PubMed, Embase, Emcare, Global Health, the Cochrane Library, PsycInfo, and CINHL databases were searched for studies published before December 2021 reporting the implementation of mHealth app for COVID-19 contact tracing. WHO Global Research on COVID-19 database was searched for gray literature.

An example of the search terms used is shown in Box 1, and the full search strategy is available in Supplementary material S1 files. The review was designed in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analysis guidelines and was prospectively registered with PROSPERO (CRD42021249500). PRISMA checklists are available in Supplementary Table S1.

We used the definition of mHealth from the Global Observatory for eHealth (GOe) which defined mHealth or mobile health as medical and public health practice supported by mobile devices, such as mobile phones, tablets, personal digital assistants (PDAs), and other wireless devices (10). mHealth involves the use and capitalization of a mobile phone’s core utility of voice and short messaging service (SMS), as well as more complex functionalities and applications such as general packet radio service (GPRS), third and fourth generation mobile phone telecommunications (3G and 4G system), global positioning system (GPS), and Bluetooth technology (10).

Eligibility was limited to studies where one of the primary objectives was to identify facilitators and barriers to adopting mHealth for COVID-19 contact tracing apps. Eligibility study designs included quantitative studies. We excluded qualitative studies. We only used research articles written in the English language.

As a first step in the data handling process, titles and abstracts of all papers retrieved by the search strategy were screened for relevance, and those clearly irrelevant were discarded. The full text was downloaded if the result was not clearly relevant. As a second step, two authors (SS, AM) independently assess the eligibility of the studies by using the predefined inclusion and exclusion criteria. Any disagreements on whether or not to include a specific study were resolved by discussion between the two authors. The two authors (SS, AM) then independently extracted data from included articles into an Excel template containing the following headings including (1): study design (2); country (3); the nature of mHealth app evaluated (4); facilitators and barriers; and (5) outcomes measures; and (6) the results for each outcome. The reference list of all included articles was hand-searched for additional relevant articles.

The Cochrane Collaboration’s risk of bias tool was used to assess the risk of bias in all studies (RoB 2). Using the Critical Appraisal Skills Program (CASP) quality checklist appropriate to each study design, two authors (SS, AM) independently assess the possibility of bias in the included studies’ design, conduct and analysis. Any discrepancies will be discussed until a consensus is reached. Although no studies were excluded from the review based on quality alone, studies rated as possessing higher methodological strength were synthesized first. Following this, studies appraised as evidencing methodological limitations were then added to the synthesis to investigate whether these lower-quality studies affect the review’s findings. The CASP tools were selected to identify the quality issues across a wide range of study designs.

The facilitators-barriers informed a narrative analysis of mHealth adoption framework synthesis from the unified theory of acceptance and use of technology (11, 12), the privacy calculus framework (13), and uncertainty reduction theory (14) (Figure 1). The framework identified seven key factors that explain mHealth adoption (1): perceived risks to COVID-19 (2); data privacy/security concern (3); trust (4); perceived benefits (5); social norm (6); technology readiness; and (7) social-demographic characteristics.

Based on the privacy calculus model, perceived risk is negatively associated with intention and adoption of using mHealth technology, with perceived benefits positively associated with intention and adoption of using mHealth technology (13). According to uncertainty reduction theory, increasing data privacy/security risks is negatively associated with intention and adoption using mHealth technology, with trust positively associated with intention and adoption using mHealth technology (14). The unified theory of acceptance and use of technology suggests that technology readiness is positively associated with intention and adoption of mHealth health technology. This theory also posits that social norms and socio-demographic characteristics can affect mHealth uptake. Socio-demographic characteristics such as age, gender, income, education, ethnicities, living areas, and marital status determine perceived risks to COVID-19, data privacy/security concern, trust, perceived benefits, social norm, technology readiness.

We found 2,834 entries in our initial search, 99 of which matched our inclusion criteria and were subjected to full-text inspection. As a result, 27 studies were included in the final review from 16 countries representing high income economies (France, German, Italy, United Kingdom, United States, Australia, Singapore, Belgium, Republic Ireland, Netherland, Japan, and Polland), upper middle-income economies (Fiji) and lower-middle income economies (India) (Figure 2).

Detailed characteristics of all included studies are provided in Table 1. Among 27 selected studies, 23 used cross-sectional data either single or repeated and four studies used an experimental design, longitudinal design, conjoint experiment, and discrete choice experiment. Within the cross-sectional design, two studies used cross-country data analyzes, 20 applied single-country analyzes, and one country used repeated cross-sectional data analysis within a city. The number of samples ranged between 288 and 28,246 respondents. Most of the studies use nonprobability sampling methods, such as convenience sampling method, quota sampling method, and purposive sampling method. We identified 10 studies employed certain theories or frameworks including Theory Acceptance Model (TAM), Unified Theory of Acceptance and Use of Technology (UTAUT1 and UTAUT2), Uncertainty Reduction Theory (URT), Health Behavior Models (HBM), Theory Planned Behavior (TPB), Theory of Reasoned Action (TRA), Perceived Behavioral Control (PHC), Hofstede’s Cultural Theory (HCT), Protection Motivation Theory (PMT), and Privacy Calculus Theory (PCT) to develop their hypothesis. There are two final outcomes identified (i.e., app adoption intention and app adoption). Of the 27 studies included in this review, 16 (59.2%) had app adoption as the final outcome, while 11 (40.7%) analyzed the app adoption intention as the final outcome. Those final outcomes may be related to the time of the studies. Most of the studies (7 studies or 63.6%) with app adoption intention as the final outcome were published in 2020, while three quarters (12 studies) having app adoption as the outcome were published in 2021.

The subjective quality assessment of the studies was generally positive (Supplementary Table S2). We did not think any of them were particularly prone to bias or had significant applicability issues. However, for at least one area, we assessed 8 of the 27 studies as ‘high risk’, either because there was insufficient information to assess the category or because we could not be sure what influence the reported technique for that category would have on bias.

The results of each study are summarized in Table 2. We reported based on each factor identified within the facilitators-barriers to adopt the mHealth framework. Facilitators are factors that increase the likelihood of contact tracing apps adoption, while barriers are factors that decrease the likelihood of contact tracing apps adoption. Table 3 shows the respective factors for each of the seven categories elicited from the 27 included articles. Social demographics out to be the most frequent category, while perceived benefits and technologies readiness the least frequent category.

We found 11 studies among 27 selected studies that include perceived risks as one of variables to explain COVID-19 app adoption. Among those 11 studies, 9 reported a significant association of perceived risks to app adoption. In high income economies contexts, Li et al. (2), Nguyen et al. (29), Camacho-Rivera et al. (17), and Hargittai et al. (16) reported that perceived risk to COVID-19 is associated with app adoption in the United States. In Italy, Guazzini et al. (28) reported that risk perception of COVID-19 infection was associated with apps uptake. In the Republic of Ireland, O’Callaghan et al. (41) found that COVID-related anxiety was linked to a desire to install the app. in France, Touzani et al. (40) found that feeling concerned and perceived risk about the pandemic situation is associated with adoption intention. In Poland, Wnuk et al. (23) found acceptance of the COVID-19 contact tracing app was highly correlated with a sense of personal threat and a loss of personal control. In Japan, Shoji et al. (36) found that concerns about health risks determined the mHealth uptake decision. In contrast, Walrave et al. (35) and Oldeweme et al. (31) reported that found no association between COVID-19 concern on apps adoption in German.

Among 27 selected studies, 13 studies included data privacy or security concerns in their estimation, and 12 of them reported a significant association of data privacy and security concerns on COVID-19 contact tracing adoption. In high income economy contexts, Oldeweme et al. (31) reported that adoption is linked to a reduction in privacy risks (i.e., individuals are concerned about data security, such as possible data leaks or misuse by third parties) in Germany. In another study in German, Tomczyk et al. (34) also found a trade-off between mHealth user concerns and perceived security on adoption intentions. In the United States study, Li et al. (2) found that privacy concerns generally had a significant negative effect on adoption, particularly the decision to install the apps. Secondary data use risks were also associated with lower app installation. However, no significant association was discovered between data breach risk and the risk of COVID-19 positive users being re-identified after implementation. In a Belgium study, Walrave et al. (35) found that app-related privacy concern was negatively associated with behavioral intention. In an Italy study, Guazzini et al. (28) found attitudes toward contact tracing systems that capture question such as “I’m afraid of contact tracing systems” and “Contact tracing systems excessively violate privacy” was negatively associated with adoption. In Republic Ireland study, Fox et al. (37) concerns about privacy have a detrimental, albeit little, impact on willingness to trust the app. In the United Kingdom, Panchal et al. (39) reported that privacy concerns were a possible reason why users did not download it. In a United Kingdom study, Dowthwaite et al. (26) also reported that privacy concern was associated with adoption intentions. However, in an experimental study, Horvath et al. (25) found privacy issues did not have as much of an impact on the digital app selection as previously thought or indicated by a study in the United Kingdom. In a study in Australia, Garrett et al. (21) showed that low uptake was linked to concerns about data security (e.g., maintaining privacy). In Netherland, Jonker et al. (24) found a contact tracing app that is both secure and private and has the most realistic features that can achieve a high adoption rate. Wnuk et al. (23) found a potential threat to privacy and other civil rights is associated with less acceptance of contact tracing apps in Poland. In lower-middle income economy contexts, Abuhammad et al. (22) the privacy of the data was a major ethical concern for many of the participants in Jordan.

We documented 11 studies that examined the relationship of trust to adoption. All those studies were in high income economy. Among them, only one study reported a null significant association of trust to adoption. In a study in the United States, Hargittai et al. (16) reported that trust in the medical system, federal gov, local gov, and business leaders was not associated with United States adoption intention. Other studies show a significant association of trust to adoption. In contrast, Altmann et al. (15) and Guazzini et al. (28) found that trust in the government was associated with adoption in France, Germany, Italy, United Kingdom, and United States. Likewise, Nurgalieva et al. (30) in a study involving United States, United Kingdom, and Republic Ireland respondents, documented that trust in current government and trust in politicians in the country where respondents are living, trust in medical doctors and nurses, trust in religious organizations, trust the accuracy of news from political party and leaders, and trust in religious organizations were associated with adoption intention. Dowthwaite et al. (26) found that trust in the NHS COVID-19 app is associated with the adoption of the NHS COVID-19 contact tracing app. Trust in the NHS United Kingdom has won out over privacy concerns, and the NHS centralized app is favored over both the centralized government app and the decentralized system (25). In German and France, Oldeweme et al. (31) and Touzani et al. (40) reported trust in government and trust in political representatives was indirectly associated with apps uptake. Saw et al. (32) reported that confidence in the government was associated with apps uptake in Singapore. In Japan, Shoji et al. (36) found that trust in national were key determinants of mHealth uptake for those aged between 40 and 59 years. A belief that the government was not trustworthy is associated with low uptake in Australia (21).

Ten studies included perceived benefits variables in their study. All of them were in high income economy. All those studies reported a significant association of perceived benefits measures on adoption. In German, Oldeweme et al. (31) found that reduction of performance risks (i.e., individuals are concerned that the product performs as it was designed and advertised) is associated with adoption. Tomczyk et al. (34) also found that perceived utility and hedonic motivation were both highly significant and consistent indicators of adoption intentions, but they were negative predictors of current app usage frequency. In the United States, Nguyen et al. (29) found perceived usefulness and health information orientation were associated with adoption intention. Li et al. (2) found that more prosocial people are significantly more inclined to install and use contact-tracing apps than those less prosocial in the United States. Guazzini et al. (28) found that positive perception of ‘Immuni’ apps was associated with adoption in Italy. Higher perceived usefulness is associated with adoption intention in French (40). Dowthwaite et al. (26) found that positive perception of NHS apps was associated with adoption in the United Kingdom. The early acceptability of citizens is influenced by their beliefs of health benefits in the Republic of Ireland (37). In Belgium, Walrave et al. (35) reported that performance expectancy (i.e., improving individual knowledge about the danger of being infected by COVID-19, which is beneficial for monitoring his/her risk of infection and controlling COVID-19 spread) is linked to behavioral intention. In Australia, Duan and Deng (27) reported that adoption was linked to the perceived value of information disclosure (i.e., when an individual feels that the risk of using an app is lower than the benefits, value given worth the disclosure of information, and benefits can outweigh the risk).

Twelve studies examine the relationship of the social norm to adoption. Among those studies, one study reported a null association. In high income country economy, Oldeweme et al. (31) reported there was no link between lower social risks (i.e., individuals may be concerned about losing status in their social group if they use or do not use the app) and adoption in Germany. In contrast, another study in German by Tomczyk et al. (34) found a very strong connection between injunctive social norms (i.e., referring to perceived societal expectations) and adoption intentions. Walrave et al. (35) reported that social influence was a substantial predictor of adoption in Belgium. Nurgalieva et al. (30) documented that respondents’ perception that their actions to limit coronavirus spread make a difference linked with apps adoption in the United States, United Kingdom, and Republic Ireland. Dowthwaite et al. (26) reported that to help the NHS, protect others, and reduce the spread of the virus are the main reason to adopt NHS COVID-19 apps. in Italy, Guazzini et al. (28) found that individuals who know important others who have downloaded and installed the Immuni app are more likely to download and utilize the app themselves. In the United States, Li et al. (2) reported that ‘prosocialness’ is associated with contact tracing app acceptance. The views of social influence determine citizens’ initial acceptance of mHealth app in the Republic of Ireland (37). In Singapore, Saw et al. (32) found that using hand sanitizers, avoiding public transportation, and preferring outdoor over indoor settings are all linked to app use. In Japan, Shoji et al. (36) reported that concern about social risk was the key determinant of mHealth uptake for a senior adult. However, attachment to the community was the main driver to uptake the app for the middle-age. In Australia, Duan and Deng (27) documented that social influence as measured by reasons of individual used apps (i.e., recommended by the influential people, recommended by the important people, and recommended by the trusted people) are associated with adoption. In upper middle income economy contexts, Sharma et al. (18) reported that social normative pressures perceived by individuals were associated with app uptake in Fiji. Accordingly, they also found that in cultures with a high level of collectivism, the link between privacy concerns and attitude is weaker.

Ten studies examined technology readiness in their study. Among those studies, we found 6 studies that reported a significant association of technology readiness on adoption. Using Parasuraman and Colby, Li et al. (2) documented that technology readiness was associated with app adoption in the United States. Hargittai et al. (16) reported that internet skill is associated with COVID-19 app contact tracing in the United States. However, in another study in the United States, Nguyen et al. (29) reported that perceived ease of use was not associated with app adoption in the United States. In the United Kingdom, Panchal et al. (39) found that little understanding regarding the app’s functionality is associated with less intention to uptake. Walrave et al. (35) reported that individual innovativeness and facilitating conditions were associated with using the app in Belgium. Tomczyk et al. (34) also found that perceived barriers, experience, and ease of use were not significant to COVID-19 app adoption in German. Likewise, in German, Blom et al. (19) reported potential spreaders have a high level of access to install the app, as well as a high level of competence to do so, but a low willingness to correctly adopt the app. In Australia, Duan and Deng (27) found that effort expectancy was measured by individual perception of easy to use, easy to become skillful, interaction is clear and understandable, and not much effort involved was associated with adoption. In Singapore, Huang et al. (38) documented that the app was not adopted by older persons who did not have cellphones, and those who did may have had difficulty installing and utilizing the program. In upper middle income economy contexts, Sharma et al. (18) found apps adoption was associated with habit describes the prior experience and individuals’ belief in their ability to protect their information privacy.

Among 16 selected studies, 15 studies include socio-demographic variables. In the context of high-income economy, Li et al. (2) documented females, lower income, lower education, and an older individual had lower intentions to adopt the app than males, higher income, higher education, and young individual. The significant association of race mostly appeared on the intentions to keep the app installed rather than the intentions to install the app, with Hispanics having much greater intentions to install than other races (17). Lower financial deprivation was associated with lower uptake in France (40). Walrave et al. (35) reported no association of health condition, age, gender, and education on behavioral intention. Saw et al. (32) found age-associated with apps uptake. However, gender, citizenship, household type and size, education, ethnicity, marital status were not significantly associated with app uptake in Singapore. Hargittai et al. (16) found that age and education are associated with adoption in the United States. However, female, urban, and household income are not. Altmann et al. (15) found young aged 18–30 more likely to uptake the apps than age 17 or lower in their cross-country data analyzes. A null association was found on the relationship between gender and app uptake. Males are more likely to say they will not install the app, either probably or definitely (41). The oldest and youngest generations are the most likely to say they will probably or definitely install the app (41). However, another study in Republic Ireland showed that age, gender, education, and health status all showed no correlations (37). The app was less likely to be downloaded and used by older persons in Singapore (38). Accordingly, Shoji et al. (36) found that age, female, and marriage status were not related to uptake in Japan. In Poland, Wnuk et al. (23) found male and older people are less likely accept contact tracing apps. Oldeweme et al. (31) and Blom et al. (19) reported that older individuals had lower intentions to use the apps. No associations were shown for education and gender. In the context lower middle income economy, Sharma et al. (33) reported the absence of an ASA installation was linked to being over 40 years old, being female, having fewer years of schooling, and having a lower per capita income (≤5,000 Indian National Rupee) in New Delhi India. In the context of upper middle income economy, acceptance and use of tracking technologies among Jordanians were predicted by income and living area in Jordan (22).