This was an observational cross-sectional study.

The study was conducted in 10 of the 11 primary health care (PHC) clinics in the Maruleng sub-district, Mopani District of Limpopo province, South Africa. These clinics offer a range of PHC services, including the diagnosis and treatment of commonly occurring communicable and non-communicable diseases, family planning services and immunisations. These services are provided by nurses. A medical officer consults at the clinic weekly. For people with DM, the package of services includes medication collection, patient education and scheduled diabetic check-ups. These clinics do not provide any eye health services. People requiring eye care services are referred to the optometry clinic at Sekororo District Hospital. Maruleng is a largely rural sub-district with a population of approximately 128 137 people.¹³ The majority of people residing in this sub-district are of African descent, contributing 96% of the total population.¹⁴ The most spoken language is Sepedi (89.7%). The sub-district comprises 23 rural villages and one urban area.¹⁴

The study included people with type 2 DM attending the 10 primary care clinics in Maruleng sub-district. The eleventh clinic that was excluded is a gateway clinic within Sekororo District Hospital, as patients attending this clinic tend to have easier access to services provided at the hospital than patients attending any of the other 10 primary care clinics. Patients under the age of 18 and those diagnosed with gestational diabetes were excluded. As the clinics did not have a list of patients expected per day, the principal investigator (PI) screened patients in the waiting area to see whether they were eligible to participate in the study. Thereafter, based on the headcount, every third patient at the busier clinics and every second patient at the quieter clinics were selected to spread sampling throughout the day. Data collection was done from Monday to Friday at each clinic.

The ideal sample size was calculated with Cochran’s single population formula, in which Z is the z-score representing the desired confidence level (here it is equal to 1.96 for a 95% confidence interval [CI]), p is the estimated population proportion, which was set at 0.5 when unknown, and e is the margin of error expressed as a 0.05 decimal.¹⁵ Thus, the required sample size (n) of this study is 384 (Equation 1):

The sample size was split across the 10 clinics proportional to the headcount at each clinic.

The data were gathered utilising a self-reported structured questionnaire with four sections: Sociodemographic characteristics and KAP. The questionnaire was adapted from a previously validated research instrument to measure KAP regarding DR among peoples with DM.¹⁶ A context-specific adaptation of this questionnaire included removing some demographic items, rephrasing knowledge items and adding new practice and attitude items. Face validity was achieved through expert review. A pilot study was conducted among 10 patients at the Tzaneen primary care clinic in a different sub-district to confirm content and face validity. The Pearson’s correlation score was 1.00 for all items. We calculated Cronbach’s alpha to assess the questionnaire’s internal consistency. The results showed the alpha scores of 0.8 for knowledge, 0.6 for attitudes and 0.7 for practices, indicating a reliable questionnaire. The questionnaire was created in English and translated into Sepedi, the local language spoken in the area. The Sepedi version was then back-translated to English to ensure its original meaning was preserved.

Knowledge and practices were assessed with binary responses (yes/no), and attitudes were assessed using a 5-point Likert scale (strongly disagree to strongly agree). The primary research data were collected through a paper-based survey conducted over 3 months during various clinic visits in 2024. Prior to the survey, participants were instructed to respond honestly to each question after providing their consent. For individuals who were illiterate, informed consent was read out to them, and an ink stamp was used to obtain their right thumbprints as a form of consent, and the PI read the questions and recorded the patients’ responses. Each participant received a unique identification number, and to maintain confidentiality, no personal information was collected from them.

The data were initially entered in a Microsoft Excel spreadsheet (Microsoft 365, Microsoft Corp, Redmond, Washington, United States) and exported to Statistics and Data Analysis (Stata) software edition 15 (StataCorp, College Station, Texas, United States) for analysis. Knowledge and practice responses were assigned 1 point for correct answers and 0 points for incorrect answers, and a total was calculated for each section per patient. Similarly, for attitude statements, favourable responses were assigned 1 point, and neutral or unfavourable responses were assigned 0 points. We performed the Shapiro–Wilk test to assess the normality of the data distribution normality. After confirming non-normal data distributions, the median and interquartile range were used for continuous variables with skewed distributions. Because of the asymmetrical distribution of the KAP scores, the median score, representing the 50th percentile, was used as a threshold to categorise KAP as good or positive (equal to or above the median) and poor or negative (below the median). Percentages and frequencies were used to display categorical data. We used the chi-square (χ²) test with a significance level of p < 0.05 to evaluate the association between demographic variables and KAP scores. Bivariate and multivariate logistic regression analyses were utilised to identify the determinants of KAP; odds ratio (OR) was reported with 95% CI, which aided interpretation of the findings. A stepwise selection strategy was used to select important variables for the analysis model. A statistically significant adjusted odds ratio (AOR) was defined as p < 0.05 in this analysis. Variables that lacked statistical significance in the bivariate analysis but had been found in previous studies to influence patients’ KAP were also included in the multivariate analysis. Highly related variables because of collinearity were excluded from multivariate analysis.

Ethical clearance to conduct this study was obtained from the Biomedical Research and Ethics Committee (BREC), University of KwaZulu-Natal (No. BREC/0000631172023). The World Medical Association’s Declaration of Helsinki, which provides a fundamental framework for ethical research, was upheld by this study, which complied with all ethical requirements for medical-related studies involving human subjects. The National Department of Health, the Provincial Health Research Committee under the Limpopo Department of Health (LP_202404_003) and the Mopani District authorised the gatekeeper. Participants’ information was kept confidential and not shared with third parties without permission, ensuring protection from unauthorised access or loss. Every participant gave their informed written consent.

A total of 389 people living with diabetes voluntarily participated in the study. Table 1 presents a summary of the sociodemographic characteristics of the participants. The median age was 62 years (interquartile range [IQR]: 55–69 years), and the median duration since diagnosis of DM was 7 years (IQR: 2–14 years). There was no significant difference in the median age between males and females (results not shown). Most participants were female (n = 255; 65.6%). More than half of the participants (n = 231; 59.4%) were pensioners, and 103 (26.5%) had no education.

The median knowledge score was 80% (IQR: 40–100). Approximately 268 (61.2%) participants had a score above the median, indicating good knowledge of diabetic eye complications. More than two-thirds of participants (n = 277, 71.2%) reported that eye problems because of diabetes can be treated, and just over half of the participants (n = 233, 59.9%) were aware that they should undergo an eye examination (Table 2).

The median attitude score was 85.7% (IQR: 71.4–100), with the majority of patients (n = 335, 86.1%) having a score above the median, indicating a positive attitude towards diabetic eye complications. More than three-quarters (n = 301, 77.4%) of the patients disagreed that there are more important issues to worry about than the eye-related problems caused by DM. Eighty-four patients (21.6%) agreed that they do not go for an eye examination because no one told them, and 44 (11.3%) agreed that they do not go for an eye examination because the hospital was far from where they live (Table 3).

The median practice score was 42.9% (IQR: 28.6–71.4). Only 216 (55.8%) of the patients scored above the median, indicating good practices towards DR prevention. Of the 115 people living with DM who missed their treatment, 56 (48.7%) skipped a dose for just 1 day. Of the 183 patients (47%) who reported having visual problems, 148 (80.9%) informed their doctor or nurse about the issue; 90 (23.1%) of these patients were subsequently referred to the nearest optometrist for an eye examination and further management, 76 (19.5%) were treated with spectacles and eye drops by optometrist and 16 (4.1%) reported that nothing was done to address their visual problems. More than 50% of patients had never undergone an eye examination, and 110 (28.3%) reported visiting an optometrist without referral (Table 4).

Patients with educational background including primary and above (secondary and tertiary educations) were twice more likely to know about diabetic eye complications compared to those without any education, with an AOR of 1.9; 95% CIs: 1.1–3.3; p < 0.05 and AOR: 2.2; 95% CI: 1.3–3.9; p < 0.05, respectively. Additionally, patients diagnosed with DM for more than 5 years were more likely to have good knowledge than those diagnosed less than 5 years ago (AOR: 1.9; 95% CI: 1.1–2.9; p < 0.05) (Table 5).

In multivariate analysis, patients with good knowledge of diabetic eye complications were twice more likely to have positive attitudes towards the prevention of diabetic eye complications than those with poor knowledge (AOR: 2.0; 95% CI: 1.3–3.1; p < 0.005) (Table 6).

Patients diagnosed with DM more than 5 years were twice more likely to attend regular eye examinations than those diagnosed less than 5 years ago (AOR: 2.3; 95% CI: 1.5–3.6; p < 0.005) (Table 7).

The authors suggest adopting a system-level approach that uses mobile units to reach patients in rural communities or underserved areas. This strategy should involve a multidisciplinary team, including eye care professionals such as optometrists, as part of the diabetes management team. The use of mobile fundus cameras at the primary care level could be a highly effective strategy for accelerating retinal screening among people living with diabetes. Training PHC nurses to capture high-quality retinal images can bring screening directly to underserved or rural areas, thereby addressing barriers such as transportation and long appointment wait times. By integrating eye care earlier in the treatment process, we can ensure patients receive consistent, specialised care that addresses knowledge gaps, fosters positive attitudes and promotes best practices related to DR. Healthcare professionals can empower patients to take proactive measures to prevent vision loss and maintain their eye health.