Sampling

This online anonymous cross-sectional survey was conducted from November 2023 to March 2024 via the QuestionStar platform in six regions: Shanghai, Guangzhou, Xinjiang, Shaanxi, Yunnan, and Shenyang.Due to the typically discreet nature of the MSM, we had specifically commissioned social non-governmental organizations (NGO) that had access to and served the MSM population to assist us in conducting this research.

The staff of NGO leveraged their established trust and resource networks within the MSM community to actively promote and mobilize participation in the study. Through various channels such as social media platforms, offline events, and face-to-face interactions, they disseminated detailed information about the research, thoroughly explaining its background, objectives, and potential societal significance, including raising awareness of MSM health issues and contributing to improved disease prevention and management strategies. They also addressed any questions from MSM, including those related to the participation process, privacy protection measures, and participants’ rights and responsibilities, ensuring a clear understanding of the study’s purpose. Based on this transparent communication, MSM individuals expressing interest were further screened to confirm they met the inclusion criteria.To fully protect the privacy of the subjects, the subjects in this study completed the survey using mobile devices such as phones on the QuestionStar platform.Informed consent was obtained, ensuring participants understood the study’s purpose, procedures, rights, and potential risks.

Data collection adhered to privacy regulations. The inclusion criteria are: (1) being male and at least 18 years old; (2) having engaged in male-to-male sexual behavior in the past six months; (3) primarily residing in one of the selected six provinces, municipalities, or autonomous regions over the past six months. The exclusion criteria are: (1) completing the questionnaire in less than 300 s; (2) providing incorrect responses to quality control questions; (3) having an IP address outside the provinces or cities under investigation.All participants were required to complete the MMHSBS and respond to questions related to their knowledge of Mpox, demographic information, sexual behavior and sexual health, history of sexually transmitted disease diagnoses, and health behaviors.The survey is estimated to take approximately 20 min to complete, and upon completion, each MSM participant was provided with a $7 remuneration as compensation for their time and effort.

Sample size

A total of 2481 questionnaires were collected in this survey, with 2403 being valid, resulting in a validity rate of 96.86%. The distribution of valid questionnaires was: 569 from Shanghai, 500 from Guangzhou, 320 from Xinjiang, 199 from Shaanxi, 313 from Yunnan, and 502 from Shenyang. The focus of this study was on the healthcare-seeking intentions of MSM who were unaware of their Mpox infection status. Therefore, it was necessary to exclude 56 participants who self-reported that they had already been diagnosed with the virus by a doctor. After exclusions, the sample sizes were: 546 from Shanghai, 490 from Guangzhou, 317 from Xinjiang, 193 from Shaanxi, 307 from Yunnan, and 494 from Shenyang, totaling 2347 questionnaires. Figure 2 illustrates the geographical locations of these provinces and municipalities in China.

The sampling distribution of this study in China.

Quality control

The design of the survey questionnaire incorporated advice and suggestions from multiple experts. After completing the initial draft, a pre-survey was conducted with 12 MSM participants. Based on the pre-survey results, modifications were made to the questionnaire, including adjustments in vocabulary and expressions to better suit the MSM population, enhancing the survey’s acceptability. For survey implementation, the study primarily relied on local NGO platforms, with NGO staff serving as researchers. Uniform training was provided to ensure these researchers understood the survey’s objectives, content, procedures, methods, and precautions. A data quality monitoring team was established to regularly check and provide feedback on the data within 48 h during data collection. After data collection, steps for data cleaning and processing were undertaken, including checking for outliers, missing data, and logical errors, and making necessary corrections and adjustments. Participants were given a malpractice allowance for completing the questionnaire to increase compliance with the survey.

Development of the MMHSBS

Firstly, a comprehensive literature review and expert interviews were conducted to gather existing research findings and professional opinions, which helped identify the key factors the scale should cover. Subsequently, interviews with MSM individuals were performed to understand their perspectives on seeking healthcare after Mpox infection, as well as perceived barriers and worries, to refine the initial items. Based on the literature review, expert opinions, and MSM interviews, initial questionnaire items were generated and formatted using a 5-point Likert scale to facilitate quantitative analysis. Next, a pilot survey was conducted with a small sample of 30 MSM individuals to collect preliminary data and obtain feedback, assessing the clarity and applicability of the items, and identifying and correcting potential issues. Finally, a 14-item MMHSBS was developed. The specific items of the MMHSBS are presented in (Supplementary Table 1).

Statistical methods

Content validity

Content validity assesses how well the items of a scale align with and represent the intended construct for a specific assessment objective. The predominant method for evaluating content validity is the content validity index (CVI)¹⁶. The CVI is calculated both at the item level (I-CVI) and for the entire instrument (Ave-CVI).

The CVI is based on expert evaluations of each item’s relevance or representativeness, typically rated on a 4-point Likert scale from 1 (not relevant or representative) to 4 (highly relevant or representative). The I-CVI is determined by the proportion of experts who rate an item as 3 or 4, indicating a consensus on the item’s content validity¹⁷. An I-CVI of 0.8 or higher is considered acceptable.

The Ave-CVI represents the average proportion of items rated 3 or 4 by different experts, providing an overall measure of the instrument’s content validity^18,19. An Ave-CVI of at least 0.8 is required for acceptability. In this study, a total of 9 experts from the fields of public health, epidemiology, medical statistics, mental health, and health policy evaluated the content validity of the MMHSBS.

Structural validity test by factor analysis

We assessed item and scale reliability, performed exploratory factor analysis (EFA) to identify underlying relationships, and confirmatory factor analysis (CFA) to validate constructs. To ensure stability and avoid overfitting, we split the dataset (n = 2347) into two subsets: dataset A (n = 1173) for EFA and dataset B (n = 1174) for CFA.

Exploratory factor analysis

Bartlett’s test of sphericity and the Kaiser-Meyer-Olkin (KMO) measure assessed sample adequacy and data suitability for factor analysis. Bartlett’s test required p ≤ 0.05, and KMO needed to be > 0.7²⁰. The Kaiser criterion identified significant factors with eigenvalues > 1.0, but recent research advises comparing solutions with ± 1 factor^21,22. The Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) helped select the best model, with lower values indicating better fit. We used principal axis factoring with oblique rotation, retaining factors with loadings > 0.4²³. Dual-loading items are assigned to the most appropriate factor based on theoretical relevance and conceptual alignment, with retention in the higher-loading factor if the difference is significant.It is recommended to include a minimum of 3 items per factor²⁴.

Confirmatory factor analysis

CFA assessed the structural validity of constructs identified in the EFA. We reviewed literature and summarized 8 metrics for validation models. The Standardized Root Mean Square Residual (SRMR) and Root Mean Square Error of Approximation (RMSEA) are key measures, with SRMR ≤ 0.08 and RMSEA ≤ 0.08 indicating good fit^25,26. The Goodness-of-Fit Index (GFI) has a threshold of ≥ 0.90, measuring variance explained by the model. The Normed Fit Index (NFI) and Relative Fit Index (RFI) also require values ≥ 0.90^27,28. Incremental fit indices, including the Incremental Fit Index (IFI) and Tucker-Lewis Index (TLI), need thresholds of ≥ 0.90, and the Comparative Fit Index (CFI) requires ≥ 0.95 for acceptable fit^25,29.

Reliability

Split-half reliability, Cronbach’s α, and McDonald’s ω assess internal consistency of scales. Split-half reliability divides the scale into two halves, correlating scores to ensure each half measures the same construct, with ideal correlations > 0.70. Cronbach’s α evaluates all items collectively; α ≥ 0.9 indicates excellent consistency, 0.7 ≤ α < 0.9 is good, 0.6 ≤ α < 0.7 is acceptable, and α < 0.6 is poor. McDonald’s ω accounts for varying item loadings, with ω > 0.80 indicating good reliability.

Convergent validity

The average variance extracted (AVE) measures the average proportion of variance explained by the indicator variables for a construct. It is the mean of the squared factor loadings, indicating the explanatory power of each variable. An AVE of 0.50 or higher signifies that the construct’s variance is well captured by its indicators³⁰. Composite reliability (CR) assesses the internal consistency of the indicators, factoring in loadings and error variance. A CR of 0.70 or higher indicates high internal consistency and good convergent validity³¹.

The formulas for AVE and CR are as follows:

,

.

(n = number of items for each factor; i = constant = 1,2,3…; λ = factor loadings;δ = item error.)

In this study, a more conservative judgment condition was adopted, requiring both AVE ≥ 0.5 and CR ≥ 0.7 to be considered a factor with Convergent Validity.

All statistical analyses in this study were performed using IBM SPSS Statistics (Version 27), IBM SPSS Amos (Version 27), and R (version 4.3.3).