Diabetes is a life-threatening chronic disease that requires effective and sustainable care and treatment. In 2021, it was responsible for 6.7 million deaths worldwide, and the number of people affected was estimated at 537 million, projected to rise to 783 million by 2045 [1]. Despite the increasing burden of diabetes, there is a shortfall of 5.9 million health care professionals required to provide quality care for people living with diabetes [2]. Moreover, many low-income countries face challenges in translating evidence-based knowledge, cost-effective guidelines, and electronic records into actionable solutions to enhance the ability of frontline health workers to deliver quality diabetes services [3].

Health IT has great potential for enhancing diabetes management by saving time and costs involved in data interpretation. Digital knowledge tools can serve as an effective resource for bridging knowledge gaps among health care providers; however, the integrated effect of knowledge tools and diabetes registry on provider performance is not well known [4]. The World Health Organization (WHO) indicates that only 50% of countries use electronic diabetes registries and expects national data to be standardized when registry coverage exceeds 75% [5]. Many low-income countries still struggle to provide comprehensive digital tools for knowledge and registries for frontline health workers. Myanmar should seize these opportunities to improve community-level diabetes care management.

Cost-effective interventions are urgently needed to address the diabetes burden, which cost US$ 966 billion globally in 2021. Southeast Asia’s expenditure was significantly lower at US$ 10.1 billion, compared to North America’s US$ 414.5 billion [6]. Therefore, it is vital to implement high-impact and affordable solutions in low-income countries. A meta-analysis shows that digital tools for diabetes knowledge can be cost-effective [7], yet there is limited evidence in low-income settings [8]. Therefore, it is essential to perform integrated and interdisciplinary research on digital tools for the effective implementation of diabetes control programs in the region.

Myanmar is one of the countries in Southeast Asia with a high burden of diabetes, with a prevalence among 10.5% of the population, which is comparatively higher than that in other countries in the region [9]. A diabetes prevalence survey conducted in 2014 revealed that the burden of the disease had doubled over a decade, and there were no effective strategies or guidelines implemented to raise awareness about diabetes management [10]. Additionally, health workers need to enhance their knowledge, and further research is necessary to improve the quality of diabetes control services at the primary health care level [11].

The rapid growth of mobile technology in Myanmar has created new opportunities for digital health. In 2014, the Ministry of Health established an electronic health management information system and a real-time District Health Information System for all townships. They also distributed 26,000 tablets with essential guidelines for frontline health workers [12]. However, the digital health information system is still in its early stages, and no specific digital application for the diabetes control program exists.

Unlike other studies, this study aimed to evaluate the novel integrated effects of digital tools on diabetes knowledge and registry in relation to diabetes control performance among frontline health care workers in Myanmar. The primary outcomes of the study were the knowledge and performance levels of the health workers, while the secondary outcome was the cost-effectiveness of the intervention.

Researchers designed the study based on two main theories: attribution theory, which examines how knowledge affects diabetes management [13]; and an economic principle assessing the cost-effectiveness of digital health technology [14]. The research hypothesized that integrated digital tools would enhance frontline health workers’ knowledge and performance by reducing diabetes program costs.

A quasi-experimental study was conducted in Naypyitaw, the capital of Myanmar, between October 2022 and April 2023 (spanning 6 months). Two townships were selected for the intervention and control groups based on matched population characteristics, geographical conditions, and access to essential diabetes control packages offered by the diabetes control program. The selection criteria for the study areas included a high unknown prevalence of diabetes among the population and inadequate knowledge among health workers [15]. Baseline and endline assessments were carried out for both the groups.

Frontline health staff, including midwives, lady health visitors, and public health supervisors, were selected for the study based on specific inclusion criteria: involvement in the diabetes control program, ability to use digital tools, and willingness to participate. Exclusion criteria included those absent for over 1 month, nearing retirement, deemed unfit for intervention, or not approved by supervisors. Using G*Power software (version 3.1.9.2; Heinrich-Heine-Universitat Dusseldorf), the sample size was calculated for multiple linear regression with a 95% CI and power. The reference minimum effect size of the intervention on diabetes control was F_4,75=.25 [16]. The minimum sample required was 86 participants, and 96 were recruited to account for dropouts, finally assigning 48 participants to each group.

The new intervention involved integrating the two new digital tools developed by the authors, Myanmar Diabetes Guides and Digital Registry. This is unique and significant because other studies measure the silo effects [17]. Researchers installed digital tools and provided 3 days of intensive training on how to use the tools. Additionally, researchers conducted monthly follow-ups and provided reorientation sessions to the intervention group.

The researchers developed self-administered questionnaires for data collection. The questionnaire was created in English and subsequently translated into the Burmese language. It included a scoring system for the primary outcomes, which assessed the knowledge and performance in five key domains of diabetes management: (i) health promotion, (ii) diabetes screening, (iii) care and treatment, (iv) referral, and (v) reporting, all referencing the WHO and National Diabetes Guidelines. The researchers established an expert panel to review the questionnaire to ensure content validity. This panel included a diabetes program manager, an expert clinician, and township health officers. The validity index for the questionnaire was scored at 0.6. Additionally, the reliability of the questionnaire was pretested by two different observers, with an interrater kappa value of 0.68 between the two observers.

The WHO-CHOICE (WHO’s Choosing Interventions that are Cost-Effective) Analysis tool for noncommunicable diseases was used to measure costing data [20]. This costing data included direct costs (intervention costs, program costs, and treatment costs) and indirect costs (communication, consulting, value of time, and work). The cost-effectiveness analysis aimed to demonstrate the economic benefits of the intervention for future investments.

The principal investigator and two research assistants collected baseline and endline data. Before data collection began, the study’s purpose was explained to authorities and participants. Primary cost data were obtained from participants, while secondary data on treatment costs came from the township hospital and program costs from the township health departments. The team checked the accuracy of the questionnaires, addressed any missing responses, and cross-verified participant data with the secondary data. Health assistants from the townships were followed-up monthly on using the tools.

Data analysis was conducted using SPSS software (version 22.0; IBM Corp). Knowledge and performance scores followed the National Diabetes Control Guideline criteria. Sociodemographic data differences were assessed with the χ² test, mean outcome data with the t test, and outcome correlations with the Pearson correlation test. The impact of the intervention was evaluated using multiple logistic regression analysis. All tests were statistically significant at a 95% CI. Cost-effectiveness was assessed through the cost-effectiveness ratio and incremental cost-effectiveness ratio.

The study received approval from the Chulalongkorn University Ethics Review Committee (090.2/64, COA No. 177/2022). The Ministry of Health, Naypyitaw Department of Health, granted permission for data collection (NPT/NCD/007-2021/5925). Participation in the research was voluntary; informed consent was obtained, and data confidentiality was ensured. During the registration, an internet package (US$ 2) and a 50 pcs box of surgical masks were provided to participants as compansation for participating the research.

Ninety-six participants enrolled in the study, and 1 patient dropped out. Figure 1 shows the flow chart of the quasi-experimental study design.

The study used the χ² test to examine participant characteristics in relation to the sociodemographic variables, work-related variables, and variables related to diabetes control (Table 1). Significant differences between the two groups were observed based on gender (P=.02), job designation (P=.02), distance from the township (P=.001), and duration of internet use (P=.001). Independent t tests analyzed differences in mean knowledge and performance scores. The mean knowledge scores were statistically associated with gender (P=.001), job designation (P<.001), and diabetes control training (P=.02). The mean performance scores were statistically associated with gender (P=.04). job designation (P=.009), diabetes control training (P=.02), diabetes registry training (P=.02), number of postings (P=.03), level of facilities (P=.02), and experiences with diabetes campaigns (P=.007).

An average of 94.2% of the intervention group (45/48) regularly utilized diabetes knowledge tools. During the baseline assessment, no significant difference in the mean knowledge scores was noted between the two groups (P=.20). However, a significant difference emerged between the two groups at the endline (P<.001). The mean (SD) knowledge score in the intervention group rose significantly from 85.04 (9.73) to 99.25 (5.33; P<.001). In contrast, the mean (SD) knowledge score in the control group slightly declined from 83.58 (11.17) to 80.47 (16.99) (Figure 2); however, this change was not significant (P=.21; Table 2). The multiple linear regression model, after adjusting for potential confounding factors, showed a significant effect, with an adjusted r² of 0.43, an unstandardized β coefficient of 17.769, and a standardized β coefficient of 0.569 (P<.001; Table 3).

A total of 91.2% (44/48) of the intervention group regularly used the registry tool and registered 1747 diabetes patients within 6 months. The mean (SD) performance score in the intervention group significantly increased from 71.22 (28.35) to 107.16 (15.62; P<.001). In contrast, the control group experienced a decrease in the mean (SD) performance scores from 73.29 (35.66) to 70.04 (29.72; P=.22). At baseline, the two groups had no significant difference in the mean performance scores (P=.75). At the endline, a significant difference was noted between the two groups (P<.001; Table 1). A multiple linear regression model, adjusted for potential confounding factors, indicated a significant effect, with an adjusted r² of 0.642, an unstandardized β coefficient of 33.143, and a standardized β coefficient of 0.554 (P<.001; Table 4).

The study showed a significant positive correlation between total knowledge and performance scores, with a correlation coefficient of r=0.45 (P<.001) at the endline (Figure 3). A significant correlation was found in the intervention group compared to the control group, with r=0.34 (P=.02). However, no significant correlation was observed in the control group, with r=0.02 (P=.89)

Researchers categorized the cost data into three main categories: (1) intervention costs, which included web application development, training, internet usage, and stationery; (2) program implementation costs, covering travel expenses, costs for consultation and communication, loss of income due to diabetes-related work, and other miscellaneous costs; and (3) treatment costs for complications. The total cost for the intervention group was 22,213,000 Kyats (US$ 10,586.71), while the control group incurred a total cost of 22,779,000 Kyats (US$ 10,856,4714) (Table 5). The intervention was deemed cost-effective, with a cost-effectiveness ratio of 0.711, and it was considered cost-effective when the cost-effectiveness ratio was less than 1. The study also analyzed the incremental cost-effectiveness ratio for comparative investment. The incremental cost-effectiveness ratio indicated saving 10,127.04 Kyats (US$ 4.83) for both outcomes, 30,154.50 Kyats (US$ 14.40) for knowledge, and 15,247.84 Kyats (US$ 7.27) in performance. The visibility of cost-effectiveness was demonstrated by plotting bootstrap results against outcomes and expenses (Figure 4).

The intervention of new integrated digital tools saved costs and yielded significant positive outcomes in knowledge and performance of diabetes control among frontline health workers. Although the study is quasi-experimental, the researcher minimized confounding by matching selection criteria and adjusting covariates through multiple linear logistic regression. Consequently, the study achieved its goals through a consistent design and reliable analytical methods, resulting in valid outcomes. The overall results align with other research on digital solutions to enhance diabetes control [21].

According to this study, despite several benefits, the rollout and sustainability of the digital diabetes registry encountered some limitations. Initially, the studies intended to measure baseline, midterm, and endline assessments. However, the authority approved only two measurements based on the country’s political situation and the expectation of no significant variation in the midterm. Although the initial goal was to collect real-time data, health staff could only upload information monthly due to being overburdened with competing priorities. The study focused on Naypyitaw, which limits generalizability to the entire country, and only has a 6-month duration, so it cannot evaluate the long-term impacts. The study population focused solely on public providers, excluding private providers, patients, and the wider community. Furthermore, the digital tools were not interoperable with the District Health Information System.

The intervention used a multidisciplinary approach for frontline health personnel at the grassroots level, significantly improving knowledge and performance and reducing program costs. Unlike other studies, this research demonstrated the integrated and correlated effects of digital knowledge and reporting tools. Given these strengths and limitations, researchers recommend that policy makers replicate the intervention nationwide, develop clear standard operating procedures, establish a reporting schedule, and provide an internet data package to enhance the use of digital tools. Furthermore, the diabetes registry operates in isolation, necessitating the creation of an interoperable system to connect with the District Health Information System. Additionally, extensive studies on long-term population research and economic evaluations are essential to evaluate the sustainability of digital tools. We suggest engaging outstanding community nurses as champions to share their best practices of digital applications, and these measures could ensure data quality and sustainability of digital tools to enhance diabetes control in Myanmar and other low-income countries.