The primary objective of this study was to evaluate the psychological resilience of the MSM population and identify the key factors that influence it. The MSM population is particularly vulnerable to mental health issues such as depression2,3. Therefore, understanding the determinants of psychological resilience is essential for developing targeted interventions. Numerous traditional statistical methods have been employed to investigate the psychological resilience and its influencing factors in the MSM population6,10,18,19. However, as data on psychological resilience in the MSM population often exhibits complexity, including nonlinear relationships and high-dimensionality, these traditional methods are less effective. In contrast, machine learning methods can efficiently manage such data34. This study employed both the CHAID decision tree algorithm and logistic regression, leveraging their combined strengths to provide a comprehensive analysis of how social support, self-esteem, depression, and education impact psychological resilience in this population. The findings aim to inform policy-making and guide the implementation of effective mental health and HIV prevention strategies for MSM.
In the present study, among the 1,070 participants from the MSM population sample, 749 individuals with low psychological resilience (scoring ≤ 60 points) were identified, accounting for 70.0% of the MSM sample. This suggests that the psychological resilience of this population is relatively low. It is recommended that a variety of strategies and measures be used to increase the psychological resilience of this population in the future. These strategies and measures include: (1) psychological education and awareness enhancement: Conduct mental health education activities through online and offline mental health lectures, workshops, and training courses to improve the understanding of psychological resilience among the MSM population35. Meanwhile, provide mental health information resources, develop mental health handbooks and online resources specifically for the MSM population to help them understand how to cope with stress and enhance psychological resilience; (2) strengthen social support networks: Establish community support groups, organize and support mutual aid groups within the MSM community to provide a safe environment where members can share experiences and support each other. In addition, enhance cooperation with mental health institutions to provide professional psychological counseling services for the MSM population; (3) skill training: Conduct stress management training to help the MSM population learn effective coping skills, such as meditation, deep breathing, and mindfulness exercises36. Furthermore, social skills training can be provided to help the MSM population build a broader social network and strengthen their social support system; (4) policy and environmental interventions: Advocate for inclusive policies and encourage policymakers to develop and implement policies that reduce discrimination and prejudice against the MSM population;37 improve the community environment through community activities and projects to enhance the living environment of the MSM population and reduce social pressure.
Our findings underscore that depression, social support, self-esteem, and educational background are closely linked to low psychological resilience in this population. It has been shown that there is a significant negative correlation between psychological resilience and depression14,18,25. The prevalence of depression in the MSM population is higher than in the general population38. The prevalence of depression in the present study was 28.9%, which is similar to the 26.2% reported by Wang in the MSM population39. Individuals with depressive disorders typically have reduced levels of psychological resilience, suggesting that they have a reduced ability to cope with stressors and are more vulnerable to the onset of depressive episodes. Furthermore, psychological resilience has a partial mediating effect on the relationship between positive coping and depressive mood40. This suggests that psychological resilience not only influences depressive mood directly, but also indirectly by influencing coping styles. Consequently, enhancing psychological resilience is an effective approach to improving depressed mood in depressed patients. This can be achieved by strengthening social support, promoting self-esteem and developing positive coping strategies.
In the MSM population, the relationship between social support and psychological resilience showed a significant positive correlation41. with lower levels of social support reducing the level of psychological resilience in the MSM, which in turn affected their psychological mood. Hussain et al.42 reported that the majority of the MSM in Pakistan lacked family support, either emotional or financial. Specifically, 76% of MSM received no financial support from their families. In addition, 29.45% of MSM did not receive family support after disclosing their sexual orientation. Liu et al.41. showed that 34.50% of the MSM had low social support, whereas the prevalence of low social support in the MSM population sample of the present study was 12.90%, which is lower than the findings of both Liu et al.41. and Hussain et al.42. Social support can be conceptualised as a social network comprising three dimensions: family support, friend support and other forms of support (e.g. social relationships with neighbours, leaders, etc.)43. A significant positive correlation was found between all dimensions of social support and psychological resilience. It is therefore recommended that additional social support channels be made available to MSM, especially those who are HIV-positive, in order to increase their psychological resilience and reduce the prevalence of depression10.
There is a significant positive correlation between self-esteem and psychological resilience in the MSM population. Self-esteem can be defined as an individual’s assessment of their own worth and abilities. It plays an important protective role in the face of adversity. Individuals with high self-esteem tend to be more psychologically resilient and better able to cope with challenges and stress. This is particularly important for the MSM population, especially those who are HIV positive or depressed. Akhtar and Bilour44 conducted a study on the mental health status of the MSM population in Pakistan. They found that 29% of the MSM population had low levels of psychological resilience, while 74% had low to moderate levels of self-esteem. Notably, there was a significant positive correlation between their psychological resilience and self-esteem, and MSM living with a mentor had significantly higher levels of psychological resilience and self-esteem than those living alone or with friends. The present study showed that the prevalence of low self-esteem in the MSM population sample was 34.49%, which is slightly higher than that reported by Akhtar and colleagues44. A Classification Tree Analysis (CTA) showed that self-esteem was the most significant predictor of psychological resilience45, a finding consistent with the results of the present study. Consequently, it is imperative to implement strategies that promote psychological resilience in the MSM population in order to improve their mental health18.
Numerous studies have demonstrated a significant positive correlation between educational attainment and psychological resilience in the MSM population44,46,47. As an individual’s level of education increases, so does their psychological resilience, enabling highly educated individuals to cope with stress and adversity. Akhtar and Bilour44 have shown that educational attainment is significantly and positively associated with psychological resilience. Our study’s multivariable ordinal logistic regression results have also showed that education is significantly associated with psychological resilience after controlling for other variables, aligning with the findings of Akhtar and Bilour44. These findings have been confirmed in other populations as well46,47. This study underscores the importance of education in enhancing psychological resilience.
Logistic regression is a statistical method rooted in probability theory. It predicts the likelihood of a specific outcome by using a linear combination of independent variables to estimate the log-odds of the dependent variable48. The model is built on the assumption that there is a linear relationship between the log-odds of the dependent variable and the independent variables. Logistic regression’s coefficients directly indicate the extent of each independent variable’s influence on the dependent variable, making it particularly suitable for situations demanding clear explanations. In contrast, the CHAID decision tree is an algorithm that relies on chi-square test. It recursively partitions the data into a hierarchical structure to highlight differences among categorical variables49. The CHAID decision tree’s structure is visually intuitive and easy to grasp. It presents the data partitioning process and decision rules in a tree diagram, allowing researchers to readily understand the interplay between different factors. The CHAID decision tree excels at capturing non-linear relationships and interaction effects. For instance, it can discern the more pronounced negative impact of depression on psychological resilience when self-esteem is low, thanks to its hierarchical structure. In our study, both logistic regression and the CHAID decision tree pinpointed depression, social support, and self-esteem as pivotal factors shaping psychological resilience in the MSM community. However, logistic regression additionally revealed that education level significantly influences psychological resilience, a finding not identified by the CHAID decision tree. This discrepancy may be attributed to the logistic regression model’s ability to quantify linear relationships through estimated coefficients, whereas the CHAID decision tree is more adept at capturing non-linear relationships and interaction effects. Additionally, differences in variable selection and model interpretability may also contribute to these divergent results51. The integration of these two methods facilitates a comprehensive and nuanced analysis of psychological resilience and its determinants. This dual-method approach not only broadens the scope of understanding but also mitigates the risks associated with relying solely on one method, thereby safeguarding against the omission of critical insights52. Furthermore, it enhances the depth of inquiry into the underlying mechanisms that shape psychological resilience.
Logistic regression has been used in a number of studies examining psychological resilience in diverse populations53,54. Similarly, the CHAID decision tree algorithm has been used in studies of psychological resilience in college students and older adults55,56. Some researchers have integrated decision trees with logistic regression to predict diseases and associated factors57,58,59,60, such as non-suicidal self-injury in adolescents57, the association between anthropometric measures and total cholesterol in a large population58, HIV infection in the MSM population59, and suicidal ideation in psychiatric patients60. The combination of these two methods has not been used in studies of psychological resilience and its determinants in the MSM population, according to a search of several databases, including PubMed. In our study, a CHAID decision tree combined with logistic regression are used to assess the factors influencing psychological resilience in the MSM population sample. The results of both methods indicated that depression, social support and self-esteem are significantly correlated with psychological resilience. The results of the two machine learning models show high consistency and good performance.
The generalizability of our findings should be approached with caution. The sample, drawn from specific online platforms and social media groups, may not fully represent the entire MSM population due to differences in regional characteristics, socioeconomic status, culture, and HIV prevalence. Additionally, the study’s main focus on Zhejiang, China, could limit the broader applicability of the results. Given these limitations, caution is advised when extending these conclusions to other regions or countries. MSM in different regions may vary significantly in social support, self-esteem, depression, and education levels, all of which in turn affect psychological resilience. Moreover, reliance on self-reported data may introduce bias, which could affect the accuracy and generalizability of the results. To address these limitations, further longitudinal studies are essential to explore the causal relationships between the identified factors and psychological resilience in the MSM population. Additionally, it is recommended to conduct intervention studies to evaluate the effectiveness of educational programs and social support initiatives in enhancing psychological resilience.
Study limitations
The following limitations are inherent to this study: (1) The presence of depressive symptoms, self-esteem, social support and HIV status in our study was determined through self-report by the study participants, which may have introduced information bias, including recall bias, subjectivity bias, and reporting bias, among others, For example, given that the content of this survey is based on self-reports, reporting bias may occur. Participants may report having stronger social support, higher self-esteem, and so on; (2) As this study was a cross-sectional network survey, causal inferences cannot be made based on the results obtained; (3) Due to the need to protect the privacy of the study participants and maintain the confidentiality of the data, some potentially influential factors, such as social discrimination, mental health history, and the details of the social support network (including family, friends, and community support), were not included in this study; and (4) The respondents were required to recall their sexual behaviour over the past six months, which may be subject to recall bias.
