The Role of Digital Health Technology Interventions in the Prevention of Type 2 Diabetes Mellitus: A Systematic Review

Objectives: Diabetes in the 21st century presents one of the greatest burdens of disease on the global population. Digitally mediated interventions have become imperative in alleviating this disease epidemic. We aimed to systematically review randomized controlled trials (RCTs) on different health technologies for preventing Type 2 diabetes mellitus, and their efficacy in decreasing diabetes risk-related outcomes in at-risk patients in comparison to standard care. Methods: Five electronic databases were searched between October 2021 and December 2022. Studies including digital health technology interventions used for preventing diabetes development by reducing diabetes risk-related outcomes in at-risk adults (⩾18 years) were identified. Data on glycemic levels, incidence of T2DM, weight, and intervention descriptions were extracted, and the risk of bias (ROB) was assessed. Results: Nine studies met the inclusion criteria and 5 studies (56%) achieved clinically significant outcomes in at least one of the following: decreased weight (22%), glycemic levels (22%), or incidence of T2DM (11%). Two of the 3 (67%) computer-based interventions effectively reduced the HbA1c levels and mean weight of their study population, and 3 of 6 (50%) mobile based interventions (text messages, mobile app, and telehealth) decreased the incidence of T2DM and HbA1c levels. Four studies each had an overall low ROB and one had a high ROB due to attrition. Conclusion: Preliminary evidence identified in our review demonstrated that health technologies for diabetes prevention are effective for improving diabetes risk-related outcomes. Future research into digital technology protocol and studies of longer duration and more diverse populations are needed for clinical feasibility.


Introduction
Type 2 Diabetes Mellitus (T2DM) is defined as a metabolic disease involving defective insulin action. 1 The resulting hyperglycemia places individuals at high risk for all-cause mortality by health expenditure of around 966 billion USD. 2 An estimated 50% of diabetes cases globally remain undetected or undiagnosed 3 and over half of diabetes' economic cost lies in its treatment. 2Diabetes prevention programs (DPPs) are increasingly 2 Clinical Medicine Insights: Endocrinology and Diabetes crucial in mediating this exponential disease epidemic.With healthcare systems' gradual digitalization, the mediation of DPPs through technological platforms can aid in translating and disseminating these preventative interventions into clinical and public health settings.
Diabetes prevention is traditionally grounded in behavioral change theory, where lifestyle modifications are normalized through educational curriculum, health coaching and peer support in individual and group settings. 46][7][8] They yield long-term results at follow-up with a T2DM reduction of up to 43% that persists for decades for particular populations, 5,6 but with the growing global incidence of diabetes, this evidence requires translation into a multifaceted systems approach. 9In a public health setting, literature indicates the need to develop local health economies that facilitate communication and accessibility between the different levels of health care. 9At the patient level, barriers to accessibility and adherence need to be overcome. 1,10At the clinical practice level, the increasing prevalence of diabetes-related complications 9 requires a shift of patient management from episodic primary physician checks to a chronic care model. 11At an administrative level, the economic burden on public health systems can be ameliorated by socio-structural change via health policies, health education 12 and the development of supportive community environments. 13igital health technology refers to the use of digital and information technologies, such as mobile devices, wearables, health information systems, software applications, and telehealth services, to improve the delivery and management of healthcare. 14,15It covers a wide range of tools and solutions crafted to improve the efficiency, accessibility, and quality of healthcare services while empowering individuals to actively engage in their health management.The origins of digital health technology can be traced back to the integration of computers into healthcare settings during the mid-20th century.Initially used for administrative tasks, these technologies gradually expanded to encompass clinical and patient care. 16he rapid evolution of digital health technology was further fueled by the advent of the internet and the widespread use of personal computing devices.Over the past few decades, there has been a notable surge in the development and acceptance of digital health solutions.The widespread adoption of smartphones, wearable devices, and the availability of high-speed internet have been pivotal factors contributing to this growth. 17igital health technologies now encompass a diverse array of applications, including electronic health records (EHRs), telemedicine, mobile health apps, remote patient monitoring, health informatics, and the utilization of artificial intelligence for healthcare analytics. 18,19rman 20 identified the key shared features accounting for the success of major DDP trials.These include individualization of patient interventions, flexibility between group and individual settings, a mixture of physical and behavioral activities and an "extensive network of training, feedback and clinical support." 1 Evidence indicates that health technologies fulfill all these criteria by extending diabetes prevention from clinical settings to creating an individual's daily health-protective environment. 10Health technologies act as intermediary platforms that transfer information between patient and provider whilst normalizing behavioral interventions and lifestyle modifications in the patients' surroundings and communities.
Regarding computer-based interventions, there is little evidence of its impact on diabetes prevention as software focuses on known diabetes and population management between various health providers, remote patient monitoring and diabetes self-management. 21A Cochrane meta-analysis comparing 16 computer-mediated diabetes education programs demonstrated a small decrease in HbA1c in patients diagnosed with T2DM and has potential in translation to preventative interventions. 22here is also a paucity of data on the efficacy of digital health technology in diabetes prevention.Past studies that showed evidence of positive correlations between digitally translated DPPs and improved diabetes risk-related outcomes have been limited by their short duration (5-24 months) and the lack of long-term follow-up.These affect the predictive validity of results by not accounting for fluctuating weight and glycemic levels.The heterogeneity of study location and participant demographic and high variability in program procedure affects the feasibility of public health implementation and scalability as study outcomes may not be generalizable to other national or global contexts.Due to these research gaps, this paper aims to systematically review and compare current evidence about digitalized DPPs' efficacy in reducing the incidence of diabetes and identify any shared features that may assist in optimizing current interventions and improving health outcomes for at-risk individuals.

Methods
This systematic review was conducted using guidelines from the Cochrane Handbook for Systematic Reviews of Interventions 23 and the protocol was registered with the International Prospective Register of Systematic Reviews PROSPERO (#: CRD42022344100).

Search strategy
A systematic literature search was conducted using 5 electronic databases (PubMed, EBSCO, Ovid Medline, EMBASE and Cochrane Library) in October 2022 with a search strategy structured using the PICO tool 24 consisting of related keywords and medical subject headings (MeSH) as seen below:

Data extraction
Due to the small scale of the systematic review (2 reviewers and less than 20 studies), data was stored using standard data extraction templates on Google Forms based on Cochrane guidelines. 23Data was extracted from all studies by 1 reviewer (VN) then a second reviewer (PG) independently repeated extraction on a random 20% of selected studies to confirm accuracy.Disagreements were resolved with discussion.Data about primary outcomes related to reducing diabetes risk were sought for extraction to compare intervention efficacy.
Studies were grouped by these primary outcomes (decreased BMI, weight, glycemic levels, and T2DM incidence) and the mean and standard deviation of baseline and end-of-study results were tabulated with P-values and confidence intervals, where available, to compare statistical significance.Study variables extracted include the following intervention characteristics: author, year of publication, population sample size, study design, country of study, the study's definition of "at-risk" and intervention characteristics, primary outcome unit of measurement and study duration.Narrative synthesis was chosen to summarize these intervention characteristics and outcomes due to clinical heterogeneity (population baseline characteristics and intervention methods) and methodological heterogeneity.

Assessing bias
The quality of selected studies was assessed for risk of bias (RoB) using the Cochrane Collaboration Tool 25 by 1 reviewer (VN).A random 20% of studies were screened by a second reviewer (PG) to confirm reliability.Disagreements were resolved with discussion.Studies were rated either "low risk," "some concerns" or "high risk" over 5 different domains.They were deemed "low risk" if they scored "low risk" on all domains or if the domains of "some concern" were highly unlikely to affect study results and deemed "high risk" if they scored "high risk" in at least 1 domain.Results were presented in Robvis 26 traffic light and weighted bar plots.
Quality of evidence was assessed using the GRADE scale 27 where reviewers' confidence in the similarity of the true effect and the estimated effect were graded by 4 levels of certainty-"high," "moderate," "low," and "very low."

Results
Reporting of the study findings followed the Prisma protocol as shown in the checklist (see Supplemental Table 1).The PRISMA flow diagram 28 shown in Figure 1 details the process of study selection and reasons for exclusion.The initial literature search identified 226 articles and after duplicates were removed 151 publications were screened by title and abstract.Forty studies were sought for retrieval and 1 study was not available on public databases.Thirty-nine studies from the following databases, PubMed (8), Ebsco (3) Ovid Medline ( 14), Cochrane (6), and other sources (8) underwent double screening and were assessed for eligibility after which 9 studies were included in this systematic review.

Risk of bias
Using Cochrane Collaboration Tool's RoB 2, 25 Figure 2 details the result of assessment by domains for the selected studies.32]34 There was no selection bias in the reviewed studies due to the adequate randomization and allocation concealment in the studies.Regarding performance bias, by nature of the interventions, the majority of patients and caregivers could not be blinded but this did not affect study outcomes.Detection bias was present in 2 studies whose outcomes were not statistically significant 29,36 and no reporting bias was identified as all outcomes were reported in the final analysis.

Discussion
This review study identified 5 studies 30,31,[35][36][37] that showed clinically significant reductions of various diabetes risk related outcomes especially glycemic levels, weight, BMI and reduced incidence of T2DM, possibly strengthening available literature on the efficacy of 2 main digital modalities (computer and mobile devices) in diabetes prevention.Despite the statistical significance of these RCTs, the studies were significantly heterogeneous in their outcomes due to the large variation in the intervention methods.The study findings suggest that the efficacy of individual DPPs may be due to the synergistic effect of various modes of delivery (technological and in-person) rather than the superiority of one modality over the others.However, due to the small number of studies about each digital modality, the impact of these variations in the intervention could not be properly assessed.In this study, there were 2 RCTs 31,33 involving computerbased DPPs that yielded statistically significant results, proving the capacity of computers to fully automate DPPs or to act in a supplementary role -providing resources alongside in-person curriculums.For one study, 31 after 6 months, the use of individualized websites and interactive emails improved the glycemic level, body weight and 8-year diabetes risk of people with prediabetes.The use of computer-mediated peer support groups, tracking and behavioral change curriculums yielded significant weight loss at 12 months, 33 whereas mobile-mediated interventions led to positive outcomes such as greater weight loss over time and reduction in the risk of developing diabetes at 12 months. 30,35,36A text-messaging intervention where delivered content was individualised 39 and one where messages were standardized to collective participant preferences, 36 both decreased the cumulative incidence of T2DM.In agreement with previous evidence, 39,40 supporting in-person DPP sessions with a mobile app for tracking progress and mediating educational resources achieved clinically significant weight loss, 30 thus affirming the strength of mobile-mediated interventions in increasing patient adherence through positive reinforcement. 39,40][37] On the other hand, mobile-mediated DPPs are effective in consistently reinforcing patient education, 39,41 self-monitoring and social support 39 which prevents non-adherence.The high global levels of smartphone ownership 40 further decrease economic expenditure and accommodate a widespread dissemination of intervention programs.Four text message-based programs focused on stage-based behavioral change were shown to significantly reduce the progression of prediabetes to diabetes with preventative outcomes of weight loss and diet logging. 29,35,38,42In examining a broad digitalization of DPPs, a 2017 meta-analysis correlates smartphone apps mediating health professional personal contact with increased weight loss, in comparison to complete automated contact. 43vidently, by replacing in-person group exercise sessions with phone-app delivered home regimens, Fukuoka et al's 30 intervention yielded no change in glucose levels, supporting how retaining partial human factors in delivery is still imperative in patient adherence and response.
Regarding the effect of computer-based therapies in decreasing HbA1c in at-risk patients, the evidence is judged to be moderate in quality.The studies have a generally low risk of bias besides some concerns of attrition for participant retention. 31,33There is no great imprecision, indirectness, or publication bias but the heterogeneity of the intervention method and the data of statistical significance only being sourced from 3 studies, lowers confidence in the review's results.For mobilebased therapies and their impact on reducing the incidence of T2DM, the evidence is judged to be of low quality.There is moderate to high risk in 2 out of 3 studies of statistical significance and the sample size in those studies was small (61 and 104 participants). 30,36The RCTs have no obvious imprecision, indirectness, or publication bias but the heterogeneity of intervention and study protocol were variable.
Digital health technology interventions play a pivotal role in augmenting existing strategies for the prevention of type 2 diabetes mellitus (T2DM).These interventions leverage innovative technological solutions to enhance outreach, education, monitoring, and management of individuals at risk for T2DM.By incorporating mobile applications, wearable devices, and online platforms, digital health technologies provide personalized and real-time interventions, promoting lifestyle modifications and behavior change.These tools facilitate continuous glucose monitoring, physical activity tracking, and dietary management, empowering individuals to take an active role in their health.Moreover, digital interventions enable healthcare providers to remotely monitor patients, offer timely if not real-time feedback, and tailor interventions based on individual progress.The integration of data analytics and machine learning further refines predictive models, identifying high-risk populations and customizing intervention strategies.In essence, digital health technology interventions not only complement but also transform conventional approaches to T2DM prevention, fostering a more dynamic, individualized, and proactive healthcare paradigm.

Implications for practice
This study provided evidence for the generalized efficacy of health technologies in preventing diabetes by reducing riskrelated outcomes as 5 studies within this review demonstrated statistically significant reductions in HbA1c and diabetes incidence.Further research is needed to translate this data into clinical settings.The feasibility of digitally mediated DPPs is supported by the gradual modernization of the healthcare system and public accessibility to technology but this can be better understood with more pragmatic trials involving diverse populations or interventions that are culturally adapted (which showed a positive outcome in one study. 29We also provided evidence of appropriateness, as diabetes prevention involves self-management and literature which identifies mobiledevices efficacy in increasing patient engagement 39,40 and the high retention rates of studies within this review 29,30,32,33,[35][36][37] support this.Evidence of meaningfulness was unavailable as patient-rated outcomes were not examined in this review and effectiveness is difficult to determine with the moderate to low quality evidence of the chosen studies.
Establishing a sustainable infrastructure for digital diabetes prevention programs is imperative for its long-term success.This infrastructure should prioritize interoperability with existing health information systems, ensuring seamless data exchange and continuity of care.Robust cybersecurity measures, including secure data storage and transmission, are essential to protect patient information and comply with healthcare data protection regulations.The use of scalable, cloud-based architecture enables flexibility and accommodates a growing user base, while user-friendly interfaces cater to a diverse range of users.Integration with wearable and IoT devices allows for real-time data collection, and analytics tools provide insights for continuous program improvement.Training modules and support mechanisms ensure both healthcare providers and end-users can effectively utilize the program.Regulatory compliance frameworks, reliable telecommunication infrastructure, and continuous monitoring and maintenance further contribute to the infrastructure's resilience.By addressing these components comprehensively, digital diabetes prevention programs can seamlessly integrate into healthcare systems, promoting sustained engagement and positive health outcomes.

Limitations and future research
The following limitations should be considered when interpreting this data.First, there was a small number of studies that met the eligibility criteria in this systematic review and the heterogeneity of study design (population, intervention mode of delivery, method, duration) collectively led to difficulty in comparison and inability to perform a meta-analysis to assess the strength of data.Although we have restricted the studies to increase the robustness of our study, future studies might include other non-RCTs for a better understanding of the impact of DPPs.Second, the study populations in the included studies lacked diversity which could affect the scalability and accessibility of these interventions.The fact that some studies were single-gendered, others recruited from specific workplaces or ethnic groups without cultural adaptation of the DPP program, may have affected the participant engagement.Third, the study durations were short ranging from 5 to 36 months with the longest follow-up at 24 months and both these factors limit the external validity of DPP interventions which target a chronic and global health issue.Also, the reviewed studies were all written in English including the Hong Kong Study.This could potentially exclude crucial findings from studies not written in English.Articles not written in English from non-English speaking jurisdictions should also be reviewed in future studies.
In this study, we identified a research gap with studies that examined specific aspects of health technologies for increased patient engagement.Two studies imply the importance of partial human factors in interventions, attributing high retention rates to healthcare staff delivering the intervention 27 and the participatory in-person activities. 37Therefore, studying the interactive nature of digital and in-person modalities may optimize intervention protocol.Understanding the difference between personalized or standardized interventions and full automation or partial digital mediation is important and can be further explored in future studies.
An examination of diverse digital health approaches across different populations, healthcare systems, and cultural contexts is needed to develop targeted and culturally sensitive digital intervention solutions.Also, research should focus on identifying determinants of successful implementation, such as user engagement, usability, and adherence, to optimize intervention designs for maximum impact.Thirdly, continuous evaluation of digital interventions over the long term is crucial to assess their effectiveness and sustainability, informing potential updates or modifications.Fourthly, exploring the integration of digital health technologies into existing healthcare systems, considering interoperability, data security, and impact on workflows, is essential for successful implementation on a broader scale.Lastly, investigating the cost-effectiveness and scalability of these interventions will be vital for their widespread adoption and integration into routine healthcare practices.

Conclusion
In conclusion, this review supports the current literature's generalized perspective on the efficacy of health technology in diabetes prevention and presents some evidence of positive correlations between digitally translated DPPs and improved diabetes risk-related outcomes.Yet, the heterogeneity in intervention protocol and study procedure and the lack of diverse populations or long-term data and follow-up undermine the possibilities of clinical effectiveness and the feasibility of public health implementation and scalability due to the risk of economic burden and the chance of patient non-adherence.Future research should include studies of larger diverse populations with longer duration and followup data and specific aspects of health technologies that increase patient engagement and the possibility of clinically significant outcomes.

Figure 2 .
Figure 2. Risk of bias assessment for the various domains.

Identification of studies via databases and registers Identification Screening Included Figure
determined intervention 1. Flowchart of study selection.

Table 1 .
Characteristics of the selected articles.

Table 2 .
The primary outcomes of the reviewed randomized controlled trials.