Between January 2022 and July 2022, a multi-stage stratified sampling method was employed to select the study subjects. In the first stage, five cities, namely Nanning, Guilin, Hechi, Chongzuo, and Yulin, were randomly selected based on the geographical regions of the Guangxi Zhuang Autonomous Region (central, eastern, southern, western, and northern). In the second stage, three counties were randomly chosen within each of the selected cities, resulting in a total of 15 counties for the survey. In each county, one internal medicine department from a local hospital was selected for the investigation. A total of 152 individuals were surveyed in each hospital, resulting in 2280 patients being surveyed in total.Inclusion criteria were as follows: (1) Conforming to the diagnostic criteria for diabetes established by the Chinese Diabetes Society in 2020 (8); (2) Age ≥ 18 years; (3) Registered and currently residing in the surveyed area. Exclusion criteria were as follows: (1) Patients with gestational diabetes; (2) Patients with a known history of dementia or other mental disorders based on medical records or information obtained from family members; (3) Patients in the acute phase of their illness, unable to cooperate with the investigation. Written informed consent was obtained from all participants, and informed consent was provided by the administrators of all hospitals participating in this study.

The hospital where the authors are affiliated received approval from the Ethics Committee (Approval Number: KT-KJT-2021-26). The questionnaire outlined the voluntary and confidential nature of participation, and all hospital administrators and participants provided informed consent.

This study involved questionnaire surveys, blood collection at local hospitals, and assisting patients in filling out standardized questionnaires. Researchers recorded general information and glycated hemoglobin values for diabetes patients in 15 hospitals and entered the results into Excel, which were then transmitted to us via the internet. Upon completion of the questionnaires, they were also sent to us electronically. We compiled data from the 15 hospitals and stored it in Excel spreadsheets. All data were independently verified and input into SPSS by two researchers. All information is stored on a secure computer to prevent data leakage.

This study employed SPSS 22.0 and Mplus 8.3 software for data statistical analysis and processing. (1) The health literacy scores of rural diabetes patients, treated as a continuous variable, were used as the manifest variable for latent profile data processing with Mplus 8.3 (13). The initial model started from “1” and gradually increased the number of latent categories until the best model fit was achieved. Fit indices included: (i) Akaike Information Criteria (AIC), Bayesian Information Criteria (BIC), and adjusted Bayesian Information Criteria (aBIC), with smaller values indicating better fit; (ii) Entropy, which represents classification accuracy, and when Entropy > 0.8, it indicates classification accuracy > 90%; (iii) Model comparisons were primarily based on the Bootstrap Likelihood Ratio Test (BLRT) and Likelihood Ratio Test (LMR). When the p-values corresponding to LMR and BLRT reached a significance level, it indicated that the model had a better fit than the previous one. This study comprehensively evaluated the fit of various category models and selected the best model.(2) Data were analyzed using SPSS 25.0 software. For quantitative data that followed a normal distribution, descriptive statistics were presented as x ± s (mean ± standard deviation). For data that did not follow a normal distribution, descriptive statistics were presented as M (Q1, Q3) (median with interquartile range). Qualitative data were described using frequency counts. Multivariate logistic regression analysis was used to determine whether statistically significant differences existed among different latent profile groups concerning their manifest characteristics. Based on these results and considering the normal distribution of the total health literacy scores, different latent profile groups’ health literacy levels were compared using t-tests, one-way analysis of variance, or non-parametric tests such as the Mann-Whitney U test and Kruskal-Wallis H test. All tests were two-tailed, and p<0.05 was considered statistically significant.

A survey was conducted on 2,280 rural diabetic patients, of which a total of 2,178 rural diabetic patients were effectively surveyed. The basic profile of the patients surveyed is shown in Table 1 . The surveyed rural diabetic patients had an average age of (63.25 ± 12.71) years and an average duration of illness of (7.96 ± 4.07) years. The male proportion was 55.28%, indicating a higher representation of males than females. In terms of educational background, the majority had completed junior high school education, accounting for 31.43%. The proportion of patients with per capita disposable annual family income ≥18931 yuan was 37.74%. The proportion of patients with a family history of diabetes was 14.55%, and the proportion of patients with HbA1c control <7.0% was 22.68%.

The fitting statistics for the profile models are shown in Table 2 . According to the entropy, the class 2 model fits the best. And the P-values of both LMR and BLRT were statistically significant, so the class 2 model was chosen as the final model.

The health literacy of diabetic patients in rural areas is categorized into two potential groups, showing distinct characteristics across three dimensions. Category 1, with a total health literacy score of (39.53 ± 4.795), exhibits relatively higher levels of health literacy across all dimensions. Among these, the functional health literacy dimension scored the highest at (14.72 ± 1.366), thus labeled as the high literacy-high functionality group. Category 2, with a total health literacy score of (22.82 ± 3.768), demonstrated lower scores across various dimensions of health literacy. Particularly, the critical dimension scored the lowest at (6.24 ± 1.280), hence named the low literacy-low criticality group. The scores for each dimension of health literacy for the two categories of rural diabetic patients are detailed in Table 3 .

Comparison of socio-demographic characteristics of different potential categories of health literacy among diabetic patients in rural areas is shown in Table 4 , which shows that the differences between the two groups were statistically significant (p < 0. 05) in terms of educational level (χ² = 83.169, p < 0. 001), per capita annual disposable income of the family (χ² = 64.792, p < 0. 001), the duration of the disease with or without complications (χ² = 10.106, p<0.001), family history of diabetes (χ² = 27.362, p<0.001) the difference was statistically significant (p<0. 05).

The standardized HbA1c rate for all study subjects was 22.68%. The high literacy-high functionality group demonstrated a blood sugar control rate of 39.12% (390/997), while the low literacy-low criticality group exhibited a blood sugar control rate of 9.66% (104/1077). The blood sugar control rate in the high literacy-high functionality group was significantly higher than that in the low literacy-low criticality group (P < 0.001). The results are presented in Table 5 .

The HbA1c standardization rate was considered as the dependent variable, while health literacy served as the independent variable. After controlling for variables such as age, gender, educational level, family per capita disposable income, family history of diabetes, medication usage, insulin use, and presence of comorbidities, a variable screening using the entry method (α = 0.05) was conducted for logistic regression analysis. The results showed that age, gender, presence of medication, insulin use, education level, per capita disposable household income, and health literacy category were the influencing factors of HbA1c attainment rate, (p<0.05),The attainment rate of HbA1c control was higher in males as compared to females (OR:0.764,CI:0.606-0.964). Patients in the high-literacy-high-functionality group may have a higher rate of achieving HbA1c control compared to patients in the low-literacy-low-criticism group (OR: 5.295, CI 4.103-6.827). Patients taking oral hypoglycaemic agents had a higher rate of achieving HbA1c control compared to patients not taking oral hypoglycaemic agents (OR: 2.270, CI: 1.750-2.943). Patients using insulin had a higher rate of achieving HbA1c control compared to patients not using insulin (OR: 2.957, CI:2.338-3.740). Compared with patients with per capita disposable annual household income <18931 yuan, patients with per capita disposable annual household income ≥18931 yuan had a higher rate of achieving HbA1c control (OR: 1.682, CI: 1.331-2.125). The results are presented in Table 6 .

The use of latent profile analysis aids in identifying heterogeneity among rural diabetes patients, facilitating the identification of variations in health literacy levels among individuals. This approach helps pinpoint issues related to health literacy among patients and allows for targeted interventions aimed at improving their knowledge level, thereby serving as a reference for interventions in glycemic control. This study employed latent profile analysis to classify rural diabetes patients into two latent categories based on health literacy: the high literacy-high functionality group and the low literacy-low criticality group. The latter, constituting 54.22% (1181/2178) of the total, exhibits relatively lower scores across health literacy dimensions compared to the overall level. This suggests difficulties in accessing and utilizing health information and inaccurate judgment of health information among these patients. Probable causes include the relatively underdeveloped economic status of Guangxi, situated in the western region of China (12), with rural diabetes patients having lower educational levels and relatively lower incomes (14, 15). Consequently, these patients have limited disease-related knowledge and information acquisition abilities, compounded by long disease durations leading to decreased disease awareness (16). Healthcare providers should encourage these patients to acquire knowledge through various channels and offer personalized or group health education to enhance their understanding and acquisition of health information, especially for those with lower literacy and limited learning abilities. The high literacy-high functionality group comprises 45.78% (997/2178) of the patients, displaying higher scores in all dimensions of health literacy compared to the overall level. Notably, this group excels particularly in the functional aspect of health literacy, indicating their willingness to invest time and effort in understanding diabetes-related health knowledge. This may be attributed to their relatively higher cultural and knowledge levels, enhancing their reading and information utilization abilities. Healthcare providers should encourage patients in this group to foster a strong desire to improve health and enhance their awareness of health.

By exploring the differences in disease-related information and demographic characteristics among different health literacy categories, it was found that patients in the high literacy-high functionality group have higher educational levels and relatively younger ages compared to those in the low literacy-low criticality group, which is consistent with other studies (17–19). Additionally, the low literacy-low criticality group of diabetes patients showed a higher prevalence of comorbid complications, suggesting an inverse relationship between the occurrence of diabetes complications and the level of health literacy. Ajuwon’s findings show that lower health literacy among African Americans is associated with lower self-management behaviours that can lead to complications of type 2 diabetes (20). This is similar to our findings. The results of this study also indicate that patients in the high literacy-high functionality group have higher average monthly family incomes compared to those in the low literacy-low criticality group, consistent with other research (21, 22). This could be attributed to an increased awareness of disease treatment in situations where material needs are met, prompting patients to actively invest in their health, thereby enhancing their health literacy regarding chronic diseases.

Studies (23–25) have indicated a significant negative correlation between health literacy among diabetes patients and glycemic control. In this study, the high literacy-high functionality group exhibited the highest number of patients meeting blood sugar control standards, while the low literacy-low criticality group showed fewer patients meeting these standards, highlighting differences in glycemic control between the two categories of patients. Logistic regression analysis revealed that the diverse latent profiles of health literacy among rural diabetes patients are influencing factors in glycemic control. Patients in the high literacy-high functionality group exhibited significantly better glycemic control compared to those in the low literacy-low criticality group. Achieving glycemic control reflects an individual’s spontaneous and multi-dimensional health-promoting behavior aimed at maintaining or improving their own health status (26). A higher level of health literacy in diabetes patients is associated with a better understanding of disease-related knowledge, improved treatment adherence, and a higher level of glycemic control (27). Therefore, healthcare professionals can encourage patients to learn and acquire knowledge related to diabetes. For diabetes patients in remote rural areas of Guangxi, interventions such as establishing platforms for diabetes knowledge acquisition and exchange could be implemented to enhance patient health literacy, consequently elevating their level of glycemic control.