The Risk Perception Attitude (RPA) framework theory was established by American scholar Rimal and colleagues in 2003, and it is used to predict public attitudes and behaviors when facing disease threats (19, 20). The fundamental hypothesis of this theory is that the perception of future disease risk and the sense of self-efficacy are the two main factors influencing whether the public will take preventive measures. Among these, risk perception serves as the motivational factor for preventive behavior, while efficacy beliefs act as the facilitating factor for such behavior. Therefore, based on the levels of these two influencing factors, the population can be divided into four risk attitude groups: Responsive, Avoidance, Proactive, and Indifference, as shown in Figure 2A . These different groups exhibit varying degrees of self-protective motivation, intentions to acquire knowledge and behavioral intentions. Individuals falling under the “Proactive” and “Independent” classifications in the RPA framework are less inclined to engage in active behavior modification, given their low perceived risk levels. The “Responsive” cohort, defined by high levels of both risk perception and self-efficacy, shows an increased inclination towards active health behavior changes. Meanwhile, the “Avoidance” group, despite high risk awareness, struggles with low confidence in their ability to implement changes. Compared to the “Responsive” group, they are less likely to take action to change their behavior.

Several studies have revealed that personalized intervention strategies based on the risk perception attitude framework can effectively enhance the self-efficacy of the target group, particularly exerting significant positive influences in diabetes prevention and health behavior management. For example, Rains et al (21).emphasized that personalized information dissemination is conducive to increasing the screening rate. Simonds et al (22). indicated that risk perception and self-efficacy are crucial factors influencing diabetes prevention. Zou Xiaohui (23) emphasized that in the prediabetic population, self-efficacy plays a regulatory role in health behaviors. Nevertheless, the study by Jones et al. (24) discovered that even if patients possess a high level of risk perception and health knowledge, these factors do not effectively enhance self-efficacy, especially in preventing cardiometabolic diseases after gestational diabetes. Overall, although the RPA framework holds certain guiding significance in improving patients’ health behaviors, in different populations and intervention contexts, a single risk perception and self-efficacy do not always fully account for the mechanism of behavioral change. These disparities suggest that the application effect of the RPA framework in diverse situations still requires further exploration and validation. By categorizing patients according to their risk attitudes, targeted interventions and educational materials can be provided to enhance glycemic control. However, there has been no specific research focusing on postpartum glycemic management in GDM patients. It is essential to fully understand how their perception of future diabetes risk and their sense of self-efficacy influence medication adherence, in order to optimize healthcare services.

To address this objective, the present investigation advances the following set of hypotheses ( Figure 2B ):

The study included a total of 80 participants. Participants were recruited from pregnant women with GDM registered in the obstetrics department of a tertiary hospital in China between January 1, 2020 and December 31, 2022.

The study enrolled patients who met the following criteria: (1) Patients who delivered in the obstetrics department of our hospital between 2020 and 2022 and were diagnosed with GDM during pregnancy; (2) Patients diagnosed with abnormal glucose metabolism during postpartum blood glucose screening conducted 4 to 12 weeks after delivery; (3) Patients who continued to use anti-hypoglycemic drugs for more than 3 months postpartum.

Exclusion criteria included: (1) Patients with severe comorbidities involving the heart, brain, lungs, kidneys, liver, or immune system; (2) Patients who failed to undergo follow-up blood glucose tests after being diagnosed with abnormal glucose metabolism postpartum; (3) Patients experiencing impaired consciousness, inability to communicate normally, or refusal to complete the questionnaire; (4) Patients with significantly incomplete records.

Approaches to evaluating medication adherence are commonly segregated into direct and indirect categories. The direct methods comprise observational techniques, physiological marker assessment, and blood drug level quantification. Indirect methods encompass scale evaluation, pill count, and tracking clinic visits, among others. Clinical researchers and medical practitioners frequently opt for the scale evaluation method, given its inherent simplicity, ease of administration, low resource requirements, and high operational feasibility. This study employed the 8-item Morisky Medication Adherence Scale (MMAS-8), developed by Professor Morisky in 2008 (25). The original MMAS-8 has a Cronbach’s α coefficient of 0.83. The Chinese version of the MMAS-8, translated by Wang Jie and others (26), has a reliability coefficient of 0.65 and a validity coefficient of 0.80, demonstrating good predictive validity. It has been extensively used by Chinese researchers to assess medication adherence in patients undergoing long-term diabetes treatment (27, 28).

Please note that, unlike the original MMAS-8 scale where a score of ≥8 indicates good adherence, we have made adjustments to the classification criteria for the total score of MMAS-8 in consideration of the characteristics of the participants and the nature of antidiabetic medication treatment. During the assessment process, it was observed that participants in this study, especially postpartum women with a history of gestational diabetes and recent abnormal glucose metabolism, exhibited extremely poor medication adherence due to various factors such as infant care, postpartum depression, inadequate medication education, and neglect of their own health. As a result, there was a severe differentiation in the scale scores and an uneven distribution. In this study, researchers first provided the respondents with a detailed explanation of the study’s purpose and the content of the scale. The respondents were then asked to truthfully answer the 8 items based on their recent medication behaviors (see Table 1 ). The scoring method is as follows: for items 1/2/3/4/6/7 a response of “Yes” is scored as 0, while “No” is scored as 1; item 5 is reverse scored;Item 8 is scored using a Likert 5-point scale, with scores assigned as 1/0.75/0.50/0.25/0, from highest to lowest. The total score for the 8 items is then calculated. A score of ≥6 indicates good adherence, 4 to 6 indicates moderate adherence, and <4 indicates poor adherence (30).

A total of 2765 patients with GDM who were recorded and delivered in our hospital during 2020~2022 were retrospectively collected in HIS system. Of the 1153 (41.7%) patients who returned to the hospital at 4 to 12 weeks postpartum for blood glucose screening, 619 (22.4%) underwent a complete 75g OGTT and 534 (19.3%) had only FBG and/or HbA1c tested. Among the patients who underwent blood glucose screening 4 to 12 weeks postpartum, 425 (36.9%) had abnormal glucose metabolism. After excluding patients with severe diseases (9 patients), no blood glucose review (72 patients), refusal to fill in the questionnaire (47 patients), and serious data missing (23 patients), 80 of the remaining 274 patients received hypoglycemic drug treatment lasting more than 3 months after delivery and were included in the study.

A total of 80 postpartum GDM patients who received pharmacological interventions were surveyed to assess their medication adherence. The median score on the MMAS-8 scale was 3.75 (3.50, 5.50). Among these patients, 11 (13.8%) exhibited good adherence, 28 (35.0%) showed moderate adherence, and 41 (51.3%) had poor adherence. Significant differences in medication adherence were observed among patients with different educational levels, postpartum health education status, and treatment regimens, with these differences being statistically significant (P < 0.05), thereby confirming research hypothesis H1. Detailed data are presented in Table 2 .

The correlation matrix for the total scores of RP-DD, GSES, SMSPD, and MMAS-8 is shown in Table 3 . Spearman’s test results indicate that the perception of future diabetes risk among GDM patients is positively correlated with postpartum medication adherence (r = 0.778, P < 0.001), confirming research hypothesis H2. Additionally, self-efficacy was also positively correlated with medication adherence (r = 0.631, P < 0.001), confirming research hypothesis H3. Furthermore, a positive correlation was found between risk perception and self-efficacy (r = 0.241, P = 0.031), while no significant correlation was observed between self-management levels and the other three variables (P > 0.05).

A linear hierarchical regression model was selected to examine the moderating effect.

In the initial Model 1, only three control variables were included: patient education level, whether postpartum health education was received, and the treatment regimen. In Model 2, with the addition of independent variables, the significance of the three control variables diminished, while the centered total scores of RP-DD and GSES became significant. This indicates that both risk perception and self-efficacy jointly influence patients’ medication adherence, confirming research hypothesis H4. Model 3 further introduced an interaction term based on Model 2. The change in R² was significant, and findings showed a meaningful interaction, suggesting that self-efficacy moderates how risk perception affects an individual’s adherence to prescribed medications, thereby confirming research hypothesis H5. The model parameters are detailed in Table 4 .

To visually represent this moderating effect, patients were categorized into two groups – those with high and low self-efficacy – determined by their GSES scale scores, with a threshold of 24 points. A moderation effect plot was created, with RP-DD total score on the x-axis and MMAS-8 total score on the y-axis ( Figure 3 ). As shown in the figure, irrespective of self-efficacy grouping, findings indicate a propensity for risk perception to correlate positively with adherence to medication protocols; Nevertheless, the high self-efficacy group demonstrated a markedly steeper gradient in the trend line compared to their low self-efficacy counterparts. This indicates that higher self-efficacy enhances the positive predictive effect of risk perception on medication adherence. Conversely, when self-efficacy is lower, this positive effect is diminished.

Quantification of adherence levels among study participants was accomplished through the application of the 8-item Morisky Medication Adherence Scale (MMAS-8). The maximum possible score on the MMAS-8 scale is 8 points; however, the median score among patients in this study was only 3.75 (3.50, 5.50). Defining good adherence as a score of ≥6 points, only 13.8% of the patients in this study met this criterion, which is significantly lower than the levels reported in other studies. For instance, in the study conducted by Wang Jie et al. (26), 73.1% of T2DM patients met the criterion for good adherence, while in the studies by Yu Yebo et al. (27) and Wu Ping et al. (28), the rates were 58.7% and 53.0%, respectively. These data suggest that, compared to general T2DM patients, GDM patients exhibit a significant deficiency in postpartum medication adherence. One possible explanation for this discrepancy might be that individuals might be experiencing the initial phases of the illness and have not yet encountered the typical symptoms and complications as with diabetes, leading to a diminished emphasis on the regularity and consistency of glycemic control treatment.

Based on the RPA framework theory, this study analyzed the impact of diabetes risk perception and self-efficacy on the postpartum medication adherence of patients with gestational diabetes mellitus (GDM). The results of the Spearman test showed that both the risk perception of future diabetes (r = 0.778, P < 0.001) and self-efficacy (r = 0.631, P < 0.001) of GDM patients were positively correlated with medication adherence. The results of further hierarchical linear regression indicated that self-efficacy played a moderating role in the impact of risk perception on medication adherence. Jones et al. (24) found that although some groups of people (such as American Indian women) have a high level of disease awareness and risk perception, if they lack self-efficacy, they may still find it difficult to take action, which is similar to the performance of the low self-efficacy group in this study. The study by Anahita Babak et al. (34) on the impact of self-efficacy training on patients with type 2 diabetes mellitus showed that self-efficacy training effectively improved the metabolic control of patients with type 2 diabetes mellitus and, to a certain extent, had good medication adherence.

In addition, previous research results have shown that the factors affecting the medication adherence of diabetes patients include socioeconomic status, the complexity of the treatment regimen, drug side effects, patients’ awareness of the disease and treatment, and the accessibility of medical resources. This partially coincides with the results observed in our univariate analysis, that is, patients with different educational backgrounds, those who received or did not receive postpartum health education, and those using different treatment regimens showed significant differences in medication adherence (P < 0.05). However, in the multivariate analysis, the impact of these three patient characteristics on medication adherence was not statistically significant. It may be because the effects of educational background, postpartum health education, and treatment regimen on medication adherence were masked by the effects of risk perception and self-efficacy. It is necessary to further examine the differences in risk perception and self-efficacy among patients with different characteristics and gain an in-depth understanding of the significance of these differences for improving patients’ medication adherence and optimizing blood glucose management.

This study still has certain limitations: (1) The sample size included in the study is relatively small (80 cases); (2) The single-center research design implies that the geographical distribution of the study population is relatively concentrated, and the conclusions should be interpreted with caution when extended to populations in different regions or ethnic groups; (3) Although a pre-experiment was conducted to optimize the questionnaire design, it is still impossible to completely eliminate the bias caused by participants’ poor understanding of the questions.