Using the UTAUT model to understand HIV-negative MSM’s intention to use a proposed HIV prevention mHealth management system: A cross-sectional study

Abstract

Background

Men who have sex with men (MSM) are key to China’s HIV prevention. In China, the proportion of HIV transmission attributed to homosexual sexual behavior increased from 9.1% in 2009 to 23.3% in 2020, highlighting the growing importance of HIV prevention among MSM. Mobile health (mHealth) interventions offer distinct advantages: lowering barriers to service access, mitigating social stigma, and improving accessibility of HIV-related interventions. However, evidence remains limited regarding HIV-negative MSM’s pre-implementation intention to use HIV prevention mHealth management systems.

Objective

Guided by the Unified Theory of Acceptance and Use of Technology (UTAUT), this study examined factors associated with HIV-negative MSM’s behavioral intention to use a proposed HIV prevention mHealth management system.

Methods

A cross-sectional survey was conducted in Chongqing, China, from 2024 to 2025. A total of 469 HIV-negative MSM were recruited through NGO and VCT-related channels. Participants evaluated a proposed system concept introduced through a standardized description; the system had not yet been deployed or user-tested. Structured questionnaires assessed demographic characteristics and UTAUT constructs. Descriptive analyses were performed using R, and structural equation modeling was conducted using Mplus.

Results

The final SEM showed acceptable model fit (χ²/df = 3.798, RMSEA = 0.076, CFI = 0.959, SRMR = 0.047). Performance expectancy (β = 0.431, P < 0.001) and social influence (β = 0.412, P < 0.001) were positively associated with behavioral intention, while effort expectancy (β = 0.039, P = 0.700) and facilitating conditions (β = 0.157, P = 0.106) were not.

Conclusions

Pre-implementation intention to use the proposed HIV prevention mHealth management system was mainly associated with perceived system effectiveness and social influence. Future HIV-related mHealth interventions should emphasize system effectiveness and leverage peer/community support to enhance adoption.

Introduction

According to the latest report by the Joint United Nations Programme on HIV/AIDS (UNAIDS), approximately 40.8 million people worldwide were living with HIV as of 2024, with around 1.3 million new infections reported that year. ¹ Men who have sex with men (MSM) are at high risk of HIV infection ² due to behavioral patterns such as multiple sexual partners, unprotected sex, occasional encounters, and drug use. ³

In China, the epidemiological characteristics of HIV have changed substantially. The proportion of HIV transmission attributed to homosexual sexual behavior increased from 9.1% in 2009 to 23.3% in 2020, ⁴ making MSM a key population in current HIV prevention and control efforts. Influenced by traditional cultural concepts and social structures, many MSM face identity concealment, social stigma, and family pressure. Some also maintain heterosexual marital relationships,^5,6 which may increase the complexity and concealment of HIV transmission.

Influenced by traditional cultural norms and social structures, many MSM in China may experience identity concealment, social stigma, and family pressure. Some MSM also report heterosexual marriage or heterosexual sexual partnerships, which may increase the complexity and concealment of HIV prevention needs. HIV-related risks among MSM are shaped by multiple factors, including diverse sexual networks, partner communication, access to prevention services, HIV testing, condom use, and stigma-related barriers. Recent epidemiological evidence further underscores the continuing importance of HIV prevention among MSM in China. For example, homosexual transmission accounted for 25.7% of newly reported HIV infections in China in 2023 ⁷ ; a recent large-scale web-based survey found that only 45.2% of HIV-negative or status-unknown MSM were fully aware of their sexual partners’ HIV status. ⁷ In addition, recent multi-city evidence among HIV-negative high-risk MSM in China has highlighted persistent gaps in PrEP awareness, willingness, and engagement. ⁸ These findings suggest that HIV-negative MSM remain an important population for early HIV prevention and therefore require confidential, accessible, and digitally supported interventions.

With the development of mobile internet technology, mobile health (mHealth) interventions have become an important supplementary approach to HIV prevention and control. They have been applied in health education, risk assessment, HIV testing promotion, and related health management services. Previous studies have shown ⁹ that mHealth interventions can improve HIV-related health behaviors, including promoting testing, increasing risk awareness, and enhancing compliance with health management. In China, the rapid expansion of mobile internet use provides a practical foundation for implementing mHealth-based HIV interventions.

For MSM, who often experience identity concealment and limited access to offline health services, mHealth interventions may offer distinct advantages. Mobile platforms can provide continuous and personalized health information without requiring direct identity disclosure, thereby reducing service avoidance related to discrimination or shame. They can also be integrated into users’ daily lives with relatively low additional learning costs. However, the effectiveness of these interventions depends largely on the target population’s acceptance and willingness to continue using them.

Although mHealth interventions have garnered considerable attention in HIV prevention and control in recent years, related studies have provided substantial evidence in promoting testing behaviors, ¹⁰ increasing risk awareness, ⁹ and improving treatment compliance. ¹¹ However, significant gaps remain in the current research. Firstly, most studies have primarily focused on HIV-infected individuals, while research targeting the key “window period” population of HIV-negative MSM ¹² is relatively scarce. Secondly, existing literature mainly evaluates the effectiveness of mHealth interventions, with insufficient attention paid to the behavioral motivations behind individuals’ willingness to use and continue using these systems, as well as the influencing mechanisms.^9,13

Therefore, understanding HIV-negative MSM’s willingness to use the HIV prevention mHealth management system and identifying the factors influencing this willingness are important for improving the acceptability and effectiveness of such interventions. The Unified Theory of Acceptance and Use of Technology (UTAUT) integrates multiple technology acceptance models ¹⁴ and explains technology adoption through constructs such as performance expectancy, effort expectancy, social influence, and facilitating conditions. In the present study, the original UTAUT model was selected rather than UTAUT2 or other extended models because the study focused on initial behavioral intention toward an HIV prevention mHealth management system, rather than actual use, continued use, or post-adoption behavior. Several UTAUT2 constructs, such as habit, ¹⁵ hedonic motivation, and price value, were less directly applicable because participants had not yet developed real-world usage experience with the system and because the system was framed as an HIV prevention and health management service rather than a consumer-oriented commercial technology. Specifically, habit generally assumes repeated prior use, which could not be established in a pre-implementation survey. Price value requires users to evaluate perceived benefits against monetary or other costs, whereas no concrete user-cost information was provided in the present system concept. Hedonic motivation may also be less central in this context because HIV prevention and health management are primarily risk-reduction and utility-oriented activities. Therefore, including these constructs could have increased speculative responses and diluted the focus on the core pre-adoption determinants examined in this study. At the same time, the use of mHealth systems among MSM may be shaped by both functional perceptions and social-contextual factors, making the original UTAUT model an appropriate and parsimonious framework for examining the core determinants of behavioral intention in this population. However, we recognize that MSM populations involve important social and behavioral complexities, including privacy concerns, stigma, trust, confidentiality, and community support. Therefore, although this study adopted the original UTAUT model to maintain theoretical parsimony, future studies should consider extended UTAUT models incorporating MSM-specific psychosocial variables. Accordingly, this study applied the original UTAUT model to examine behavioral intention to use an HIV prevention mHealth management system among HIV-negative MSM in China and to identify its associated factors.

Methods

Participants

This study was conducted in Chongqing, China, from 2024 to 2025, using a cross-sectional design. It was led and organized by the Disease Prevention and Control Center of Jiulongpo District, Chongqing, in collaboration with non-governmental organizations (NGOs) and voluntary counseling and testing (VCT) clinics. Participants were recruited through non-probability sampling. The study population consisted of MSM. Inclusion criteria were: (1) age ≥ 18 years; (2) having engaged in same-sex anal or oral sexual activity within the past year; (3) HIV-negative status confirmed by HIV antibody testing conducted at local VCT clinics before questionnaire administration; (4) no severe physical or mental disorders, the ability to understand the survey content, voluntary participation, and signing of written informed consent before participation. Exclusion criteria included individuals with incomplete questionnaires or responses showing significant inconsistencies. After eligibility was confirmed, participants completed the questionnaire through either online or offline methods, resulting in 469 valid responses. Because recruitment was conducted through offline, service-based NGO outreach and VCT-related channels rather than through an open web-based survey or a fixed sampling frame, the total number of individuals approached or invited was not predefined. Only completed questionnaires from eligible and consenting participants were recorded; refusals and non-returned incomplete questionnaires were not systematically documented. Therefore, a precise sampling denominator, response rate, and refusal rate could not be calculated.

Model construction

This study constructed the research model based on the original UTAUT framework. Proposed by Venkatesh et al., ¹⁶ UTAUT explains individuals’ acceptance and use of new technologies through four core determinants: performance expectancy, effort expectancy, social influence, and facilitating conditions. The original UTAUT model was used because this study examined initial behavioral intention toward an HIV prevention mHealth management system, rather than actual usage behavior or continued post-adoption use. Although UTAUT2 and other extended models provide additional constructs, such as hedonic motivation, price value, and habit, these constructs were less suitable for the present context. Participants evaluated an HIV prevention and health management system without long-term real-world use experience, and the system was conceptualized as a public health-oriented service rather than a commercial consumer technology. Therefore, the original UTAUT model provided a theoretically appropriate and parsimonious framework for examining the core determinants of behavioral intention among HIV-negative MSM. At the same time, given the social and behavioral complexity of MSM populations, future research may extend this framework by incorporating context-specific psychosocial factors, such as privacy concerns, stigma, trust, confidentiality, and community support.

According to the UTAUT theoretical framework (Figure 1), an individual’s behavioral intention to use a technology is primarily influenced by four core factors: performance expectancy, effort expectancy, social influence, and facilitating conditions. Based on the context of HIV health interventions for the MSM population, the following definitions are provided for each latent variable: Performance Expectancy: MSM’s perception of whether using the mHealth management system can enhance the efficiency of obtaining HIV-related health information and the effectiveness of health management. Effort Expectancy: MSM’s perception of the ease of system use and the level of effort required. Social Influence: The extent to which MSM perceive recognition and support from significant others (e.g., peers, medical staff, or social organizations) regarding their use of mHealth systems. Facilitating Conditions: MSM’s assessment of whether they possess the necessary technical resources, skills, and supportive environment to use the system. Behavioral Intention: MSM’s subjective willingness and tendency to use mHealth systems for HIV prevention and health management in the future. Based on this theoretical framework and existing empirical evidence, the following research hypotheses are proposed:

Figure 1.

UTAUT model schematic diagram.

H1

Performance Expectancy has a significant positive effect on Behavioral Intention.

H2

Effort Expectancy has a significant positive effect on Behavioral Intention.

H3

Social Influence has a significant positive effect on Behavioral Intention.

H4

Facilitating Conditions have a significant positive effect on Behavioral Intention.

Survey instrument and measurement

Based on the UTAUT framework and a review of relevant domestic and international literature, a structured questionnaire was developed to measure the five latent variables in the research model. The UTAUT-related items were adapted from previously published studies rather than adopted as an intact existing instrument, and the wording was modified to fit the context of the HIV prevention mHealth management system among MSM. To improve transparency, we added an item-level source mapping table as Supplemental Table S1, linking each construct and item to the corresponding published UTAUT-based or mHealth acceptance scales from which the item wording was adapted. The items retained the core wording of the original UTAUT constructs to preserve theoretical comparability, while the system description and item context were adapted to the HIV prevention mHealth setting among MSM. Thus, MSM-specificity was reflected primarily through the standardized introduction of the proposed HIV prevention mHealth system and its intended functions, rather than through identity-specific wording in every item. No formal cognitive interviewing or independent translation validation was conducted before the main survey. However, the questionnaire was reviewed by the research team and local public health staff familiar with HIV prevention services for MSM to ensure clarity, contextual appropriateness, and consistency with the standardized system description.

Although the items were adapted from published studies and evaluated for reliability and validity in the present sample, the complete questionnaire was not a formally validated instrument specifically developed for HIV-negative MSM in China. The questionnaire used in this study has been provided as a Supplementary File. The questionnaire consists of two sections: the first section collects demographic information of the participants, and the second section constitutes the main scale, designed to measure latent variables including Performance Expectancy, Effort Expectancy, Social Influence, Facilitating Conditions, and Behavioral Intention (Table 1). Each item is rated on a 5-point Likert scale, ranging from “comparatively disapproved” to “very approved,” scored from 1 to 5. Before completing the questionnaire, all participants received a standardized explanation of the HIV prevention mHealth management system, including its intended use in HIV prevention and health management, HIV-related health education, access to prevention services, risk reduction, and continuous health management support. Participants were informed that all items referred specifically to this HIV-related mHealth management context, rather than to a general mobile system. Although the items retained relatively general UTAUT wording, they were administered after this standardized explanation and were explicitly framed within the context of HIV prevention and health management among MSM. Thus, the MSM-specific context was reflected through the study population, the standardized system description, and the system’s intended functions, including HIV prevention knowledge, risk reduction, access to intervention services, and continuous health management support.

Table 1.

Questionnaire.

Factor	Question
Performance Expectancy17–21	PE1. The intelligent management system can help me better learn about AIDS prevention.
	PE2. The intelligent management system allows me to access intervention services more conveniently
	PE3. The intelligent management system can help me value my health more.
	PE4. The intelligent management system can help me reduce the risk of infection.
Effort Expectancy18–20	EE1. Learning to use this intelligent management system is easy.
	EE2. Using this intelligent management system is easy.
	EE3. My interaction with the system will be clear and understandable.
Social Influence18–20	SI1. Friends around me think I should use this intelligent management system.
	SI2. People close to me think I should use this intelligent management system.
Facilitating Conditions19,21,22	FC1. I have the resources needed to use this intelligent management system (e.g., mobile phone, internet, WeChat, etc.).
	FC2. I have the relevant knowledge to use mobile phones, the internet, and WeChat official accounts.
	FC3. I do not need special training to use this intelligent management system.
	FC4. When I have difficulties using the intelligent management system, I can get help from others.
Behavioral Intention19,22,23	BI1. I am willing to use or continue using this intelligent management system.
	BI2. I would recommend this system to others.

Note. In the original questionnaire, “the intelligent management system” or “the system” referred to the HIV prevention mHealth management system. Before completing the questionnaire, participants received a standardized explanation of the system’s HIV prevention and health management objectives, functions, and expected use scenarios.

Reliability analysis indicated that, except for Behavioral Intention, the Cronbach’s α coefficients and composite reliability (CR) of the remaining latent variables exceeded 0.70, demonstrating good internal consistency. For Behavioral Intention, Cronbach’s α was 0.679, and CR was 0.638, both slightly below the conventional 0.70 threshold but above the 0.60 minimum criterion commonly considered acceptable in exploratory research. The AVE value was 0.471, slightly below the recommended 0.50 threshold for convergent validity. Given that BI was measured using only two items, we additionally examined the inter-item correlation (r = 0.462) and Spearman-Brown coefficient (0.632), as reliability assessment for two-item scales should not rely solely on Cronbach’s α. ²⁴ The Behavioral Intention construct was retained because it is theoretically central to the UTAUT framework, both BI items showed statistically significant standardized factor loadings, and its reliability indices met the minimum criteria for exploratory analysis. Nevertheless, the slightly lower AVE value was reported transparently and considered as a measurement limitation. Validity assessment showed that the overall KMO value was 0.837, and Bartlett’s test of sphericity was significant (P< 0.001), indicating that the data were suitable for factor analysis. Confirmatory factor analysis revealed that the average variance extracted (AVE) for each latent variable was all greater than or close to 0.50, suggesting good convergent validity of the scale.

The Behavioral Intention construct was retained because it is theoretically central to the UTAUT framework, both BI items showed statistically significant standardized factor loadings, and its reliability indices met the minimum criteria for exploratory analysis. Nevertheless, the slightly lower AVE value was reported transparently and considered as a measurement limitation.

Sample size estimation

The minimum required sample size for structural equation modeling (SEM) was calculated based on the formula $N=\left(\frac{L}{\mathrm{\lambda }}\right)+1$ , where N represents the minimum sample size, L is the number of observed variables, and λ is the anticipated factor loading. In this study, the measurement and structural models together contained 26 free parameters, resulting in an estimated minimum sample size of approximately 260 participants. The actual sample size was 469, which satisfied the statistical requirements for SEM fitting and path analysis.

Statistical analysis

Descriptive statistics and correlation analyses were conducted using R software, with categorical variables summarized as frequencies and percentages. Prior to confirmatory factor analysis and structural equation modeling, the distributional properties of the observed variables were assessed. The absolute values of skewness were below 2 and the absolute values of kurtosis were below 7 for all measurement items, indicating no severe univariate non-normality. Therefore, confirmatory factor analysis and structural equation modeling were conducted in Mplus using maximum likelihood estimation. Model adequacy was evaluated using commonly accepted fit indices and criteria: root mean square error of approximation (RMSEA) < 0.08, standardized root mean square residual (SRMR) < 0.08, comparative fit index (CFI) > 0.90, and Tucker-Lewis index (TLI) > 0.90.

Model refinement was conducted using a theory-guided and stepwise strategy rather than relying solely on data-driven modification. The initial measurement model included all 15 theoretically specified UTAUT items. During confirmatory factor analysis, standardized factor loadings, statistical significance, modification indices, and item content were examined jointly. Items were considered for removal only when they showed inadequate psychometric performance, such as a standardized factor loading below 0.60 or a non-significant loading, and when their removal was theoretically justifiable and did not compromise the conceptual coverage of the corresponding construct. Modification indices were used only as supplementary diagnostic information to identify potential sources of local misfit. No residual covariance, cross-loading, or additional structural path was added solely on the basis of modification indices. After each modification, the model was re-estimated, and model fit, reliability, convergent validity, and construct meaning were re-evaluated.

To examine whether demographic characteristics moderated the relationships between UTAUT constructs and behavioral intention, exploratory multi-group analyses were conducted using Mplus. Multi-group SEM was performed by age group, educational level, and marital status. For age, participants were divided into a younger group (≤40 years) and an older group (>40 years). For educational level, participants were divided into a lower-education group (high school or below) and a higher-education group (college or above). For marital status, the original three categories (unmarried, married, divorced/widowed) were collapsed into two groups, not married (including unmarried, divorced, and widowed) and married, due to the small sample size of the divorced/widowed group, which caused estimation instability in the three-group mode. Differences in structural paths across groups were assessed using Wald tests of parameter constraints.

Results

Demographic characteristics of MSM participants

The demographic characteristics of the participants are summarized in Table 2. The majority of respondents (92.32%) were of Han ethnicity. Over half (68.02%) had an education level below high school. A total of 277 (59.06%) were unmarried, while 115 (24.52%) were married. Additionally, 73 participants (15.56%) reported a monthly disposable income of no more than US$740 (RMB 5,000), while 73 participants (15.56%) reported a monthly disposable income exceeding US$1,480 (RMB 10,000). The average age of the participants was 42 years, with a maximum age of 82 and a minimum age of 19. The sample showed a relatively older age distribution, with 121 participants (25.80%) aged 41-50 years and 137 participants (29.21%) aged over 50 years.

Table 2.

Demographic characteristics of the group.

Variables	Category	Count (%)
Ethnicity	Han	433 (92.32)
	Other	36 (7.68)
Educational Level	High school or below	319 (68.02)
	College or above	150 (31.98)
Marital Status	Unmarried	277 (59.06)
	Married	115 (24.52)
	Divorced/Widowed	77 (16.42)
Monthly Disposable Income	≤US$740 (≤RMB 5000)	73 (15.56)
	US$741-148 (RMB 5001-10000)	323 (68.88)
	>US$1480 (>RMB 10000)	73 (15.56)
Age	18-30	105 (22.39)
	30-40	106 (22.60)
	41-50	121 (25.80)
	>50	137 (29.21)

Note. Currency conversion was based on an approximate exchange rate of US$1 =RMB 6.78.

Correlations among latent variables

Regarding the correlations between latent variables (Table 3), significant correlations were found between performance expectancy and effort expectancy (r= 0.619, P<0.001), social influence (r=0.158, P= 0.002), and facilitating conditions (r= 0.471, P<0.001). A significant correlation was also observed between effort expectancy and social influence (r=0.141, P=0.016), as well as between effort expectancy and facilitating conditions (r=0.741, P<0.001). However, the correlation between facilitating conditions and social influence was not significant (r = 0.050, P = 0.414).

Table 3.

Correlation between behavioral intention and its factors.

​	Performance expectancy	Effort expectancy	Social influence	Facilitating conditions	Behavioral intention
Performance Expectancy	1	​	​	​	​
Effort Expectancy	0.619***	1	​	​	​
Social Influence	0.158**	0.141**	1	​	​
Facilitating Conditions	0.471***	0.741***	0.050	1	​
Behavioral Intention	0.594***	0.480***	0.493***	0.409***	1

Note. ***P < 0.001, **P < 0.05.

Measurement model

Based on UTAUT theory, an initial measurement model comprising five latent variables and 15 observed variables was first constructed and tested. The initial model showed inadequate fit: χ²/df = 6.441, RMSEA = 0.106, CFI = 0.878, TLI = 0.839, and SRMR = 0.095. Therefore, the model was refined following the theory-guided strategy described in the Statistical Analysis section. Three items were removed from the initial measurement model. PE3 was removed because its standardized factor loading was below the recommended threshold of 0.60 (λ = 0.392), suggesting that it did not adequately represent the Performance Expectancy construct in this sample. Conceptually, PE3, “The intelligent management system can help me value my health more,” reflected general health awareness rather than the immediate functional utility of the HIV prevention mHealth management system. PE4 was removed because it had a very low and non-significant loading (λ = 0.093, P = 0.063). Its content, “The intelligent management system can help me reduce the risk of infection,” may represent a more distal health outcome that depends not only on system performance but also on users’ subsequent preventive behaviors. FC4 was removed because its standardized factor loading was slightly below the 0.60 threshold (λ = 0.577). Its content, “When I have difficulties using the intelligent management system, I can get help from others,” overlapped with interpersonal help or social support rather than directly reflecting participants’ own technical resources and facilitating conditions.

Importantly, modification indices were used only as supplementary diagnostic information. No item was removed solely to improve global model fit, and no post hoc residual covariance, cross-loading, or additional structural path was added without theoretical justification. The final measurement model retained 12 items and preserved the conceptual meaning of all five UTAUT constructs. Compared with the initial model, the final model showed substantially improved and acceptable fit: χ²/df = 3.798, RMSEA = 0.076, CFI = 0.959, TLI = 0.939, and SRMR = 0.047. To provide a transparent comparison of the model refinement process, the fit indices of the initial and final measurement models are summarized in Table 4. Standardized factor loadings for all retained items exceeded 0.60 and were statistically significant (P<0.001). The specific indicators are presented in Table 5. CR for each latent variable ranged from 0.638 to 0.882, and AVE ranged from 0.471 to 0.791 (Table 6). Most constructs demonstrated acceptable internal consistency and convergent validity. However, the CR (0.638) and AVE (0.471) values for Behavioral Intention were slightly below conventional thresholds. Because the CR value remained above 0.60, the reliability of Behavioral Intention was considered marginal rather than unacceptable, but its convergent validity was still interpreted with caution. ²⁵ Discriminant validity was assessed using the Fornell-Larcker criterion. As shown in Table 7, the square root of the AVE for each construct was greater than its correlations with other constructs, supporting acceptable discriminant validity and indicating that the latent constructs were empirically distinct from one another.

Table 4.

Comparison of initial and final measurement models.

Model	χ2/df	RMSEA	CFI	TLI	SRMR
Initial measurement model	6.441	0.106	0.878	0.839	0.095
Final measurement model	3.798	0.076	0.959	0.939	0.047
Recommended criteria	<5.0	<0.08	>0.90	>0.90	<0.08

Table 5.

Measure model.

Dimension	Item	Standardized factor loading	Standard error	Critical ratio	P
Performance Expectancy	PE1	0.889	0.059	16.713	-
	PE2	0.830	-	-	<0.001
Effort Expectancy	EE1	0.815	0.050	22.258	-
	EE2	0.878	0.055	19.311	<0.001
	EE3	0.816	-	-	<0.001
Social Influence	SI1	0.967	0.091	9.147	-
	SI2	0.804	-	-	<0.001
Facilitating Conditions	FC1	0.827	0.060	18.839	-
	FC2	0.826	0.079	16.367	<0.001
	FC3	0.735	-	-	<0.001
Behavioral Intention	BI1	0.620	0.150	9.392	-
	BI2	0.746	-	-	<0.001

Table 6.

CR and AVE indicators for the factors.

Factors	CR	AVE
Performance Expectancy	0.850	0.740
Effort Expectancy	0.875	0.700
Social Influence	0.882	0.791
Facilitating Conditions	0.839	0.635
Behavioral Intention	0.638	0.471

Table 7.

Discriminant validity assessment.

Construct	PE	EE	SI	FC	BI
Performance Expectancy	0.860	​	​	​	​
Effort Expectancy	0.619	0.837	​	​	​
Social Influence	0.158	0.141	0.889	​	​
Facilitating Conditions	0.471	0.741	0.050	0.797	​
Behavioral Intention	0.594	0.480	0.493	0.409	0.686

Note. Diagonal values are the square roots of AVE. Off-diagonal values are inter-construct correlations. PE = Performance Expectancy; EE = Effort Expectancy; SI = Social Influence; FC = Facilitating Conditions; BI = Behavioral Intention.

Structural equation model

Based on the measurement model, a structural model was further constructed to examine the Behavioral Intention of MSM to use the HIV prevention mHealth management system. The structural model demonstrated good overall fit, with χ²/df = 3.798, RMSEA = 0.076, CFI = 0.959, TLI = 0.939, and SRMR = 0.047. The model explained 54.2% of the variance in Behavioral Intention (R² = 0.542), indicating moderate-to-substantial explanatory power for MSM participants’ intention to use the HIV prevention mHealth management system.

The results of the structural model path analysis (Figure 2, Table 8) indicate that performance expectancy positively and significantly influences behavioral intention (β = 0.431, P < 0.001), and social influence also has a significant positive effect on behavioral intention (β = 0.412, P < 0.001), supporting hypotheses H1 and H3. In contrast, the effect of effort expectancy on behavioral intention was not significant (β = 0.039, P = 0.700), and the direct effect of facilitating conditions on behavioral intention did not reach statistical significance (β = 0.157, P = 0.106), indicating that hypotheses H2 and H4 were not supported.

Figure 2.

Structural equation model diagram. Note. Black solid lines indicate significant paths (***, P < 0.001), while black dashed lines indicate non-significant paths. Standardized path coefficients are shown, with standard errors in parentheses. PE = Performance Expectancy; EE = Effort Expectancy; SI = Social Influence; FC = Facilitating Conditions; BI = Behavioral Intention. R² for Behavioral Intention = 0.542. ***P < 0.001; **P < 0.01; *P < 0.05.

Table 8.

Estimation of the modified model.

Path	Path coefficient	Standardized path coefficient	t	P
Performance Expectancy	0.275	0.431	5.273	<0.001
→Behavioral Intention
Effort Expectancy	0.027	0.039	0.385	0.700
→Behavioral Intention
Social Influence	0.196	0.412	6.106	<0.001
→Behavioral Intention
Facilitating Conditions	0.118	0.157	1.616	0.106
→Behavioral Intention

Multi-group analysis by demographic characteristics

Measurement invariance was tested across age, education level, and marital status using a three-step procedure (configural → metric → scalar). As shown in Table 9, all three grouping variables achieved full scalar invariance, with ΔCFI values ranging from –0.004 to +0.001, all within the recommended threshold of ≥ -0.01. ²⁶ These results confirm that the measurement model is equivalent across groups. Subsequently, multi-group analyses were conducted to examine the moderating roles of these variables (Table 10). For age, participants were divided into a younger group (≤40 years, n = 211) and an older group (>40 years, n = 258), based on the original age categories and sample distribution. The Wald test showed that structural paths did not differ significantly between the two age groups (χ² = 2.972, df = 4, P = 0.563), indicating that age did not significantly moderate the relationships between UTAUT constructs and behavioral intention.

Table 9.

Measurement invariance testing across demographic groups.

Group	Model	χ2/df	RMSEA	CFI	TLI	SRMR	ΔCFI
Age	Configural	2.361	0.076	0.953	0.940	0.077	​
	Metric	2.419	0.078	0.955	0.937	0.072	0.002
	Scalar	2.375	0.077	0.951	0.939	0.080	-0.004
Education	Configural	2.165	0.070	0.959	0.947	0.064	​
	Metric	1.224	0.072	0.960	0.944	0.060	0.001
	Scalar	2.163	0.070	0.957	0.947	0.076	-0.003
Marital Status	Configural	2.301	0.074	0.955	0.942	0.070	​
	Metric	2.403	0.077	0.955	0.937	0.068	0
	Scalar	2.225	0.072	0.956	0.945	0.071	0.001

Table 10.

Multi-group analysis of demographic moderators.

Moderator	Group	PE→BI	EE→BI	SI→BI	FC→BI
		β (P)	β (P)	β (P)	β (P)
Age	≤40 years	0.562 (0.045)	0.291 (0.477)	0.254 (0.169)	-0.03 (0.908)
	>40 years	0.550 (<0.001)	-0.224 (0.109)	0.365 (<0.001)	0.226 (0.060)
Marital status	Not married	0.454 (<0.001)	-0.081 (0.597)	0.424 (<0.001)	0.243 (0.075)
	Married	0.422 (<0.001)	0.055 (0.740)	0.409 (<0.001)	0.145 (0.378)
Educational level	High school or below	0.530 (<0.001)	-0.047 (0.796)	0.141 (0.482)	0.139 (0.411)
	College or above	0.517 (<0.001)	-0.128 (0.404)	0.421 (<0.001)	0.307 (0.029)

Note. PE = Performance Expectancy; EE = Effort Expectancy; SI = Social Influence; FC = Facilitating Conditions; BI = Behavioral Intention. β values are standardized path coefficients.

For marital status, participants were divided into not married (including unmarried, divorced, and widowed) and married groups. The Wald test showed no significant differences in structural paths across marital-status groups (χ² = 0.765, df = 4, P = 0.943), suggesting that marital status did not significantly moderate the UTAUT relationships in the present model.

For educational level, participants were divided into a lower-education group (high school or below) and a higher-education group (college or above). The Wald test indicated significant differences in structural paths across educational groups (χ² = 12.911, df = 4, P = 0.012), suggesting that educational level significantly moderated the relationships between UTAUT constructs and behavioral intention. Specifically, in the lower-education group, only performance expectancy was significantly associated with behavioral intention (β = 0.530, P < 0.001). In the higher-education group, performance expectancy (β = 0.517, P < 0.001), social influence (β = 0.421, P < 0.001), and facilitating conditions (β = 0.307, P = 0.029) were significantly associated with behavioral intention. Effort expectancy was not significant in either educational group.

Discussion

This study, grounded in the UTAUT model, examined the behavioral intention of Chinese HIV-negative MSM to use an HIV prevention mHealth management system and the factors influencing it. The results indicate that performance expectancy and social influence have significant positive effects on behavioral intention, whereas effort expectancy and facilitating conditions are not statistically significant.

The results of this study indicate that performance expectancy is one of the most important factors influencing MSM users’ Behavioral Intention to use the HIV prevention mHealth management system (β = 0.431, P < 0.001). When users perceive that the system can effectively enhance the efficiency of HIV-related health information acquisition and improve the effectiveness of health management, their behavioral intention significantly increases. This finding is consistent with previous mHealth studies among MSM populations. For example, Mi and Lu et al. reported that perceived health benefits were the core motivators driving the use of HIV-related mobile intervention tools among both adolescent and adult MSM groups.^27,28 Similarly, Zhang et al.’s research on mHealth use among MSM reached comparable conclusions. ²⁹ Hoque et al. further demonstrated in a study on elderly adoption of health information technology that performance expectancy is a significant predictor of behavioral intention.^18,30 Collectively, these findings indicate that performance expectancy has strong explanatory power across different age groups, cultural contexts, and types of health interventions.

In the MSM population, the effect of performance expectancy may also be partly shaped by privacy- and confidentiality-related considerations. Previous research among Malaysian MSM showed that privacy and confidentiality concerns are important factors affecting the acceptability and uptake of HIV-related mHealth apps, and that such platforms need to be perceived as safe and secure to ensure public trust and engagement. ³¹ Similarly, formative research on HIV prevention chatbots among MSM indicated that users’ acceptance of digital HIV prevention tools is influenced by perceived benefits, limitations, and preferred privacy-related features. ³² Therefore, in the present study, the significant effect of performance expectancy may not only reflect users’ evaluation of functional effectiveness, but also their perception that the HIV prevention mHealth management system can provide discreet, low-exposure, and convenient access to HIV prevention and health management services.

The significant positive effect of social influence on behavioral intention (β = 0.412, P < 0.001) suggests that MSM users’ behavioral intention is strongly shaped by peers, community networks, and social recognition. ³¹ This finding aligns with previous UTAUT-based health intervention studies in developing countries^18,33 and is consistent with the evidence reported by Mi et al., which highlights the role of social support in promoting the adoption of HIV interventions among adolescent MSM.^27,28 For individuals who have experienced long-term stigmatization, recognition and support from peers or the broader social environment may reduce psychological resistance to using technology, thereby enhancing their willingness to adopt mHealth systems for HIV-related health management.

The observed effect of social influence may further reflect the psychosocial context in which MSM make decisions about HIV-related service use. Stigma has been shown to affect HIV testing decisions among gay, bisexual, queer, and other MSM, with stigmatization potentially prompting individuals to avoid or delay HIV testing. ³⁴ In this context, recognition and support from peers, community organizations, or trusted health professionals may help reduce psychological resistance and increase confidence in using HIV-related mHealth systems. In addition, among MSM populations who may experience identity concealment and social stigma, the effect of social influence may partly reflect broader psychosocial mechanisms, including peer recognition, perceived community support, and reduced anticipated stigma, rather than merely interpersonal pressure.

In this study, effort expectancy (β = 0.039, P = 0.700) and facilitating conditions (β = 0.157, P = 0.106) did not have significant effects on behavioral intention. One possible explanation is that many participants may already have been familiar with smartphones, mobile internet, and WeChat-based services, making operational difficulty and basic technical resources less salient barriers to their intention to use the HIV prevention mHealth management system.^29,33 In addition, because the system was introduced to participants as an HIV prevention and health management tool with expected functions and use scenarios, participants may have focused more on its perceived preventive value, privacy protection, and social acceptability than on ease of use or external technical support. ³³ The relatively strong effects of performance expectancy and social influence may also have overshadowed the contributions of effort expectancy and facilitating conditions.^27,28 However, these non-significant paths should not be interpreted simply as evidence that effort expectancy and facilitating conditions are unimportant. Ceiling effects may have reduced the variability of these constructs if most participants perceived mobile phone access, internet use, or basic system operation as relatively easy. Measurement insensitivity may also have contributed to the non-significant findings, as the items may not have fully captured MSM-specific barriers to system use, such as concerns about privacy, stigma, trust, confidentiality, or sustained engagement. Therefore, future studies should refine the measurement of effort expectancy and facilitating conditions in HIV prevention mHealth contexts.

The multi-group analyses further refined the interpretation of these findings by examining whether demographic characteristics moderated the UTAUT relationships. Given the relatively older age distribution of the sample, age was specifically tested as a potential moderator, as age-related differences in digital literacy, mHealth experience, and mobile internet use may influence technology acceptance. And the relatively older age profile of the sample may partly explain why effort expectancy and facilitating conditions were not significant in the overall model, as participants recruited through NGO and VCT channels may already have had some degree of health-service engagement and familiarity with commonly used mobile platforms. Age and marital status did not significantly moderate the structural paths, suggesting that the overall pattern of associations between UTAUT constructs and behavioral intention was relatively consistent across age and marital status groups. However, educational level showed a significant moderating effect. Among participants with high school education or below, behavioral intention was mainly driven by performance expectancy, indicating that perceived preventive value may be the primary factor shaping acceptance in this group. In contrast, among participants with college education or above, performance expectancy, social influence, and facilitating conditions were all significant predictors, suggesting that higher-education participants may be more responsive not only to perceived system benefits, but also to social support and available enabling resources. This pattern may partly reflect differences in digital health literacy and the ability to use enabling resources, as digital literacy can influence health information access, interpretation, and engagement with digital health services. ³⁵ These findings suggest that the non-significant overall effect of facilitating conditions may mask subgroup-specific effects, particularly among participants with higher educational attainment, whereas the non-significant effect of effort expectancy was relatively consistent across demographic subgroups. From a practical perspective, these results support more tailored HIV prevention mHealth interventions. For participants with lower educational levels, intervention promotion may need to emphasize clear and concrete prevention benefits, whereas for participants with higher educational levels, strategies that combine system functionality, social support, and accessible service resources may be more effective.

From a theoretical perspective, the contribution of this study lies primarily in the contextual application of the original UTAUT model rather than in theoretical advancement. By applying the UTAUT framework to HIV-negative MSM’s intention to use an HIV prevention mHealth management system, this study provides empirical evidence on how established technology-acceptance constructs operate in a specific public health and sexual-minority context. Previous UTAUT-based health technology studies have often focused on general users or patient populations, whereas HIV-negative MSM represent a prevention-oriented population with specific service needs, privacy concerns, stigma-related barriers, and potential avoidance of offline HIV-related services. The findings suggest that, in this context, performance expectancy and social influence are the most salient determinants of behavioral intention, whereas effort expectancy and facilitating conditions play a comparatively limited role. This pattern indicates that MSM’s initial acceptance of HIV prevention mHealth systems may depend less on perceived ease of use or basic technical resources and more on perceived preventive value and social-contextual support. Therefore, this study adds context-specific empirical evidence to the UTAUT literature by examining the relative relevance of its core constructs in an HIV prevention mHealth setting for HIV-negative MSM.

More specifically, performance expectancy in this study reflects participants’ expectation that the system could improve access to HIV prevention information, intervention services, risk reduction support, and continuous health management. Social influence may reflect the role of peer recognition, community-based support, and encouragement from trusted health or social service providers in shaping MSM’s willingness to engage with HIV prevention services. In this sense, the study contextualizes the application of UTAUT by illustrating how the relative importance of its core constructs may vary in a sensitive HIV prevention context. These findings also offer guidance for future HIV-related mHealth interventions: developers should optimize system functionality and enhance social support and interaction mechanisms to increase users’ behavioral intention.^27,28,33

It should be emphasized that the system examined in this study was not presented to participants as a general health management application, but as an HIV prevention mHealth management system for MSM. Although the original questionnaire used the abbreviated terms “intelligent management system” or “the system,” these terms were used after a standardized explanation of the system’s HIV prevention objectives, functions, and expected use scenarios. Therefore, participants’ responses to the UTAUT constructs should be interpreted within the specific context of HIV prevention and health management, including HIV-related information acquisition, access to prevention services, risk reduction, and continuous health management support. This context is important for interpreting the observed effects of performance expectancy and social influence, because perceived usefulness and social support in this study were closely linked to the system’s ability to provide discreet, convenient, and HIV-specific preventive services rather than to general mHealth use.

This study has several limitations. First, although SEM was used to examine theoretically specified pathways among UTAUT constructs, the cross-sectional design does not allow causal inference or confirmation of temporal ordering among the variables. Therefore, the observed associations should be interpreted as statistical relationships rather than evidence of causal pathways. Longitudinal or experimental studies are needed to further examine whether performance expectancy, social influence, effort expectancy, and facilitating conditions causally influence subsequent adoption or sustained use of HIV prevention mHealth systems. Second, participants were recruited through NGO and VCT-related channels using a non-probability sampling strategy, which may have introduced self-selection bias. MSM who were connected to NGOs, VCT clinics, or community-based prevention networks may have greater HIV-related health awareness, higher willingness to engage with prevention services, stronger trust in health-service providers, or more positive attitudes toward digital health interventions than MSM who are hidden, less service-connected, or less willing to participate in HIV-related studies. This recruitment strategy may also partly explain the relatively older age profile of the sample. Therefore, the findings should be generalized with caution to the broader HIV-negative MSM population, especially younger MSM or MSM who are not connected to community-based HIV prevention services. A related limitation is that the total number of individuals approached, refusals, and unreturned questionnaires were not systematically recorded; therefore, the response rate could not be calculated, which limited our ability to assess potential non-response bias. Third, self-reported data from a single time point may be subject to social desirability bias and common method bias.^27,28 Fourth, behavioral intention was measured rather than actual usage behavior, so we cannot determine real-world adoption or continued use. Fifth, the Behavioral Intention construct relied on only two items, and its internal reliability was slightly below the recommended threshold, though the items were administered in a clear HIV prevention context. Although the Behavioral Intention construct demonstrated acceptable reliability for exploratory research, its AVE was slightly below the conventional threshold, indicating that the convergent validity of this construct may be suboptimal. This measurement limitation should be considered when interpreting the structural relationships involving Behavioral Intention. In addition, the UTAUT model omitted MSM-specific psychosocial factors (e.g., privacy concerns, stigma fear, trust, confidentiality, community belonging), which may moderate the effects of performance expectancy and social influence. Finally, although the questionnaire was adapted from prior UTAUT-based and mHealth acceptance studies and an item-level source mapping table was added, it was not formally validated specifically for HIV-negative MSM in China. In addition, no formal pilot testing, cognitive interviewing, or independent translation validation was conducted before the main survey, and some items retained generic UTAUT wording that may not fully capture MSM-specific concerns, such as privacy, stigma, trust, confidentiality, and community support. These issues may have limited measurement precision and construct validity in this sensitive context.

In conclusion, performance expectancy and social influence are the key determinants of behavioral intention to use HIV-related mHealth systems among the MSM population in China, whereas effort expectancy and facilitating conditions have a limited impact. These results highlight the importance of focusing on system functionality and social support in the design of future HIV mHealth interventions.

Conclusions

This study provides empirical evidence that performance expectancy and social influence primarily drive MSM users’ behavioral intention toward HIV-related mHealth systems, while effort expectancy and facilitating conditions play a comparatively minor role. Beyond identifying key influencing factors, the findings offer practical guidance for designing mHealth interventions that enhance system usability and social support mechanisms. Additionally, the study contributes to the theoretical understanding of technology adoption behaviors in proficient user populations, informing both implementation strategies and future research directions in HIV-related digital health interventions.

ORCID iD

Hong Chen https://orcid.org/0009-0007-2871-7432

Ethical considerations

This study was approved by the Ethics Committee of the Jiulongpo District Center for Disease Control and Prevention in Chongqing, China (Approval No. JLPJK-E-2023-10).

Data Availability Statement

The data that supports the findings of this study are available from the corresponding author, upon reasonable request.*