Since 1954, when the first renal transplant was successfully conducted between twins, a series of sustainable progresses such as refined surgical techniques improved immunosuppressive protocols and optimized perioperative management of transplant patients have been utilized to improve patient and graft survival following the renal transplant (Tonelli et al., 2011). However, the renal transplant community has been now challenged by acute rejection and transplant graft loss because of IM nonadherence (Prihodova et al., 2014). A recent study showed that IM nonadherence could lead to increasing risk of cancer over the long term in renal transplant patients (RTPs) (Yadav et al., 2019). However, research reports indicated that the prevalence rate of IM nonadherence in China ranges from 23.2 to 54.9% in RTPs (Teng et al., 2015; Xia et al., 2019; Liu et al., 2021), which was almost consistent with a worldwide incidence of 20–50% (Germani et al., 2011). Therefore, identifying the predictive factors with IM nonadherence would allow us to evaluate patients’ need for interventions to improve adherence, which could reduce the acute rejection rate and transplant graft loss to improve transplant outcomes. Numerous studies about nonadherence to IM in RTPs have indicated that demographics such as age, gender, working and socioeconomic status, educational level, and marital status were often associated with nonadherence rates (Spivey et al., 2014; Kung et al., 2017).

Several postulated theories or models have been utilized to predict and intervene nonadherence to IM in RTPs. The health belief model (HBM) model consisted of perceived susceptibility, perceived seriousness, perceived benefits, and perceived barriers, which has been widely used in many health-related behaviors (cancer, heart disease, hypertension, and renal transplant recipients) (Brown et al., 2011; Horwood et al., 2015; Zhao et al., 2015; Quast et al., 2016; Yang et al., 2016; Kung et al., 2017). An effective application of the HBM model has been conducted to predict IM nonadherence among RTPs in our transplant center, and our results indicated that perceived seriousness and perceived barriers were closely associated with IM nonadherence (Xia et al., 2019). Also, a recent study indicated that beliefs about medications were more powerful predictors to nonadherence than clinical or sociodemographic factors among patients with multiple sclerosis (Neter et al., 2021). The theory of planned behavior (TPB), as a cognitive model, also exhibited several advantages to predict a wide range of health-related behaviors (Andrew et al., 2016; Li et al., 2019; Caru et al., 2020). The TPB model only consisted of several variables (attitudes, subjective norms, perceived behavioral control, and intentions), which were avoided to include and test so many demographic and clinical factors. The validity of the TPB has been supported by different behavior domains and populations within large samples and applied successfully to predict IM nonadherence into solid organ transplant recipients (Chisholm et al., 2007; Hugon et al., 2014). Nevertheless, so far, all the predicted behavior models were usually tested independently to predict health behaviors, and few models were well suited to identify all the factors that contributed to nonadherence to a prescribed medical regimen as crucial as immunosuppressive therapy and each model had its limitations. For example, Chisholm et al. (2007) reported that the predictive capacity of the TPB could be enhanced by adding other variables such as the past behavior. Our previous study has demonstrated that the HBM could be used to predict health-related behaviors (Xia et al., 2019); however, a combined theory model to predict medication nonadherence has received little attention in RTPs. Therefore, we attempted to add the HBM variable to the TPB model to enhance the predicted power of combining the TPB and HBM model on IM nonadherence in RTPs. To the best of our knowledge, this is the first study to explore the combined effects of the TPB and HBM model on IM nonadherence in Chinese RTPs, and we hope to increase the prediction of the models on IM nonadherence and suggest some interventions to improve the IM nonadherence.

Our article initiated the first study of the incorporation between TBP variables and the HBM model in IM nonadherence in Chinese RTPs. Also, the purpose of our work was to conduct the combined model to determine the barriers to IM nonadherence and provide some new insights on how to improve the IM adherence among the RTPs.

In the current study, the IM nonadherence of the patients was 33.53%, which was in accordance with previously reported 5–50% range of nonadherence rate in renal transplantation (Leendertse et al., 2008; Germani et al., 2011). Forgetfulness was one of the major reasons for nonadherence to medications in other clinical settings such as HIV (Morowatisharifabad et al., 2019). Better drug reminder also effectively reduced the rate of nonadherence.

The binary logistic regression analysis suggested that several sociodemographic characteristics such as marital status, household income, preoperative drinking history, the time after transplantation, and religion were independent predictive factors of IM nonadherence. Our study indicated that distinct marital statuses exhibit different capacity of IM nonadherence; married recipients were more likely to adhere to IM than to single or divorced recipients. Additionally, findings from other studies have revealed that the family member–based supervised therapy used in patients with hypertension and type 2 diabetes may have positive effects on patient IM nonadherence (Mayberry and Osborn, 2012; Huang et al., 2014; Shen et al., 2017).

Meanwhile, family income had negative effect on IM nonadherence. Although a Brazilian study reported that higher family income was the only factor that was associated with immunosuppressive nonadherence because of lower income recipients benefiting from better access to care and coverage of health-care costs after transplantation (Marsicano et al., 2015), higher family income could support transplant recipients to pay for their IM, check the IM concentration, and improve the quality of life in China. Also, the life quality manifested a direct effect on medication adherence in Asian RTPs (Ganjali et al., 2019). In addition, religion issues also displayed a considerable influence on positive attitudes toward transplantation (Messina, 2015; Kobus et al., 2016). Religions like Christianity, Islam, Judaism, Buddhism, or other denominations provided guidance and principles for their behavior and thinking in these areas. Therefore, the recipients with religions may have some special etiquettes to limit the use of IM.

Moreover, our result showed that the recipients within half to one year were more likely adherent to IM. Interestingly, the recipients with more than three years did not show significant correlation with adherence to IM. Many studies reported that RTPs believed their allografts would become relatively histocompatible over a longer time (Chisholm et al., 2007; Zhao et al., 2017). In the present study, one of the plausible reasons was that the patients were more appreciative of their allografts and cautious in the early postoperation. Also, the more the time passes, the less the appreciation they have. Meanwhile, doctor’s advices and patients’ positive attitudes about the usage of immunosuppressive drugs for allograft might be diluted when time passes.

Finally, RTPs with preoperative drinking history were prone to IM nonadherence and studies in alcoholism have highlighted the impairments of affecting episodic memory as well as semantic and cognitive procedural learning (Pitel et al., 2007; Le Berre et al., 2010; Noël et al., 2012; Le Berre et al., 2017). So patients with preoperative drinking history tend to forget to take IM. From the above results, important strategies for intervention are needed to be utilized to minimize the barriers and improve the adherence to IM.

In the present study, TPB only explained 2% of variance in adherence to IM. Based on our previous empirical and theoretical work (Xia et al., 2019), IM beliefs were closely associated with IM nonadherence in Chinese RTPs. Our result showed that combining the TPB and HBM model could explain 52% of variance in IM adherence as well as all variables except for the subjective norms were independent predictive factors of IM nonadherence according to the logistic analysis. Several studies have noted that due to variations in the behavior under investigation, it should not be expected that all TPB factors (attitudes, subjective norms, and perceived behavioral control) will always be significant and stronger support for some components of the TPB than others, with subjective norms often emerging as the weakest predictor (Chisholm et al., 2007). Our results indicated that though the small items of the TPB model showed some advantages in predicting a wide range of health-related behaviors, it did not exhibit superiority in our large sample study. Based on our previous empirical and theoretical work (Xia et al., 2019), we integrated the HBM variables into the TPB structural equation model and found that the combined model has the ability to improve the predictive capacity of IM nonadherence than that in any single model. Therefore, we proposed that the combination of TPB and HBM exhibits more advantages in predicting IM nonadherence in studies with large samples.

In the structural equation model, we discovered that attitudes act as a salient predictor to IM intentions (0.284), both directly and indirectly. Chisholm MA et al reported that those RTPs who were particularly at risk might have a history of nonadherence to medical advice, especially when they had negative attitude about IM adherence (Chisholm et al., 2007; Hugon et al., 2014). An individual's perception of the "benefits and harms" of the target behavior is crucial to the behavioral attitude. Thus, important interventions to improve and strengthen attitudes about IM benefits were needed to promote the IM adherence. Notably, subjective norms were second only to attitudes in predicting IM intentions (0.278). In other words, the greater the influence of other individual or group giving to RTPs in IM adherence, the stronger the intentions to take medication. Therefore, medical staff should not only provide positive psychological interventions and social support to RTPs but also enhance their family’s awareness of taking IM regularly and timely.

In our study, HBM contributed the overall effect to IM nonadherence with the highest coefficient of effect and perceived barriers could play a vital influence on IM nonadherence. Our result was consistent with some previous studies showing that perceived barriers tend to be the strongest predictor of compliance (Wang et al., 2006; Carpenter, 2010). However, a recent study provided some different evidence that perceived susceptibility to rejection and perceived benefits of treatment were identified as major predictors of IM nonadherence (Kung et al., 2017). This difference might be due to different sample sizes in these studies. Nevertheless, all studies confirmed that health-related beliefs measured by HBM is an effective predictor of IM nonadherence. Based on our results, future educational programs should focus on the seriousness of not taking IM for Chinese RTPs. Some perceiving barriers, such as forgetting to take medicine and conflicts with social events, were similar to other reports. A recent study about hypertension population reported that forgetting was the most common reason for causing difficulty in taking medication (Yang et al., 2016). Understanding the barriers to optimal adherence provided information that could be used to better target interventions by doctors or health-care professionals. According to the present results, some effective methods, such as electronic medication devices with reminder and record functions, could play important roles in enhancing RTPs’ adherence. Our study also found that all variables of the TPB model did not exhibit sufficient effects on IM nonadherence, which might be one explanation of why TPB only explained 2% of variance in IM nonadherence. Until now, IM nonadherence could be predicted by several models, each having advantages and disadvantages, but none was considered as the gold standard. So we hoped that combination of the TPB and HBM model could make up for their shortcomings and provide much more interventions to improve transplant outcomes in clinical applications.

The most important aspect of this study included using the combined model and representing the larger sample size. Admittedly, the present study still had some limitations. One limitation was the potential bias, which was considered as follows: 1) A prestudy sample size calculation was absent. Hence, the findings of this study should be interpreted with caution. This is because the absence of sample size calculation may lead to systematic errors. 2) The adherence questionnaire used in this study was based on self-report, hence there could be recall bias. 3) Randomization was not employed in the selection of participants, hence the way the sample was obtained could introduce selection bias. Another limitation was that our study lacked effective external validity at present because of the COVID-19. In the future, multicenter studies with larger samples of different age-groups and some objective measures of IM nonadherence are needed to confirm our conclusions.

Our study, to our knowledge, is the first to explore the combined effect of the TPB and HBM model on IM nonadherence in Chinese RTPs, which could significantly improve the predictive capacity of any single model in IM nonadherence. Important implications for intervention can be drawn from our results. The logistic regression analysis results suggested that medical staff should focus on the RTPs with single or divorced status, families with low income, long time course after renal transplantation, and preoperative drinking history. Moreover, the combined model implied that encouraging RTPs’ willingness toward IM and enhancing their families’ awareness of taking IM regularly and timely were great ways to improve IM adherence. Meanwhile, future interventions should be conducted to increase perceived seriousness and to reduce perceived barriers for taking IM, which will effectively increase IM adherence rates and reduce the rejection risk to improve transplant outcomes. Future multicenter validation about the prediction and prevention of IM nonadherence are needed to confirm our conclusions, which would make more RTPs benefit.