), and SI-BI (r=0.551). Further analysis identified significant moderating effects of geographic region on PEOU-BI, technology type on both PU-BI and SI-BI, and visual demonstrations on PU-BI and SI-BI relationships, while age and gender showed no significant moderating effects.

The pooled results extend our understanding of how older adults accept health care technology. While a separate study noted a lower correlation between PU and BI for health care technology compared to other types of technology [65], our meta-analysis revealed that PU maintains a strong influence on older adults’ intention to use health care technology. The significant PEOU-BI relationship aligns with the understanding that older adults may face cognitive and physical challenges [66,67], making ease of use particularly crucial for technology adoption and accessibility [68]. The strong SI-BI relationship confirms the substantial influence of family members, friends, and caregivers on health care technology adoption decisions [65], providing empirical support for the importance of social networks in technology acceptance among older adults.

This meta-analysis has several important limitations that should be considered when interpreting the findings. First, the main source of analysis, which is the correlation coefficient of the TAM constructs, were used instead of the finalized path coefficient of the primary studies. While this ensures the comparability and synthesis of the weighted averages of the effect sizes, it may not fully represent the results of the primary studies as the control variables are removed in the correlation analysis. Second, the substantial heterogeneity observed across the included studies, as reflected in the high I² values (PU-BI: 95.9%, PEOU-BI: 93.6%, and SI-BI: 95.3%), represents a significant limitation. This heterogeneity likely stems from differences in study populations, geographic regions, types of health care technology, and methodologies. Although the use of a random-effects model mitigated the impact of this variability by accounting for between-study differences, it remains challenging to generalize the findings to all older adult populations or health care technologies. Researchers should interpret the pooled effect sizes cautiously, as they represent averages across diverse study contexts rather than universally consistent effects. Third, the definition of older adults varied across the included studies, with most defining participants as >60 years of age [14,15,26,40,44-50,56-58,60,61], while others set the range at 55 years [13,27,54,63,64], 50 years [23,39,41,43,55,59], or ≥65 years [20,52,53,62]. Some studies did not clearly specify participant ages, relying instead on general descriptors such as older adults or elderly. Although the meta-regression suggested that mean age did not significantly influence the results, this variation reflects a potential limitation in how age was operationalized. This lack of consistency should be considered when interpreting the findings, and future research should aim for more standardized age definitions. Finally, the use of different methods across studies to collect information has noteworthy limitations. Although most studies obtained their data through web-based surveys, other methods such as face-to-face distribution, telephone, and mobile surveys were used, creating the possibility of the mode effect.

This meta-analysis provides a framework to understand and interpret the heterogeneous nature of health care technology acceptance among older adults. Rather than viewing inconsistent findings in previous literature as methodological weaknesses, our results suggest they may reflect genuine variations in acceptance patterns across different contexts and technologies. These insights have important implications for both research and practice. Future studies should carefully consider and report contextual factors that our analysis has identified as significant moderators, while also addressing current methodological limitations through standardized age definitions and balanced sampling approaches. This could lead to a more comprehensive understanding of technology acceptance among older adults and support the development of more effective implementation strategies for health care technology—an increasingly important consideration as health care systems worldwide adapt to serve older adult populations.