Background
Massive migration poses severe challenges to public health worldwide. Globally, vulnerability to ill-health is a big issue towards to both international and internal migrant populations [
1]. Migrants are considered to be at elevated risk for sexual and reproductive health complications and diseases, such as unintended pregnancy, induced abortions, sexually transmitted infections (STI) and human immunodeficiency virus infection/acquired immunodeficiency syndrome (HIV/AIDS) [
2‐
4]. China’s internal migrant population (most coming from rural area to cities) has been increasing over the last decade and reached a historical record of 281 million in 2016 [
5]. Separation from family, unfamiliar environments, more open attitude toward sexual behavior and limited knowledge of sexual and reproductive health enhance risky sexual behaviors among migrants [
6‐
8]. A meta-analysis of 54 studies showed that among rural-to-urban migrants, populations floating out of or floating into provinces were observed to have a higher HIV prevalence (0.15 and 0.38%) than the general Chinese population (0.057%) [
9]. Additionally, household registration (a record officially identifies a person as a resident of an area)leads to unequal access to health care, with rural migrants often lacking access to reproductive health services in cities [
10]. In China, family planning services regarding reproductive health target married couples only, meaning that unmarried females are likely excluded. Previous studies have shown more than 15% of unmarried migrants have reportedly experienced unintended pregnancy [
11,
12], most of which concluded in induced abortions [
11‐
13]. Therefore, unmarried female migrants’ vulnerability to sexual and reproductive health risks requires special attention.
Condom use during sexual activity is one of the most feasible and efficient ways of preventing potential risks related to sexual and reproductive health [
14]. However, consistent condom use (CCU) remains limited in migrant populations [
15]. Therefore, to understand the related factors of CCU, it appears necessary to adopt a conceptually based, empirically tested, and highly generalizable model. To demonstrate what factors are related to CCU, several theories and models have been adopted, such as the Health Belief Model (HBM) [
16,
17], Social Cognitive Theory (SCT) [
18], Theory of Reasoned Action (TRA) [
19] and Theory of Planned Behavior (TPB) [
20]. Integration of these theories may yield a more complete characterization of the framework. Many of the constructs from HBM, SCT, and TRA/TPB have been incorporated into a generalized form, the Information-Motivation-Behavioral skills (IMB) model [
21]. This theoretical model was developed by Fisher and it predicts AIDS-risk behavior change [
22]. In particular, the IMB model explores the direct or indirect effects on behavior changes. The IMB model proposes three essential factors that bring about behavior change: information about transmission and prevention, motivation to reduce risk, and behavioral skills for performing risk reduction acts [
21‐
23]. Information refers to knowledge concerning AIDS transmission and AIDS prevention. Motivation was measured in accord with the constructs and operation of the TRA (attitude, subjective norm, and intention). Behavioral skills includes objective skills for promoting a certain act and a sense of self efficacy for doing so. The IMB model has already been used to understand condom use in a variety of populations and its assertions have also received considerable empirical support, e.g., among male street laborers [
24], students [
25,
26], and sex workers [
27]. However, few studies have focused on CCU by using health promotion theoretical frameworks and no study to date has applied the IMB model to investigating CCU among unmarried rural-to-urban female migrants of China.
Although the IMB model is a suitable framework for studying CCU related factors, it can be further improved. For example, the model could consider psychological problems and personality changes because they play important roles in sexual behavior among migrants, which may eventually influence their sexual and reproductive health status. Previous studies have found migrants to be more susceptible to psychological or personality problems, especially loneliness [
28,
29], depression [
30‐
32], and low self-esteem [
33,
34]. These psychological and personality factors and condom use have also been found to be significantly correlated. Condom use and loneliness have been found to be negatively associated among migrant workers [
29], men who have sex with men (MSM) [
35] and people living with HIV [
36]. Self-esteem was identified as one factor that was strongly correlated with condom use [
29,
37,
38]. Moreover, depression was either correlated with condom use or a predictor of condom use in MSM [
39], bisexual young men [
38], and disadvantaged adults [
40]. Aside from the correlation with condom use, psychological and personality factors have also been found to be related to some IMB model constructs. In particular, self-esteem was associated with condom use attitudes [
41] and self-efficacy [
41‐
43], while depression was related to attitudes toward condom use [
37] and communication skills for negotiating safer sex behaviors before sexual intercourse [
44]. All of these findings combined serve as an interesting supplement to the IMB model. Since psychological and personality factors cannot be integrated in information, motivation, or behavioral skill, their role in the IMB model therefore needs to be further explored. One study targeting male street laborers in urban Vietnam incorporated several new variables into a common construct labelled ‘psychosocial factors’ and assumed it worked directly or through behavioral skills to influence condom use [
24].
By using structural equation modeling (SEM), we examined the associations between condom use information, motivation, and condom use behavior skills. We incorporated ‘psychological and personality factors’ as an additional construct to form a modified IMB model. Then we tested the modified IMB model to add explanatory power to the original IMB model, thus providing more evidence to guide the design of future intervention. Our hypotheses are as follows:
(1) Condom use information and motivation will indirectly affect CCU through behavioral skills. Information or motivation will also directly affect CCU. Information and motivation are interrelated.
(2) Psychological and personality factors will directly or indirectly affect CCU, and the modified IMB model is suitable for our study.
Discussion
Psychological and personality factors have rarely been considered when a theory-based framework is applied to understand condom use. In this study, we examined whether or not the IMB model and the modified IMB model can be used to predict CCU and how these models’ constructs influence CCU among unmarried rural-to-urban female migrants in Shanghai. Our research showed that both the IMB model and the modified IMB model appeared to be suitable for predicting CCU in the studied population. Further, motivation, behavioral skill, and psychological and personality factors all influenced CCU.
Our findings confirm that unmarried rural-to-urban female migrants are a high-risk group in terms of sexual and reproductive health. The rates of unintended pregnancy and induced abortion among unmarried female migrants were 29.9 and 28.2%, respectively. These rates are higher than that of previous regional findings, including 15.3 and 14.6% in Shanghai in 2012 [
12], 10.4 and 9.8% in Qingdao in 2012 [
56], and 18.4 and 17.7% in Shanghai in 2013 [
11], respectively. Unprotected sex and failure of contraceptive methods are considered the main reasons for unintended pregnancies. In our study, the percentage of CCU among unmarried rural-to-urban female migrants in Shanghai during the 6 months prior to the survey was 13.8%, which was lower than that in two studies conducted in Shanghai (14.7% [
11], 53.3% [
12]) and another in Guangxi (26.4% [
45]). The difference may be due to two reasons. One is that our study measured CCU within the last 6 months, which was a much longer period than those of previous research. Another reason is that the actual prevalence of CCU in our participants was lower than that of previous studies, which reflects the higher rate of unintended pregnancy and induced abortion in our study.
In both models examined in our study, information was not associated with behavioral skills or CCU. This inconsistency between information and behavioral skills [
57,
58] and between information and HIV-preventative behavior [
27,
59] has also been observed in many other studies. According to Fisher, information is both necessary and sufficient for relatively uncomplicated behavior changes (e.g., avoiding sexual contact). However, although information is necessary, it is not a sufficient condition for a complicated behavior change (e.g., CCU) because many other factors may play more important roles [
22,
60]. Besides, information may impact initial behavioral change rather than maintain such behavior over time, which indicates information’s impact on maintaining condom use may be weak [
61]. In addition, condom use (and condom skills) involves communication and negotiation with a partner, but previous studies have suggested that one partner’s information may have a low impact on dyadic behavioral change [
62]. Additional evidence has shown that due to power dynamics between men and women, female migrants usually take a subordinate position in sexual interactions, which further lowers the impact on behavioral change [
63‐
65].
Consistent with our hypothesis, in both models, behavioral skills directly influenced CCU, while motivation contributed to CCU indirectly by affecting behavioral skills. However, no direct influence between motivation and CCU was found. Unmarried rural-to-urban female migrants who possess higher levels of behavior skill were observed to more likely perform CCU, while highly motivated individuals were more likely to acquire the requisite skills to perform CCU. These results highlight the crucial role of behavioral skills. Interventions targeted at enhancing self-efficacy and at promoting condom use skills may have a strong effect on CCU. However, to the best of our knowledge, no published study has focused on intervention for both CCU self-efficacy and condom use skills among unmarried rural-to-urban female migrants, which should be addressed in future studies.
In the modified model, while psychological and personality factors influenced CCU directly, they were not mediated by behavioral skills that affected CCU. Participants with higher levels of depression, higher levels of loneliness, and lower levels of self-esteem were less likely to use a condom consistently. Our SEM results confirmed that condom use and psychological and personality factors were negatively associated, which is consistent with the findings of previous studies that used multiple logistic regression analyses (condom use vs loneliness, or condom use vs low self-esteem, or condom use vs depression) [
36,
66,
67]. However, our finding that psychological and personality factors’ direct influence but nonexistence of an indirect influence is contrary to Nguyen’s findings [
24]. The IMB model in Nguyen’s study was modified by an additional construct labelled ‘psychosocial factors’, which eventually showed an indirect influence but nonexistence of a direct influence. We believe this difference may be explained by several reasons. First, the additional construct did not consist of the same variables. The psychosocial factors in Nguyen’s study included some variables from our additional construct (e.g., depression) and was not limited to the psychological and personality factors in our study (e.g., low access to aids prevention, or alcohol use). Moreover, the two studies’ target populations were different, with one focused on male street laborers in Vietnam, and the other on unmarried female migrants in China. Despite these differences, both studies showed a significant effect of additional constructs on condom use, thus providing alternative variables for other researchers attempting to modify the IMB model in predicting condom use.
Examination of the mediation effect suggested the relation between motivation and CCU is fully mediated by behavior skill. Additionally, we found that in the modified IMB model, the indirect effect of motivation was larger than the direct effect of psychological and personality factors, thus suggesting prioritization of intervention targeted at motivation and behavioral skills. Although the effect of psychological and personality factors was small, it influenced CCU directly, which indicates another effective target for intervention. Our findings suggest that a shift from information-motivation-behavioral skills-focused interventions to a more comprehensive strategy that addresses psychological factors may be necessary.
Both the IMB model and the modified IMB model can be used to understand CCU in unmarried rural-to-urban female migrants in Shanghai. Overall, we tend toward adopting the modified IMB model because our purpose is not to find a hypothetical model that best matches the observed data, but to explore the rationality and appropriateness of building a model based on theory and the results of previous research.
The results of our study should be interpreted within the context of study limitations. First, the data from the two chosen factories may not be representative of unmarried rural-to-urban female migrants in other regions. Because migrants are generally mobile, obtaining random samples is difficult. Therefore, we used convenience sampling, which limited the representativeness of our sample. Second, data were gathered from a self-report questionnaire, so the reliability of the responses to sensitive questions may be questionable. There may exist some information bias. For example, “never” and “always” may be selected when measuring condom use even if they are not accurate. We minimized this problem by implementing various interventions and tools such as using anonymous questionnaires and providing private rooms for the survey. Third, due to the design’s cross-sectional nature, measures of information, motivation, behavioral skills, and the personality and psychological variables are assessed contemporaneously with consistent condom use, which made it difficult to decide the causality. This also limits the ability to control for habitual or past behavior - such as the amount of past sex or number of partners. Fourth, we noticed the relevance of power dynamics between men and women, and the subordinate position taken by female migrants incontraceptive use after the implementation of the project. It would be better to set up a variable that measures the subordinate position of female migrants in relationship. Fifth, the latent variable “psychological and personality factors” didn’t contain all the variables that may influence CCU. We preliminary include three variable because 1) migrants had been found to be more susceptible to them; 2) these three variables had been found correlated with CCU and 3) these three variables had also been found related to IMB model constructs. Further researches can be conducted to explore more variables related to psychological and personality factors. Sixth, combining three variables into one “psychological and personality factors” variable appeared to limit straightforward interpretation of the results because each variable may have different influences. However, separating these factors would complicate the model, which violates the rule of parsimony in structural equation modelling.