The International Diabetes Federation estimated that in 2021, in South Africa, 4.2 million adults, aged between 20 and 79 years, will have diabetes, with almost 100,000 diabetes-related deaths annually [1]. The number of adults with diabetes is expected to reach 5.4 million by 2030.

There are several reasons for this increase in prevalence, mainly driven by type 2 diabetes. The risk of type 2 diabetes increases with age. However, the most important risk factor for diabetes is obesity [2]. In South Africa, the recent increases in urbanization and consumption of high-calorie diets, accompanied by diminished physical activity, have resulted in increased obesity: 69% of women and 39% of men are reported to be overweight or obese [2]. Obesity is estimated to account for 87% of type 2 diabetes cases in South Africa [2].

The South African health care system is already burdened by high rates of HIV infection, AIDS, tuberculosis, noncommunicable diseases, maternal and childhood mortality, and injury-related disorders. The predicted diabetes epidemic, with sequelae of microvascular and macrovascular complications, will further burden the already fragile health care sector and the South African economy [2]. In 2021, the total diabetes-related health expenditure in South Africa is estimated at US $7.2 billion [1]. Diabetes is known to have severely impacted the South African economy [3]. Indeed, about two-thirds of diabetes-related deaths occurred in working-class people younger than 60 years.

The occurrence of diabetes-related complications severely impacts patient health and quality of life, reduces productivity, and incurs high costs. The risk of developing micro- and macrovascular diabetes-related complications is known to increase with increasing glycated hemoglobin (HbA_1c) levels. Microvascular complications increase above a threshold of 6.5%, and macrovascular complications increase above a threshold of 7% [4].

In South Africa, many people with diabetes do not receive optimal treatment [5-7]. A retrospective cross-sectional study reported that, in 2013, only 15.5% of people with type 2 diabetes reached the recommended target HbA_1c of less than 7% [8,9]. Furthermore, in the South African cohort of the International Diabetes Management Study, the 899 patients with type 1 or type 2 diabetes had a mean HbA_1c of 8.2%, with only 30% of patients reaching an HbA_1c below the 7% target [6]. The international real-world DISCOVER study program was a prospective observational study that assessed diabetes management in 15,992 patients with type 2 diabetes in 38 countries [10]. The South African cohort of the DISCOVER study had a mean HbA_1c of 9% at baseline, 8.2% after 6 months of follow-up, 8.4% after 12 months, and 8.3% after 24 months [5]. In comparison, in the global DISCOVER study, mean HbA_1c was 8.3% at baseline [11]. It is well established that suboptimal treatment increases the risk of diabetes complications [4], reduces quality of life [12], and increases the economic burden [13].

Digital tools, with or without support by a health care professional, have proven to be effective in lowering HbA_1c levels and promoting diabetes self-management [14-16]. These benefits have been confirmed in meta-analyses based on systematic reviews [14,15,17]. However, in South Africa, evidence to support the use of digital tools for diabetes management is scarce.

The real-world Diabetes Nurse Educator (DNE) study was designed to assess the benefit of the MyDiaCare program, combining digital tools (patient mobile apps and a health care professional platform) with DNE support, for diabetes management support in South Africa after at least 6 months of use. Furthermore, an economic study estimated the cost-effectiveness of MyDiaCare for patients with type 2 diabetes in the South African private sector.

The real-world DNE study was designed as an observational, retrospective, multicenter, single-group study. The study was performed on South African patients with diabetes who had participated in the MyDiaCare program for at least 6 months.

The MyDiaCare program combines digital tools with regular DNE face-to-face interventions (See Figure 1).

Patients older than 18 years with type 1 or type 2 diabetes who had participated in MyDiaCare for at least 6 months were eligible for the study. Patients using another telemonitoring program, without complementary health insurance, or participating in another diabetes education program were ineligible. Furthermore, patients with limited or complete legal incapacity were not enrolled.

All eligible patients who presented at any one of the 2 participating health care centers for a routine visit between November 25, 2019, and June 30, 2020, and who agreed to participate in the study were enrolled. At enrollment, the patients performed the only study-specific visit.

During this study-specific visit, patient and diabetes-specific data were collected. In addition, relevant data collected in the MyDiaCare program were pseudonymized and extracted from the digital platform from the date the patient joined MyDiaCare until the date they left the program or at the latest on the June 30, 2020. These data included diabetes-related data (type of diabetes, date of diagnosis, diabetes treatment, concomitant treatments, diabetes complications, and comorbidities), clinical assessment (blood pressure, body weight, and waist circumference), and standard laboratory tests used for diabetes monitoring (HbA_1c, low-density-lipoprotein [LDL] cholesterol, high-density-lipoprotein [HDL] cholesterol, total cholesterol, and triglyceride levels). Furthermore, patients’ diabetic care plans were assessed, including clinical targets (blood pressure, body weight, and waist circumference) and HbA_1c and blood lipid levels targets.

The BMI was calculated from the weight and height data collected. Moreover, patient, physician, and DNE satisfaction with MyDiaCare was assessed using self-administered and study-specific questionnaires.

The outcome measures were the change in HbA_1c and cardiovascular risk factors (LDL, HDL, total cholesterol, triglyceride, systolic blood pressure [SBP], and diastolic blood pressure [DBP] levels, as well as body weight and waist circumference). Furthermore, the proportions of patients that achieved their targets for these parameters, the adherence to the care plans, and the satisfaction of the patients, physicians, and DNEs with MyDiaCare were also assessed. Adherence was defined as the percentage of tests or consultations performed over the tests planned, adjusted to the duration of follow-up for each patient. Patients’ satisfaction was assessed at least 6 months after joining the MyDiaCare program.

The data are described using descriptive statistics. Quantitative variables are presented as means and SD and medians with IQR. While qualitative variables are presented as frequencies with percentages. The numbers of missing data are indicated. Missing data were not replaced. Analyses were performed using SAS (version 9.4; SAS Institute).

The study was approved by the Research Ethics Committee of the South African Medical Association (registration number 1927/000136/08 NPC). All patients provided written informed consent before participating in the study. Furthermore, the study was conducted in compliance with Good Clinical Practice (International Council for Harmonization E6, the current version), the provisions of the Declaration of Helsinki (Fortaleza, Brazil, October 2013), and the applicable local regulatory requirements. The data was processed in accordance with the local regulations, notably the European Union Regulation 2016/679 (General Data Protection Regulation).

Overall, at 2 study centers, a total of 118 patients were included instead of the approximately 200 planned due to the COVID-19 pandemic. One patient, with a follow-up duration of only 3.45 months, was excluded from the analysis. In total, 117 patients were finally analyzed: 93% (109/117) with type 2 diabetes and 7% (8/117) with type 1 diabetes. Details concerning the demographic characteristics as well as diabetes treatments, complications, and comorbidities are summarized in Table 2.

In the 79/117 patients with data at baseline and at 6 months, there was a clinically relevant mean decrease in HbA_1c levels of –0.6% (SD 1.3); median (IQR) –0.5 (–1.2 to 0.1); range (–3.5 to 3.5): from 7.8% at baseline to 7.2% at 6 months after starting MyDiaCare. These 79 patients comprised 8 patients with type 1 diabetes and 71 patients with type 2 diabetes. At 6 months, the mean HbA_1c level decreased from 8.2% to 7.4% in patients with type 1 diabetes and from 7.8% to 7.2% in those with type 2 diabetes.

Among the 54 of 79 patients with a decrease in HbA_1c levels at 6 months, 80% (43/54) had a clinically relevant decrease of at least –0.4%. The clinically relevant decrease in the mean HbA_1c absolute change was also observed at time points ranging from 3 to 12 months. Moreover, the proportion of patients with HbA_1c levels <7% increased from 30.4% at baseline to 50.6% at 6 months (Figure 3).

Most patients (43/79, 54%) reached or maintained their HbA_1c target levels and a further 20% (16/79) had improved HbA_1c levels but failed to reach their HbA_1c target (Figure 4).

At 12 months, there was a slight deterioration in the proportion of patients that reached their HbA_1c targets. However, only half of the patients had HbA_1c data at 12 months.

Concerning targets related to cardiovascular risk factors, more than 50% of patients reached or maintained their targets for the following parameters: 69% (49/71) for LDL cholesterol, 58% (41/71) for HDL cholesterol, 83% (59/71) for total cholesterol, 67% (53/79) for SBP, and 89% (70/79) for DBP. In contrast, less than 50% reached or maintained their targets: 46% (32/70) for triglycerides, 17% (13/76) for body weight, and 18% (12/68) for waist circumference. These data were not integrated in the economic evaluation.

Regarding adherence to the MyDiaCare care plan, a patient-specific schedule for laboratory tests: HbA_1c, lipid tests, MAU (determined by ACR), for consultations: DNE, doctor, and dietetic consultations, and for examinations: eye and foot screenings. The mean overall adherence to the MyDiaCare care plan was 93%; 86% (100/117) of patients adhered to more than 70% of the scheduled tests and consultations in their care plan. The mean adherence to each test and consultation was between 98% and 100%, including physician and DNE consultations. However, for dietetic consultation and eye and foot examinations, the mean adherences were 59%, 57%, and 44%, respectively.

From the patients’ perspective, most patients (87/117, 74%) were satisfied with their diabetes management, and 85% (99/117) agreed that MyDiaCare improved their diabetes management. The program allowed 62% (72/117) of patients to better understand their diabetes. Furthermore, most patients (105/117, 90%) noted that the presence of the DNE was useful or very useful. Finally, when patients completed the study-specific survey after participating for at least 6 months in the MyDiaCare program, 84% (98/117) wanted to continue using the program, and 89% (104/117) would recommend it to other patients.

All investigators reported that they were satisfied with the patients’ management, that they would use MyDiaCare long term, and that they would propose the program to at least 50% of their diabetes patients. They all felt that the program allowed them to better anticipate the patients’ complications and empowered their patients.

The 2 participating DNEs, 1 at each site, agreed that the program made them feel part of a medical team and allowed them to better anticipate patient complications. Furthermore, they found the tracking platform user-friendly, with the patients’ data easily available.

In the DNE study, after 6 months of using the MyDiaCare program, the mean HbA_1c levels decreased by a meaningful –0.6%. from 7.8% at baseline to 7.2% at 6 months. This relevant decrease of –0.6% is substantial considering the low HbA_1c baseline level.

This decrease in mean HbA_1c levels is comparable to what has been reported in systematic reviews and meta-analyses after 3 to 10 months using mobile health interventions or telemedicine [14,15,17]. In the South African cohort of DISCOVER, the mean HbA_1c level was 9% at baseline and decreased to 8.2% after 6 months [5].

Several studies have investigated digital tools with limited or no intervention by a diabetes health care worker [16,31,32]. In South Africa, where health literacy is expected to be low, it is critical that digital tools be accompanied by regular health care visits, particularly DNE and physician visits. Indeed, a South African study conducted at a single institution found that knowledge concerning diabetic foot among patients with type 2 diabetes was suboptimal [24].

Regular and frequent visits with DNEs, in between visits with physicians, might play a role in decreasing HbA_1c. The RENEWING HEALTH randomized controlled trial evaluated the benefit of a mobile health intervention with or without health counseling in patients with type 2 diabetes [31]. In the control group, there was only 1 telephonic counseling session by a DNE (4 months after baseline) compared to more frequent DNE face-to-face visits (every 3 months) in the health counseling group. In the Norwegian cohort of this trial, at 4 months, the HbA_1c levels were reduced on average by –0.23% with the mobile app, –0.41% with the mobile app and DNE counseling, and –0.39% with standard of care. These results are comparable with the –0.6% reduction in HbA_1c levels observed in this study.

Other studies that assessed mobile health interventions showed larger decreases in HbA_1c levels compared to this study. However, these larger decreases were to be expected since baseline HbA_1c levels were substantially higher than those in this study. After 3 months of using a mobile app with integrated coaching by a diabetes educator, the mean baseline HbA_1c level of 9.87% had decreased by –0.86% in patients that had completed the single-arm study and by –0.96% in those that actively used the mobile app and coaching. Furthermore, among active users, those with baseline HbA_1c levels ≥9% reduced these levels by on average –1.32% [32]. Similarly, after 4 months of using a digital tool comprising remote monitoring and lifestyle changes with coaching by a diabetes specialist, the mean HbA_1c level had significantly decreased by –0.8% from a baseline level of 8.9%. In patients with HbA_1c levels >9% at baseline, the HbA_1c decrease was even higher at –1.4% [16].

At 6 months, nearly 75% (59/79) of patients improved their HbA_1c levels: 54% (43/79) reached or maintained their HbA_1c targets, and a further 20% (16/79) improved their HbA_1c levels. Also, after 6 months of MyDiaCare, 50.6% (40/79) of patients had achieved the recommended HbA_1c level target of 7%, as compared to 30.4% (24/79) of patients at baseline. In comparison, after on average 2 years in the Central Chronic Medicine Dispensing and Distribution program, only 29.2% of patients had HbA_1c levels below 7% [7]. In the South African cohort of DISCOVER, the mean HbA_1c level was 9% at baseline and decreased to 8.2% after 6 months. Both the baseline HbA_1c levels (7.7%) and proportion of patients with HbA_1c levels below the targeted 7% at baseline (30.4%) and after 6 months (50.6%) in the MyDiaCare study were substantially better than those reported [5,7].

In addition to achieving HbA_1c targets, a large proportion of patients also achieved their targets for other cardiovascular risk factors, in particular LDL and HDL cholesterol and blood pressure levels.

It is noteworthy that both patients and health care professionals were satisfied with the MyDiaCare program. The MyDiaCare program provides innovative tools for patients and health care professionals. For patients, the mobile app allows them to self-manage their diabetes. While the health care platform allows physicians and DNEs to monitor diabetes evolution from a distance. In this program, the role of the DNE is critical: providing frequent and regular support for patients and physicians. Overall, MyDiaCare is a program designed to empower patients but also optimize diabetes management.

The real-world DNE study showed the benefit of MyDiaCare in improving HbA_1c levels. To provide further evidence of this metabolic benefit in the long term, we used HbA_1c as a proxy to estimate MyDiaCare’s impact in terms of diabetes complications and costs. Despite the estimated increased cost of diabetes management (health care visits, tests, and investigations), these costs were largely compensated for by the estimated costs avoided by reducing the number of diabetes complications (mainly nephropathy). The model estimated that in the private sector, about 26 million ZAR would be saved in the first year, with the cost of diabetes reduced by 23% compared to current standard of care. The annual cost savings for the private sector per patient during the first year were estimated to be 71,000 ZAR (US $4089). However, when we used the same mean baseline level of HbA_1c in the standard of care (results from a sensitivity analysis), the annual cost savings per patient during the first year were estimated to be 17,652 ZAR (US $1016). These savings are comparable with those previously reported [33,34]. For example, Smith and colleagues [33] evaluated the effectiveness and economic impact of a diabetes education program among adults with type 2 diabetes in South Texas. They found an HbA_1c decrease of –0.8% and a direct cost savings of US $2780 per year and per patient during the program. In another economic study, the annual savings per patient of an integrated care team model for targeted, high-risk Medicare patients with type 2 diabetes were estimated at US $5844 [34]. To our knowledge, no other reported economic study has estimated the cost of diabetes management in the South African private sector. This study suggests that increased diabetes monitoring, as observed in this study, may benefit not only patients and health care professionals but also private insurance.

The DNE study was designed to retrospectively collect real-world data concerning diabetes management in the South African private sector. The data collected therefore reflects current clinical practice. However, the retrospective design means that there are some missing data. In addition, the enrollment period of our study coincided with the COVID-19 pandemic, preventing us from recruiting the planned number of patients. Furthermore, there is no control group to evaluate the specific clinical effects that can be attributed to the MyDiaCare program. The DNE study was only performed at 2 health care centers, with mostly patients with type 2 diabetes enrolled. Thus, the results obtained may not be representative of diabetes management in the whole South African private health care sector. Baseline HbA_1c levels were lower than those already reported in other cohorts in South Africa. Moreover, 32% of patients with type 2 diabetes had reached their HbA_1c target level at baseline. Therefore, the DNE study population seems to correspond to a better-treated population than those reported, representing a selection bias [5-7]. Finally, the study was not designed to collect safety data, in particular cases of hypoglycemia and hyperglycemia.

The budget impact model has several limitations. First, the model is highly sensitive to HbA_1c levels at baseline and at 6 months, since these levels were used to estimate the incidences of diabetes-related complications where the main cost benefits were observed. Indeed, mean baseline HbA_1c levels in the standard of care cohort without MyDiaCare were higher than in the cohort with MyDiaCare. This difference would have impacted the estimated numbers of diabetes-related complications in these cohorts and the overall economic impact. Second, the HbA_1c levels in the standard of care cohort were estimated using the South African cohort of DISCOVER, but these data may not be representative of all patients and centers in South Africa [5]. Third, in the budget impact model, the HbA_1c levels obtained at 6 months were kept constant for the remainder of the model time horizon, and the method used to estimate long-term complications was the same for the first year and for subsequent years. The maintenance of HbA_1c levels after 6 months for the model is a hypothesis based on the continual effectiveness of MyDiaCare program to maintain these levels. Finally, the legacy effect, which is the continual long-term increase in risk despite improved glycemic control, and the difference in cumulative glycemic exposure, estimated by HbA_1c, were also not considered in the economic model.

Diabetes is a major public health challenge in South Africa. The increase in prevalence and related severe complications is expected to impact not only patient health and quality of life but also productivity and the South African economy. Many figures confirm that diabetes management in South African patients is, at present, suboptimal, with low percentages of patients reaching their targets for HbA_1c and other cardiovascular risk factors and adhering to guidelines. The MyDiaCare program, which combines digital tools for patients and health care professionals with DNE support, may be a clinically effective and cost-saving solution for diabetes management in the South African private health care sector. These results need to be confirmed in further studies.