The Use of Mobile Health Interventions for Outcomes among Middle-Aged and Elderly Patients with Prediabetes: A Systematic Review

Background: There are currently limited systematic reviews of mobile health interventions for middle-aged and elderly patients with prediabetes from trial studies. This review aimed to gather and analyze information from experimental studies investigating the efficacy of mobile health usability for outcomes among middle-aged and elderly patients with prediabetes. Methods: We conducted a literature search in five databases: Clinicaltrials.gov, the International Clinical Trials Registry Platform (ICTRP), PubMed, ProQuest, and EBSCO, with a date range of January 2007 to July 2022 written in English, following a registered protocol on PROSPERO (CRD42022354351). The quality and possibility of bias were assessed using the Jadad score. The data extraction and analysis were conducted in a methodical manner. Results: A total of 25 studies were included in the qualitative synthesis, with 19 studies using randomized trial designs and 6 studies with non-randomized designs. The study outcomes were the incidence of diabetes mellitus, anthropometric measures, laboratory examinations, measures of physical activity, and dietary behavior. During long-term follow-up, there was no significant difference between mobile health interventions and controls in reducing the incidence of type 2 diabetes. The findings of the studies for weight change, ≥3% and ≥5% weight loss, body mass index, and waist circumference changes were inconsistent. The efficacy of mobile health as an intervention for physical activity and dietary changes was lacking in conclusion. Most studies found that mobile health lacks sufficient evidence to change hbA1c. According to most of these studies, there was no significant difference in blood lipid level reduction. Conclusions: The use of mobile health was not sufficiently proven to be effective for middle-aged and elderly patients with prediabetes.


Introduction
Impaired Fasting Glucose (IFG), Impaired Glucose Tolerance (IGT), or both, and/or elevated levels of Hemoglobin A1c (HbA1c) are all signs of prediabetes [1]. In 2015, there were over 318 million cases of IGT worldwide, and by 2040, there are expected to be about 481 million cases [2]. Prediabetes has become more common as people become older [3]. The transition from normoglycemia to prediabetes or diabetes, or from prediabetes to diabetes, was prevalent in middle-aged and elderly people, suggesting the need for continuing treatment [4]. Individuals with prediabetes are four times more likely than individuals with normal glucose tolerance to develop Type-2 Diabetes Mellitus (T2DM) [5]. Prediabetes can lead to major complications, whether or not patients have T2DM. Microalbuminuria

Study Selection
Three reviewers (Y.A.J., R.N. and D.D.E.) worked independently to identify articles by looking at the titles, then studying the abstracts, and finally reading the full-text form. The full articles of any possibly relevant studies were then obtained for a thorough review by all reviewers. In the event of a disagreement, a consensus decision was achieved.

Data Extraction
Three reviewers (Y.A.J., R.N. and D.D.E.) retrieved data from the included studies and entered it into a spreadsheet independently. Author, year, country, sample size, study design, details of intervention and control, outcomes of interest, and the other key results were among the data retrieved from the studies. If the needed data were not reported in an article, we contacted the article's corresponding author to obtain any missing information. If within four weeks of being contacted, the corresponding author does not respond, then the study was not included.

Quality Assessment
Using an adapted version of the Jadad score, two reviewers (R.N. and D.D.E.) independently assessed the quality and risk of bias of the included studies [33]. In the event of a disagreement, another author (Y.A.J.) joined the debate to assist in the resolution of the issues.

Data Synthesis
The Cochrane methodologies for data analysis and syntheses [34] were used to accomplish the data synthesis (narrative synthesis). Data from both qualitative and quantitative sources were described, analyzed, and classified individually. The characteristics of the research, as well as the significant disparities between them, were meticulously recorded. Topics that were relevant were highlighted, and the data were translated into a descriptive style. As stated in the original studies, the outcome results are described in detail. Due to a paucity of studies with similar settings, interventions and results, a meta-analysis could not be performed.

Search Results
For our investigation, we selected five electronic search engines (PubMed, ProQuest, EBSCO, Clinicaltrials.gov, and the International Clinical Trials Registry Platform (ICTRP)) accessed on 8 July 2022. There were 11,938 electronic database documents discovered in all. A total of 11,871 articles were removed from the study due to being irrelevant (title, conference, published, editorial, news, comments, and reviews). In addition, 11 articles were eliminated from the study because they had a title that was duplicated across four search engines. Following that, 16 papers were eliminated since the abstracts did not meet the inclusion requirements. A total of 41 full articles were reviewed for eligibility and 25 were included in the qualitative synthesis ( Figure 1). descriptive style. As stated in the original studies, the outcome results are described detail. Due to a paucity of studies with similar settings, interventions and results, a me analysis could not be performed.

Search Results
For our investigation, we selected five electronic search engines (PubMed, ProQue EBSCO, Clinicaltrials.gov, and the International Clinical Trials Registry Platform (ICTRP accessed on 8 July 2022. There were 11,938 electronic database documents discovered all. A total of 11,871 articles were removed from the study due to being irrelevant (tit conference, published, editorial, news, comments, and reviews). In addition, 11 artic were eliminated from the study because they had a title that was duplicated across fo search engines. Following that, 16 papers were eliminated since the abstracts did not me the inclusion requirements. A total of 41 full articles were reviewed for eligibility and were included in the qualitative synthesis ( Figure 1).
The youngest average age of 37.8 ± 9.2 years was found in Muralidharan's [54] study, while the oldest average age was found in McLeod's [53] study, which is 61.8 ± 9.5 years old. Hamdan's [41] study only examined female subjects and Ramachandran's [36] only examined male subjects. Bender's study only intervened in Filipino Americans because they had a higher risk of developing type 2 diabetes [49]. The shortest studies lasted 3 months, and were found in the Sevilla, Kondo, and Griauzde [46,50,60] studies. Khunti's study had the longest intervention which was 48 months. There were three types of interventions used, namely text message-based, website-based, and mobile phone apps.
Staite [39] used all three intervention combinations. Block [37] used websites and mobile phone apps. Stewart [59] used text message-based and mobile phone apps. There were 15 studies using only mobile phone apps, 3 other studies using websites, and 4 studies using text messages only. There were 6 studies that use wearable devices as part of the intervention [39,43,49,50,59]. The study outcomes were classified into five categories, which included the incidence of diabetes mellitus (DM), anthropometric measures, laboratory examinations, measures of physical activity, and eating behavior. Ramachandran [36], Khunti [42], and Nanditha [52] examined all of the five outcome categories. Kondo and Chen examined four of the five categories except incidence of DM. Meanwhile Fischer [57] (anthropometric measures) and Francis [40] (physical activity measures) only assessed one of the five clinical outcome parameters.

Quality Assessment
The Jadad score [33,62] was used to measure the quality of the trials, which reflects the quality of research based on their description of randomization, blinding, and dropouts (withdrawals). The methodology features of the Jadad score are shown in Table 3. Each study in this systematic review received a score (Table 4). In the Jadad score, the scale runs from 0 to 5, with a score of ≤2 indicating a low-quality report and a score of ≥3 indicating a high-quality report [33]. Six [55][56][57][58][59][60] studies have low quality, according to the analysis. This is due to the trial's insufficiency of randomization in selecting participants. Concerning "blinding," we understand that many studies do not receive points because the nature of the intervention does not allow for "blinding".     When the participant was seated, the arterial blood pressure was taken from the right arm.
Venous sampling was used to assess standard biomedical outcomes such as HbA1c, a lipid profile (triglycerides, HDL, LDL, and total cholesterol), urea and electrolytes (sodium, potassium, urea, and creatinine), and liver function tests (albumin, total bilirubin, alkaline phosphatase, and alanine transaminase). Venous blood samples were collected after 8-12 h of overnight fasting and processed at CAP-accredited laboratories (National University Hospital Department of Laboratory Medicine or National Healthcare Group Diagnostics). Plasma glucose was determined using a photometric assay using the hexokinase method, and HbA1c was measured using high-performance liquid chromatography. An enzymatic colorimetric assay was used to determine serum lipids and creatinine levels. Self-reported questionnaires were used to collect participants' physical activity levels in minutes per week at baseline, 3 months, and 6 months.
At the baseline, 3-, and 6-month visits, dietary intake was collected using a 2-day food diary and analyzed using the nBuddy Dashboard's nutrient analysis platform, which includes over 14,000 food items and incorporates the Singapore energy and nutrient composition of food, Malaysian Food Composition, and USDA food databases, as well as nutritional information from food packaging and nutrient analysis of recipes. A chemiluminescence assay was used to measure insulin levels (Access 2, Beckman Coulter).
HbA1c was measured using a Variant II Turbo system (BIORAD) and a 4-mL peripheral blood sample was drawn via venipuncture using the standardized technique.
Diet and physical activity questionnaires were administered. Waist circumference d.
Glucose metabolism e.
Lipid metabolism f.
Liver function Notes: only the 3-month laboratory measurement is evaluated in this paper. The 6-month attrition rate was higher dropout were associated with the COVID-19 pandemic.
The study procedure includes four visits: at baseline, three months later, six months later, and twelve months later, with anthropometric and laboratory examinations performed at each visit. Blood samples were collected after a 12-to 14-h fast. A glucose hexokinase method was used to measure glucose. An enzymatic colorimetric test was used to examine lipids. A two-step sandwich enzyme immunoassay using monoclonal antibodies was used to measure serum insulin. Matthews' formula was used to calculate HOMA-IR. Body composition was determined using bioelectrical impedance analysis with the InBody 370 and 15 impedance measurements at 5 body segments, as well as a tetrapolar 8-point tactile electrode system.      : Randomized Trial. *: Diagnosed by American Diabetes Association (ADA) screening tool (score of ≥ 5 or). **: Age (years) by mean ± SD or range age of participants. IG: Intervention Group; CG: Control Group.

withdrawals and dropouts
Was there a description of withdrawals and dropouts?
Yes 1 No 0  Red: minimum randomization criteria score. Yellow indicates the lowest possible score for the blinding criteria. Green indicates the highest possible score for all criteria. Pink: 1 point off the total score.

Mobile Health Interventions
Six studies [36,39,48,52,57,59] used text messaging or short message service (SMS) as an intervention. Furthermore, Staite [39] and Stewart [59] include interventions other than text messaging, such as mobile phone apps and website-based interventions. Ramachandran [36] and Nanditha [52] reported on text messages based on the trans-theoretical model stage. Despite this, Staite [39] reported text messages that were based on the theory of planned behavior; Stewart [59] reported text messages that were based on Bandura's self-efficacy: toward a unifying theory of behavioral change; and Fischer [48,57] did not report any details about the theory that underpins text messages to participants. Four studies [36,39,52,59] that used messages as interventions reported that content-text messages were created using curriculum material from the National Diabetes Prevention Program (DPP).
In a different manner, five studies [37,39,55,56,60] used website-based interventions. Besides website-based interventions, participants also received intervention via mobile phone apps [37]. Chen Study [55] with SOPs underlying SWAP-DM2 and Block Study [37] with DPP Curriculum demonstrate a distinct variation in the content of each website-based intervention in each study. The website-based intervention adopted the methods and theoretical frameworks, such as The SWAP-DM2 provided diabetes prevention services, ranging from uncomplicated educational websites and record-keeping to quite complex risk-scoring and individualized counseling [55]. Despite this, the Alive-PD was a fully automated and flexible online behavior change strategy [37]. The system includes tools for weight monitoring, eating, and physical activity, as well as weekly health information on diabetes and prevention strategies, quizzes, social support through virtual teams and a participant messaging program, feedback on diet and activity reports, and also goal achievement of success or failure, weekly reminders, and other features [37]. Practice the use of website-based was conducted as a basis for promoting positive Physical Activity (PA), dietary lifestyles [56] and mental resilience [39].

Outcomes Reported
Anthropometric measures, such as weight loss, changes in BMI, and waist circumference were used as the primary outcomes in twelve studies [38,39,42,45,47,[52][53][54][55][56][57]59]. Changes in HbA1c were the main result in five studies [37,44,49,51,58]. In addition, three studies [46,48,50] examined physical activity and its changes as the primary outcome of the research. In Ramachandran [36] and Nanditha's [43] studies, the primary outcome was the incidence of type 2 diabetes. In the study conducted by Griauzde [41], Bender [40], and Sevilla [60], the primary outcomes were the feasibility and acceptability of the mobile health intervention. In Xu's [46] studies, the primary outcome was a change in dietary behaviors and physical activity. The summary of intervention outcomes is shown in Table 5.

Incidence of T2DM
The findings from Ramachandran [36] showed that mobile phone messaging (SMS) could be an effective technique for lifestyle modification to reduce the incidence of type 2 diabetes. The cumulative incidence of T2DM at 24-month follow-up was lower in those who received mobile phone messages than in controls [36]. Despite this, Khunti [50] and Nanditha [43] found that SMS and mobile phone apps did not significantly reduce the cumulative incidence of T2DM at the 12, 24, and 48-month follow-ups. There were other factors that could influence these inconsistent results, such as the content of the intervention and the role of the tool or other intervention (wearable device) [50] that has the potential to have an effect. Differences in examination methods and the provision of a prediabetes diagnosis were also factors that needed to be considered in these findings.

Anthropometric Measures
An analysis of twenty studies using weight change as an outcome found that eight studies [39,43,44,47,49,50,53,60] showed no significant differences between intervention and control. Despite this, a significant difference was observed between the intervention and control groups in nine other studies [37,38,42,45,51,52,54,57,59]. Furthermore, the Sharit [56], Chen [55], and Summers [58] studies found significant changes in body weight following the intervention. Seven [37,38,42,45,51,57,59] of the 12 studies that found a significant difference in weight change from mobile health apps were based on content based on DPP.
There were two studies [39,57] that include a ≥3% weight loss as a research outcome. The results of the two are completely contradictory. According to the Fischer study in 2016, there was a significant difference in achieving ≥3% weight loss between the intervention and control groups [39], Meanwhile Fischer in 2019 [57] stated the opposite. Respondent characteristics, such as ethnicity, were suspected to be important considerations in these findings. There was a significant difference between the control and intervention groups and pre-post intervention in three [40,51,52] of the four studies that included achieving ≥5% weight loss as an outcome. Fischer [39], however, stated that there was no difference between the intervention and control groups. Two studies [40,51] used DPP-based content and mobile health apps as interventions.
Waist circumference was observed as an outcome in fifteen studies. Six studies [36,43,47,49,50,53] showed no significant effect of intervention on waist circumference. Despite this, eight studies [37,38,40,45,46,51,54,60] reported a significantly reduced waist circumference between the intervention group and the control. In addition, Sharit's study described a different mean reduction in waist circumference between baseline and postintervention [56]. Table 5. Summary of Intervention Outcome.

Type of Trial Non-Randomized Interventional Study
Randomized Trial   : Results were statistically significant. : Results were not statistically significant.

Randomized Trial
: Not observed in the study.   : Results were statistically significant. : Results were not statistically significant.

Randomized Trial
: Results were statistically significant.   : Results were statistically significant. : Results were not statistically significant.

Results of Intervention Outcome
: Results were not statistically significant.

Physical Activity
Eleven [36,38,43,48,[50][51][52][54][55][56]59] studies reported changes in physical activity as an outcome measure of the study. Five of the 11 studies stated that the mobile health intervention showed significant differences in increased physical activity between the control and intervention groups [48,51,54,59] and pre-post intervention [55]. Despite this, six studies reported that there were no significant differences in increases in physical activity between the control and intervention groups [36,38,43,50,52] and pre-post intervention [56].

Dietary Behaviors
Eight [36,43,46,49,50,54,55,60] studies reported dietary behavior as an outcome measure. Five [36,43,46,54,55] of the eight studies stated that the mobile health intervention showed significant differences in changes to healthy dietary behavior between the control group and the intervention and pre-post-intervention groups [55]. Despite this, three [49,50,60] studies reported that there was no significant difference in changes in dietary behavior between the control and intervention groups.
Reduction in blood glucose levels was one of the measures used as an outcome in eight studies [36][37][38]40,43,53,54,60]. Three [36,37,60] studies explained that there were significant differences in reduction in blood fasting glucose levels between the intervention and control groups and 1 h glucose levels in pre-post intervention. Whereas no significant difference in reduction in blood fasting glucose levels was found between the intervention and control groups in the other five studies [38,40,43,53,54]. In addition, the reduction in blood glucose levels was also significantly different in the oral glucose tolerance test (OGTT) in the control and intervention groups in the study on Vida Sana apps [60].
Seven studies [36][37][38]43,45,49,52] reported blood lipid level reduction as an outcome of the study. All of these studies stated that the mobile health intervention did not show a significant difference in blood lipid level reduction between the control and intervention groups [36][37][38]43,45,49,52].

Discussions
Traditional face-to-face treatment for achieving and sustaining weight or BMI can be replaced with technologies that are practical, affordable, and scalable [64]. Mobile phone intervention has gained widespread acceptance among people of all ages and socioeconomic backgrounds, and it provides numerous opportunities in health care, including self-management and T2DM prevention [65]. Despite growing concerns about data privacy in the fields of access, patient confidentiality, and data storage [66]. The variety of definitions and use of diagnostic tools used in these study results indicate that prediabetes was a complex condition that triggers a burden on clinical services and public health policies [3]. The use of combined diagnostic tools is expected to be able to detect conditions of undiagnosed prediabetes and diabetes [67]. Depending on the diagnostic tools used as standards, the characteristics of the intervention each subject requires, and the treatment's results can change.
During long-term follow-up, there was no significant difference between mobile health interventions and controls in reducing the incidence of T2DM [43,50]. Despite this, there was variation in outcomes that may be brought about by variances in the intervention's underlying theory and the message's content to participants, and length of observation. Participants in twenty trials used technology to reduce weight, according to this systematic review. Studies show that mobile phone messaging is an effective and acceptable method to deliver advice and support towards dietary behavior to prevent T2DM in individuals at high risk. The findings [57] showed that although SMS4PreDM was relatively low-cost to deliver and demonstrated high retention, weight loss outcomes may not be sufficient to serve as a population health strategy. It can happen because pre-diabetes is influenced by many factors [55].
In these studies, website-based interventions were used to promote positive health outcomes such as physical activity, dietary habits, and mental resilience. These interventions differed in their complexity, ranging from simple educational websites to more complex individualized counseling. The differences in the details of the form of intervention, may be one factor making the research results inconclusive.
As a simple and effective measure of central obesity, waist circumference is a major predictor of increased risk of hypertension, diabetes mellitus, dyslipidemia, metabolic syndrome, and coronary heart disease [68]. Waist circumference could potentially be used as a clinical equivalent for visceral adipose tissue, which impairs insulin sensitivity and predisposes to prediabetes when excessive [69]. A systematic review and meta-analysis study looked at the impact of technology on waist circumference reduction and found a mean change of 02.99 cm (95% CI: 03.68 to 02.30) [68]. Self-management by chronic healthcare customers can be enhanced via mobile apps, according to users of health apps. This might be used for patients with prediabetes, particularly in terms of lifestyle adjustments [70].
Using text messages instead of other reminders had some advantages. Text messages could be sent to patients at the same time, they were always available, cost-effective, and required less staff [71]. Despite all the benefits and features of text messaging on a mobile phone, there are some drawbacks, such as: the staff could not be certain that text messages were received by all participants, mobile phone numbers could have changed [71], and participants sometimes blocked their status or stopped reading the text messages [72]. A few intervention strategies were bidirectional, allowing T2DM patients to receive continuous and personalized support via SMS while also allowing T2DM patients to interact with healthcare professionals and diabetes care educators to strengthen their T2DM selfmanagement abilities and knowledge [65].
In other circumstances, mobile health technology (including social media) may provide benefits such as low or no cost, high scalability, self-tracking, and tailored feedback, image and video use for improved health literacy, wide range, and data sharing for large-scale analytics [73]. Mobile phone apps are widely utilized all around the world. Health and medical apps are increasingly being employed in a range of scenarios, according to evidence. Many authors in the medical and public health literature have highlighted the benefits that laypeople and healthcare professionals may contribute to health and medical apps [74]. Chronic disease or patients' conditions must be monitored for a long time. Using mobile phone apps for chronic self-care could be beneficial, allowing patients to monitor and regulate their illnesses [70]. Despite the fact that mobile health intervention prompts were well-received by patients, there are conflicting results regarding their influence on behavioral, laboratory, and T2DM incidence [75].
There are some limitations in this study that need to be addressed. First, in our study, we only used American Diabetes Association (ADA) criteria for prediabetes, although many other prediabetes studies used the World Health Organization (WHO) criteria. Second, although few studies on mobile health interventions with a sample adequate for prediabetes intervention used randomized controlled trials (RCTs) as their study design, we propose that further studies on mobile health apps for prediabetes intervention use RCT design. Third, we were unable to depict all studies in meta-analysis due to differences in diagnostic methods, evaluation, and reporting of intervention outcomes. Approximately half of the studies provided sufficient data to compute them. Fourth, our search strategy may have resulted in a bias for positive results because null or negative results are less likely to be published. Finally, the full text of certain studies was not available.

Conclusions
The systematic review includes twenty-five studies that were discussed about mobile health intervention for pre-diabetes among middle-aged and elderly patients. A few studies showed that each mobile health intervention promised an effective and acceptable method to deliver advice and support towards lifestyle modification to prevent diabetes. Although, evidence of the effectiveness of mobile health as an intervention for prediabetes was inconclusive.