This study received approval from the ethics review committee of the Xiangya School of Nursing, Central South University (batch 2019002). This study was retrospectively registered with the Chinese Clinical Trial Registry (ChiCTR2400090554) on October 24, 2024.

This study utilizes a randomized controlled trial design. The inclusion criteria for the participants were as follows: (1) women as defined by biological sex, aged 18 years and older; (2) having access to a smartphone (Android and iOS phones compatible with Fitbit) and being willing to wear a Fitbit sports tracker for the entire study period; (3) willing to share their behavioral data collected via Fitbit with researchers in this study; (4) WC>80 cm and BMI>24.5; (5) having at least 1 child in the age range of 1-12 years; (6) not reached menopause; (7) able to read Chinese and speak Mandarin; and (8) planning to live locally for at least 8 months. The exclusion criteria were as follows: (1) being pregnant, (2) having given birth within 12 months prior to the enrollment date, (3) having an acute or life-threatening illness (eg, kidney failure), (4) having a condition that requires dietary and activity control (eg, diabetes, hypertension, hyperthyroidism), (5) having irregular periods for 6 months, and (6) having plans to become pregnant within 1 year.

Monte Carlo simulations were conducted, aiming to estimate the power for effect sizes to be tested with multilevel regression models. Using G*power 3.1 software, sample size estimation was conducted based on the effect size of HbA_1c. The effect size of HbA_1c was calculated according to relevant literature as 0.83 [22], using a 2-tailed test with a significance level set at .05. To ensure a power of 90%, the total sample size required was determined to be 64. Accounting for a 20% dropout rate, the total sample size needed for this study was 80 individuals, with 40 participants in each group.

Participants were recruited from Health Management Centers for Adults and Children of 2 tertiary hospitals in Changsha, a provincial capital city of Hunan Province in China via posters and web-based recruitment. Two research assistants put up posters at the health management centers. The web-based recruitment was promoted in the WeChat Moments through Eqxiu. Eqxiu is an all-in-one creative design and marketing platform that uses design tools such as posters, long pages, and videos to produce marketing content for users. Interested mothers could contact the research assistants at the research sites. Research assistants measured the WC, height, and weight onsite to ensure they met inclusion and exclusion criteria.

If interested participants met the eligibility criteria, the researchers would provide them with a description of the purpose, content, process, risks, benefits, and the right to withdraw. Written informed consent was obtained from the participants in this study. All research-related information, including basic demographic data of the research participants and the original data recorded in this study, strictly adhered to the principles of research confidentiality. Individuals participating in this study received a designated compensation for the time and effort they expended in support of this research following the completion of each data collection session.

Randomization was performed using a random sequence generated by SPSS software (v. 25.0; SPSS Inc) in this study. The mHealth group was designated as number 1 beforehand and the control group as number 2. According to the time order of entry into the study, the researchers used SPSS software to generate a corresponding random number for each study participant and then coded half of the random numbers as 1 and the other half as 2. The study participants were entered into 2 groups according to this code. After randomization, the intervention and control groups included 40 participants, respectively. Neither the study participants nor the medical staff involved in recruitment were informed of the randomization allocation results. All research assistants involved in this study possessed extensive clinical and research experience. Prior to the commencement of the program, they underwent specialized training provided by the program lead. Notably, the research assistants responsible for the interventions in the mHealth group and the control group were different, with each receiving separate training to avoid potential contamination.

The mHealth-based lifestyle intervention was developed based on the social cognitive theory. The intervention consisted of 3 components: (1) recording participants’ exercise, sleep, and diet data via Fitbit during the whole study period (6 months); (2) accessing the 12 weekly lifestyle modification modules delivered on WeChat; and (3) receiving personalized biweekly messages after the modules, informed by Fitbit data, which were sent via WeChat for 3 months.

The control group interventions consisted of 3 main components: Fitbit trackers identical to those used by the mHealth group, 12 educational modules addressing general health information delivered on WeChat, and health-improvement messages. The 12 modules, which did not focus on lifestyle modification, covered a variety of topics: (1) obesity and diabetes, (2) parenting education, (3) domestic violence, (4) sourcing food, (5) Chinese Adult Physical Activity Guidelines, (6) cardiovascular disease, (7) health checkups, (8) anxiety, (9) sexually transmitted diseases, (10) contraception, (11) hepatitis B, and (12) AIDS. Figure 2 shows the samples of the weekly educational modules for the control group. The supervision pattern for module completion in the control group mirrored that of the mHealth group. Additionally, nontailored informational tips were delivered to the control group, aligning with the educational content of the modules. For instance, participants received messages such as “AIDS, a class B legal infectious disease in China, lacks effective cure drugs. Click to review the tenth module on WeChat.”

At baseline, 3 months, and 6 months, research assistants who were unaware of group allocation contacted the participants via phone or WeChat for data collection, including questionnaires and physical examinations. The research assistants responsible for contacting participants were distinct from those who implemented the interventions and were blinded to group assignments. Participants completed questionnaires in a designated hospital room, with researchers available for assistance. Physical examinations, conducted by trained nurses, included height (measured at baseline), weight, and WC, with data averaged from 3 measurements. Fasting blood samples for HbA_1c were collected by hospital nurses at baseline and 6 months. Participants received travel and meal allowances as appreciation for their participation. To avoid contamination, data collection for both groups occurred on separate dates at the study sites.

Raw data were entered in pairs using EpiData 3.1 software [31] to verify accuracy. Data were analyzed using SPSS 25.0 software [32] with a test level of α taken as .05. Statistical description methods were proposed to describe the demographic and sociological data and all scale data for both study groups. One-way analysis of variance (for measurement information), least significance difference tests (2-way comparisons between factors), and χ² tests (for counting information) were performed to compare the differences in demographic information and primary and secondary variables between the 2 groups of enrolled and study participants who did not complete the follow-up assessments. Intention-to-treat analysis was used, with all participants (N=80) analyzed based on their original group assignments, including those excluded or lost. To compare outcome indicators (including psychological, behavioral, weight-related, and diabetes risk–related variables) between the 2 groups over 6 months, we established a 3 (time: baseline, 3 months, and 6 months) * 2 (group: intervention and control) generalized estimating equation (GEE) model by using general sociodemographic information and disease risk–related data as covariates to control for the confounding factors. Additionally, we developed a 2 (time: baseline and 6 months) * 2 (group: intervention and control) GEE model to evaluate the time trend of HbA_1c levels and compare the differences in HbA_1c between the mHealth group and the control group.

The GEE model offers key advantages over traditional repeated measures analysis. It accounts for within-subject correlations in longitudinal data, provides reliable estimates even when the correlation structure is incorrectly specified, and focuses on population-average effects. GEE can also handle missing data by assuming it occurs completely at random, allowing for the inclusion of incomplete datasets. Additionally, by incorporating covariates, GEE controls for confounding factors, making it a robust choice for analyzing repeated measurements over time.

A total of 80 participants were enrolled and randomly assigned to the mHealth group (n=40) and the control group (n=40). Figure 3 lists the details of inclusion and loss to follow-up. The mean age of the participants was 34.86 (SD 4.36; range 26-45) years. About 79% (63/80) of the participants had a university or higher level of education. The mean WC and BMI of the sample were 92.61 (SD 9.58) cm and 28.12 (SD 3.93) kg/m², respectively. The mean type 2 diabetes risk score of the sample was 21.50 (SD 9.66). There were no statistically significant differences in baseline sociodemographic and clinical characteristics between the mHealth and control groups. There were no significant differences in the demographic and clinical characteristics between participants who completed the follow-ups and those who missed the follow-ups (P>.05; Table 1).

The total number of WeChat module views for the mHealth and control groups were 697 and 189, respectively, and the top 3 view sessions in the mHealth group were modules 1, 6, and 3 with 89, 80, and 77 views, respectively. All modules can be viewed repeatedly but will also be counted repeatedly. The mean score of the satisfaction questionnaire about weekly educational modules posted on WeChat was 4.37 out of 5 for the mHealth group and 3.71 for the control group.

With a reduction of 3.39 cm and 1.50 points in the mHealth group versus 2.37 cm and 0.56 points in the control group at the 6-month follow-up (Figures 4 and 5; Table 2), respectively, we evaluated the effectiveness of the intervention and accounted for missing data through an intention-to-treat analysis.

The intention-to-treat analysis of the adjusted GEE model revealed a significant group-by-time interaction effect on WC and modifiable type 2 diabetes risk scores (Table 3). We observed a gradual decrease in the primary outcome measures due to the group-by-time interaction effect. No significant effect on BMI was found (Table 3). For secondary outcomes, there was a significant group-by-time interaction effect for the mHealth group on daily steps, self-efficacy, social support for physical activity, and the physiological domain of quality of life (Table 3). Specifically, the daily steps and daily moderate to vigorous activity time in the mHealth group were significantly higher than those in the control group at baseline, 3 months, and 6 months. Although the daily moderate to vigorous activity time in the mHealth group was consistently higher than that in the control group, there was no significant group-by-time interaction effect for this variable (Table 2 and Table 3).

To the best of our knowledge, this is the first study to explore the use of mHealth technology for diabetes prevention in mothers with abdominal obesity in China. We demonstrated that a tailored mHealth-based diabetes prevention intervention is both feasible and resource-efficient, yielding significant positive effects. Specifically, our findings provide preliminary evidence that such interventions can reduce WC, increase daily step count, enhance exercise-related self-efficacy and social support, improve quality of life, and lower modifiable diabetes risk. The high level of compliance among participants suggests potential for implementation in primary health care centers.

The success of this program can be attributed to several key factors. First, the intervention was grounded in the social cognition theory [33], which posits that psychosocial influences on health behavior are shaped by 3 factors: goals, outcome expectancies, and self-efficacy. In this study, participants set realistic, personalized goals for diet and exercise, greatly enhancing the likelihood of achieving desired outcomes. In addition, Fitbit’s data visualization feature encouraged self-motivation and boosted participants’ self-efficacy [34]. Second, our study population consisted of mothers whose children were younger than 12 years, a stage when increased maternal involvement is often required in various aspects of life, including academics, physical health, and emotional well-being. At the same time, they were at a life stage marked by active professional pursuits. As a result, these mothers may have inadvertently neglected their own health due to the competing demands of family and work. However, the mHealth-based lifestyle management intervention overcame time and space constraints, enhancing adherence, particularly for busy mothers [35,36].

The mHealth-based, tailored diabetes prevention intervention effectively reduced WC in mothers with abdominal obesity, which was consistent with findings from a recent meta-analysis [37]. The intervention included components aimed at reducing body fat and WC, such as 12 web-based modules (4 on exercise and 3 on diet). The 15-minute exercise videos were designed for both individual use and joint participation with children. These short sessions made it easier for busy mothers to stay active daily. Additionally, wearing exercise trackers and receiving personalized health text messages further supported WC reduction. This approach highlights that small, consistent daily efforts play a key role in achieving long-term health improvement.

In this study, while the overall diabetes risk score did not show a significant reduction, it is noteworthy that a significant difference in diabetes risk scores emerged between the 2 groups when the 10 nonmodifiable risk factors from the CHINARISK diabetes risk scale were excluded. This left only 4 modifiable risk factors: WC, BMI, physical activity, and dietary habits. These findings suggest that tailored diabetes risk management interventions using mHealth technology can effectively target controllable risk factors such as WC, BMI, diet, and physical activity, ultimately helping to lower an individual’s future risk of developing diabetes.

Exercise self-efficacy, as an essential component of health self-efficacy, reflects an individual’s belief in their ability to resist temptations and maintain a healthy exercise routine [38]. This intervention led to significant improvements in exercise-related psychological variables, including exercise self-efficacy and exercise social support, which are likely to encourage women to engage in more physical activity. This aligns with findings from a systematic review of lifestyle interventions [39]. Our 12 web-based educational modules incorporated family-based exercise training, offering participants new formats that mobilize social resources and enhance support for exercise. However, there is limited national and international research on exercise-related social support for women with children at risk of diabetes. Therefore, the findings regarding social support in this study should be interpreted with caution.

This study shows that tailored mHealth-based diabetes risk management improved the quality of life in the physiological domain for abdominally obese mothers. Participants reported greater satisfaction with their sleep, increased energy levels for daily activities, and enhanced ability to perform everyday tasks. These findings align with weight management studies that link weight reduction to improved sleep quality [40]. Our intervention included exercise training, which encouraged behaviors such as increasing daily step counts. According to the pendulum effect [41], adequate physical activity during the day promotes deeper sleep at night, reduces sleep latency, and enhances overall sleep quality and satisfaction. This, in turn, leads to improved energy levels for daily activities, ultimately enhancing participants’ quality of life in the physiological domain.

The frequent module reviews and the high satisfaction ratings reflect the feasibility of our research. The traditional face-to-face diabetes risk management program is costly and time-consuming, posing challenges in low-income countries with a shortage of primary health care professionals and limited medical resources [16]. WeChat, a popular multifunctional mobile app, is China’s leading instant messaging tool and offers a wide range of connectivity features [42]. Utilizing WeChat as a web-based platform for health education has proven effective in promoting weight loss among Chinese obese women and even preventing breast cancer [43]. Our diabetes risk management program on WeChat leverages its flexible learning formats to engage participants and ensure high adherence to the intervention. Additionally, it requires less reliance on health care professionals, which helps conserve both human and economic resources—a significant advantage in the current context of low-income countries.

First, participants were recruited exclusively from Changsha, limiting the generalizability of the findings to mothers with abdominal obesity across China. Second, the study lasted only 6 months, which may be insufficient to fully understand the long-term development of diabetes, given its gradual onset. Third, there could be discrepancies between the participants’ actual eating behaviors and the diet data they manually entered in the Fitbit app. Fourth, the intervention and data collection mainly took place in autumn, a time when people tend to eat more [44], potentially contributing to the lack of significant changes in dietary variables in this study.

Future trials should place importance on long-term follow-ups to confirm the sustained effectiveness of this risk management intervention. Using factorial design methodology can help identify which specific elements of the intervention directly influence changes in patient behavior. This study, delivered through the WeChat platform, uses flexible learning formats to actively engage participants and ensure high adherence to the intervention. Moreover, by reducing reliance on health care professionals, this approach provides cost-effective benefits, which are particularly valuable in resource-limited health care settings.

This mHealth-based and tailored diabetes prevention model lays both the theoretical groundwork and practical strategies crucial for scaling up diabetes prevention and risk management initiatives in public health nursing practice. Although our study sheds light on the initial effects of mHealth-based interventions, it addresses the current scarcity of medical resources in China, offering a potential blueprint for tackling various chronic conditions.