Many workers spend most of their waking hours in the workplace [1], which is an environment that can have both positive and negative effects on health [2]. As such, the workplace is the best environment to apply the concept of health promotion. The World Health Organization declared that workplaces should be a priority for health promotion [3]. Workplace health promotion entails employers, workers, and communities working together to improve workers’ mental and physical health and overall well-being [4]. Elaborately designed workplace health promotion not only improves workers’ health [2] but also positively affects their productivity [5]. The primary challenge in workplace health promotion involves how to increase worker participation, given that less than 50% of participants typically remain in workplace health promotion programs [6] and the median attrition rate is 28% [7]. These obstacles can be overcome by incorporating health care technologies (eg, electronic health, mobile health [mHealth], wearable devices) into workplace health promotion strategies [8].

Health care technologies can increase workers’ interest, motivation, and participation in workplace health promotion [9,10]. These technologies function as cost-effective health promotion and disease prevention mechanisms by allowing workers to monitor their own health. Workers are at increased risk from stress caused by heavy workloads and unhealthy lifestyles, including lack of exercise and frequent drinking, relative to the general public [11]. In particular, office workers sit for long periods of time in the workplace, which exposes them to an increased risk of developing chronic diseases such as heart disease, cerebrovascular disease, and hypertension [12] that are all associated with high mortality rates as the second, third, and ninth most prevalent causes of death in Korea, respectively [13]. The cost associated with the treatment of cardiovascular disease has reached US $6.9 billion [14], which is higher than the US $4.7 billion spent on cancer, as the most prevalent cause of death, and the burden of disease in Korea is high. Moreover, mortality from cardiovascular disease has steadily increased over the past 10 years [15]. Effective prevention and management are essential because cardiovascular disease not only harms workers’ health but also increases medical expenses [16] and contributes to the social burden caused by decreased corporate productivity [17].

Workplace health promotion using health care technology has been shown to improve participants’ physical activity and eating habits [18,19]. Setting goals, supporting self-monitoring, and providing feedback on changes in physical activity and eating habits can be an effective mechanism for workplace health promotion. Various health care technologies have been studied for efficient workplace health promotion application. Mattila et al [20] conducted a 1-year randomized controlled trial to investigate the activity and usefulness of personal health technologies (web services, mobile apps, and personal monitoring devices) that support workplace health promotion. The authors showed that less than 30% of subjects continued to use mobile apps and web technologies, and that the key requirements for personal health technologies were simplicity, integration with everyday life, and clear feedback. Cook et al [9] conducted a randomized controlled trial for 3 months to evaluate the effectiveness of a web-based workplace health promotion program and found that web-based programs were more effective than a print-based intervention for improving diet and nutrition, but not for improving stress and physical activity. Balk-Møller et al [21] conducted a randomized controlled trial for 38 weeks to investigate the motivations of workers involved in web and app-based workplace health promotion and reported that social functions were more popular than personal functions, and social factors motivated continued use.

Choia et al [22] studied workers’ intention to use health care technology through an investigation of construction workers' acceptance of wearable devices (smart vests and wristbands) for occupational safety and health based on the technology acceptance model (TAM) [23]. They found that perceived usefulness, social influence, and perceived privacy risk were related to the intention to adopt wearable devices. Mohadis et al [24] investigated office workers’ acceptance of mHealth apps designed to increase physical activity based on the unified theory of acceptance and use of technology (UTAUT) [25], and found that performance expectancy and social influence had a significant effect on behavioral intention, but not on effort expectation. Sari et al [26] proposed a UTAUT-based conceptual framework to identify factors that influence worker adoption of and intention to use mHealth technology. Technology acceptance of mHealth apps has also been tested in patients with chronic diseases [27], younger adults [28], and health care professionals [29], but related research on workers in workplaces is still in its infancy [30]. Although many mHealth apps have been developed to date, few of these apps have been developed specifically to improve workplace health promotion [31].

Interest in personal health management is rising as aging and the incidence of chronic disease increase [32]. Moreover, active services focused on prevention and health promotion are needed [33] to record lifestyle factors such as exercise, nutrition, and sleep via various wearable devices [34-36] and to measure blood pressure, blood sugar, and weight via personal health devices [37]. Occupational factors such as the workplace environment should also be considered in managing chronic diseases, either by integrating occupational information into the electronic health record (EHR) [38] or implementing the occupational data for health model [39]. Recently, Health Level Seven (HL7) designed a fast health care interoperability resource (FHIR) profile [40] to represent patients’ occupational elements in personal health records (PHRs). PHRs are electronic tools that allow secure access, management, and sharing of health information [41], which is generally monitored by patients [42]. Individuals can check medical records provided by hospitals, monitor information regarding prescribed medicines and test results, and manage exercise and diet information related to health promotion [43]. Patients can use PHRs to reduce additional medical expenses, and disease management, treatment, and prevention activities can be enhanced as cooperation is improved through communication among medical personnel [44,45].

Employers expect workers to participate in workplace health promotion and enjoy the benefits provided by the organization. In this situation, the concept of PHRs is prominent owing to program technology-based attributes [46]. Employers can provide PHRs that motivate workers’ health care [47]. Employer-sponsored PHRs are driven by commercial goals to reduce productivity loss and health insurance costs by promoting a healthy lifestyle [48]. Despite the interest in and expected effects of PHRs, it is difficult to successfully provide these services [49]. Google Health, released in 2008, suspended the service in 2011 because of poor user participation [50], and Microsoft announced in November 2019 that they would stop providing their PHR HealthVault. However, Apple’s HealthRecord offers large-scale services that could be linked to over 200 medical institutions as of February 2019. These services are offered free of charge, but users’ technology acceptance for PHRs remains low and must be addressed [51]. To implement and operate a successful PHR app service in the workplace, employers, policymakers, developers, and planners must be aware of factors affecting workers’ technology acceptance.

Therefore, the purpose of this study was to analyze the factors that influence acceptance of PHR apps in the workplace and examine differences in perceptions surrounding the main factors. We applied a research model that included Bauer’s perceived risk [52] as an independent variable to the UTAUT [25], which is widely used to explain acceptance of new information technology.

The PHR app is an important element of the next generation of health care services and supports personal health promotion by storing and managing important personal medical data in one location. Moreover, using the UTAUT model to analyze and predict users’ technology acceptance of the PHR app is a rational approach. In this study, we set the main variables of effort expectancy, social influence, performance expectancy, and facilitating conditions as factors affecting behavioral intention to use the PHR app based on the UTAUT model. In the model, the dependent variable, use behavior, is affected by behavioral intention and facilitating conditions, and behavioral intention is determined by performance expectancy, effort expectancy, and facilitating conditions. However, as PHRs are currently in the introduction stage in Korea, research on actual users is limited. Therefore, in this study, we assumed that facilitating conditions also affected behavioral intention, and therefore included behavioral intention as a dependent variable without considering use behavior. Gender and age were also assumed to moderate the effects of performance expectancy, effort expectancy, and social influence on behavioral intention.

The PHR app is accompanied by various risks, which exert direct effects on behavioral intention. For example, medical data collected and utilized by the PHR app contain highly sensitive information, and behavioral intention decreases when there is a high probability of a data breach or fraud. In addition, behavioral intention decreases when users cannot securely manage as much information as they expected, or believe their information will be used for other purposes. These risks should be minimized when introducing health care services based on the PHR. Therefore, this study extended the existing UTAUT model by including perceived risk as a main variable (Figure 3).

Performance expectancy is defined as “the degree to which an individual believes that using the system will help improve his or her performance” [25]. Performance expectancy is similar to the perceived usefulness of the TAM [23]. In this study, performance expectancy refers to the degree to which users believe that using the PHR app will help them improve their health. The PHR app not only collects and manages the worker’s own data but also improves motivation and participation for workplace health promotion through personalized feedback from health experts [74-77]. Previous studies have demonstrated that performance expectancy and usefulness affect behavioral intention [51,63,78-80]. Therefore, we proposed the following hypothesis, H1: Performance expectancy in the PHR app will exert a positive effect on behavioral intention.

Effort expectancy is defined as “the degree of ease associated with the use of the system” [25], and affects the intent to use and the ease with which users can learn and use the system, similar to the TAM’s perceived ease of use [23] and innovation diffusion theory [81]. In this study, effort expectancy refers to the PHR app’s ease of use; the app should be easy to use, taking workers’ diversity into account (eg, age, type of business). Given that hard-to-use functions have a negative impact on users [82], PHR app functions should make it easy for users to access their data. Previous studies have shown that ease of use affects behavioral intention [51,83,84]. Therefore, we proposed the following hypothesis, H2: Effort expectancy in the PHR app will exert a positive effect on behavioral intention.

Social influence is defined as “the degree to which an individual perceives that important others believe that he or she should use the new system” [25]. In this study, social influence refers to the degree to which users feel that others deemed to be important or work colleagues believe that the user should use the PHR app for enhanced health management. Social influence is associated with organizational culture in the workplace [85], which is known to be an important social characteristic that affects organization, group, and individual behavior [86-89]. Social influence is an important factor in workers’ communications and interactions with their colleagues [90], which has been identified as a suitable factor for technology acceptance in previous studies [55,91-93]. Therefore, we proposed the following hypothesis, H3: Social influence in the PHR app will exert a positive effect on behavioral intention.

Facilitating conditions are defined as “the degree to which an individual believes that an organizational and technical infrastructure exists to support the use of a given system” [25]. In this study, facilitating conditions refers to a user’s belief that an organizational and technical infrastructure exists to support PHR app use. The number of aging individuals is rapidly increasing. Older workers are less capable of acquiring and adopting new technologies such as information and communication technology, and need adequate training and technical support for using such tools. Previous studies have shown that facilitating conditions affect workers’ behavioral intention to use new technology [94-96]. Therefore, we proposed the following hypothesis, H4: Facilitating conditions in the PHR app will exert a positive effect on behavioral intention.

Perceived risk is defined as “the uncertainty due to unforeseen consequences” [52]. In this study, perceived risk refers to the degree to which users are aware of possible loss associated with uncertainty surrounding PHR app use. Since workplaces require insight into the health status of their organizations and workers, personal health data should be strictly protected [97]. The health manager in the workplace handles sensitive medical data such as workers’ medical checkups and data from individuals who have abnormal findings; the PHR app collects and manages these data. Confidence in the system, information privacy, and security concerns affect the sharing behavior of PHRs [98]. Therefore, we propose the following hypothesis, H5: Perceived risk in the PHR app will exert a negative effect on behavioral intention.

In UTAUT, performance expectancy, effort expectancy, and social influence are moderated by gender, age, experience, and the voluntary nature of use [25]. Gender and age are generally perceived as significant factors in attitudes toward information technology [99], which previous studies have treated as moderating variables in technology acceptance [100-102]. Therefore, we proposed the following hypotheses: H6, Effects of performance expectancy on behavioral intention of the PHR app will be moderated by gender and age; H7, Effects of effort expectancy on behavioral intention of the PHR app will be moderated by gender and age; and

H8, Effects of social influence on behavioral intention of the PHR app will be moderated by gender and age.

The questionnaire consisted of 42 items. Responses to the 21 items (see Multimedia Appendix 1) measuring the main variables were on a 5-point Likert scale ranging from “strongly disagree” to “strongly agree” (Table 1). In addition, 12 items pertained to participants’ general characteristics and 9 items concerned the experience of and functions required of the app.

After receiving ethical approval (IRB No. P01-201902-23-014) from the Korea National Institute for Bioethics Policy, we recruited workers and health experts who visited 21 worker health centers with cooperation of the Korea Occupational Safety and Health Agency. The survey included 40 workers and 10 health experts from each worker health center and considered regional distribution.

We used a paper-based questionnaire to verify the proposed research model. The survey was conducted for approximately 7 weeks from February 12 to April 6, 2019. In total, 1050 questionnaires were distributed and 900 were collected. Of the collected questionnaires, we removed 34 that were inappropriate for analysis because they did not meet the purpose of the study or included insincere responses.

Frequency statistics were used to analyze the participants’ general characteristics. Reliability analysis, exploratory factor analysis, and confirmatory factor analysis were used to assess the instrument’s reliability and validity. Correlation analysis was used to examine bivariate associations among the main variables. Multiple linear regression analysis was employed to verify the explanatory power and hypotheses of the research model. Multigroup path analysis was performed to verify the effect of moderating variables, and critical ratios for differences were calculated to verify the statistical significance of the path coefficient for each group. Independent samples t tests were applied to analyze differences in perception of the main variables. For frequency analysis and independent samples t tests, the participants were divided into workers and health experts, and the remaining analyses were based on all respondents.

We used the factor extraction method, which implements the most commonly used maximum-likelihood estimation, in common factor analysis; for factor rotation, we used the direct Oblimin method in the square rotation to examine correlations between the factors. The determination criterion for the number of factors extracted was an eigenvalue >1. The Kaiser Meyer Olkin index was used to test the suitability of factors for analysis, where values of 0.5-1.0 indicate that the factors are suitable for analysis and those ≤0.5 indicate that the factors are not suitable for analysis. Pearson correlation analysis was performed to examine bivariate correlations before performing multiple linear regression analysis. Correlation analysis was performed using factor score storage values in the factor analysis. In the social sciences, coefficients below ±0.2 are considered low, those between ±0.3 and ±0.7 are considered moderate, and those above ±0.7 are considered high [103]. Analyses were performed with IBM SPSS Statistics ver. 25.0 and AMOS 22 (IBM Corp, Armonk, NY, USA).

The study involved 866 participants, including 179 health experts and 687 workers who visited 21 workers’ health centers nationwide. Workers’ general characteristics are summarized in Table 2. The majority of the workers were women. Most participants were older than 50 years, followed by those in their 40s and 30s. The duration of employment in the workplace was most commonly 1-4 years, and 66.1% (454/687) of workers were employed in workplaces with <50 employees. Clerical and service-based businesses were more common than production and technical businesses.

Health experts’ characteristics are shown in Table 3. Similar to the workers, the majority of respondents were women. The most common age group was 30-39 years and the duration of employment in the workplace was most commonly 1-4 years. The most common type of occupation was nursing, followed by physical therapy, industrial hygiene safety engineering, counseling psychology, occupational and environmental medicine, and other.

The results of the reliability and exploratory factor analyses are shown in Table 4. Cronbach alpha values greater than .60 and .90 are generally considered acceptable and highly reliable, respectively. Cronbach alpha values for all variables, excluding perceived risk (.69), were within the recommended range (>0.70), and thus the reliability of the main variables was considered acceptable. The analysis was performed without deleting items because none of the items impaired reliability.

The Kaiser Meyer Olkin statistic was 0.90 and the result of the Barlett test was Chi square₂₁₀=14334.09 (P<.001); thus, the factor analysis model was considered suitable. In addition, the cumulative variance was 70.63% and the explanatory power of the six factors was high. All factor loading values were above 0.4, which demonstrated the validity of the overall instrument; therefore, the analysis was performed without additional adjustment.

The fit indices for the research model were as follows: Chi square₁₇₄=819.66 (P<.001), goodness-of-fit index=0.91, root mean square residual=0.04, root mean square error of approximation=0.07, normed fit index=0.94, relative fit index=0.93, incremental fit index=0.96, comparative fit index=0.95, Tucker-Lewis index=0.95, and adjusted goodness-of-fit-index=0.88. The adjusted goodness-of-fit-index did not meet the criteria, but the overall model fit was satisfactory, and the other indices met the criteria. The results of confirmatory factor analysis (Table 5) showed that the paths to the observed variables were significant (P<.001) for all latent variables. The average variance extracted was >0.50 and the construct reliability was >0.70; therefore, convergent validity was demonstrated. In addition, the discriminant validity ensured that the average variance extracted was higher than the square value of the correlation coefficient (Table 6).

The results of Pearson correlation analysis showed that all dependent variables and the independent variable in the research model were significantly correlated (Table 7). In particular, perceived risk was negatively correlated with all other variables.

We performed multiple linear regression analysis to verify the effects of effort expectancy, social influence, performance expectancy, facilitating conditions, and perceived risk on behavioral intention. The factor analysis results showed that the regression model was statistically significant (F=194.96, P<.001), as shown in Table 8. The explanatory power of the regression model was 53.1% (R²=0.53, adjR²=0.53). Moreover, no issues were observed with respect to the independence of residuals (D-W=2.04) or multicollinearity (variance inflation factor<10).

The regression coefficients showed that effort expectancy, social influence, performance expectancy, and facilitating conditions exerted significant positive effects on behavioral intention, whereas perceived risk exerted a significant negative effect on behavioral intention (Table 8). Therefore, the results supported H1, H2, H3, H4, and H5.

We performed multigroup path analysis to verify the moderating effects of gender and age (Table 9). In the male group, effort expectancy, social influence, and performance expectancy exerted significant positive effects on behavioral intention. In the female group, effort expectancy, social influence, and performance expectancy exerted significant positive effects on behavioral intention. Only the performance expectancy (critical ratio=–3.38, P<.001) showed statistically significant differences in the path between males and females.

In the younger group, effort expectancy, social influence, and performance expectancy exerted significant positive effects on behavioral intention. In the older group, effort expectancy, social influence, and performance expectancy exerted significant positive effects on behavioral intention. Only the performance expectancy (critical ratio=1.97) showed statistically significant differences in the path between younger and older respondents. Therefore, the results supported H6, but H7 and H8 were rejected.

The path analysis of workers and health experts showed that effort expectancy, social influence, performance expectancy, and facilitating conditions exerted significant positive effects on behavioral intention, whereas perceived risk exerted a significant negative effect on behavioral intention. Only the perceived risk (critical ratio=–2.24) showed statistically significant differences in the path between workers and health experts.

The results of the independent samples t tests to analyze differences in perception of the main variables between workers and health experts are summarized in Table 10. Workers’ mean scores for the main variables were higher relative to those of health experts for all variables except perceived risk. Moreover, behavioral intention, effort expectancy, social influence, performance expectancy, and perceived risk differed significantly between groups, whereas facilitating conditions did not.