This study was conducted on an occupation-based population from August 2015 to February 2016 in the Shunyi District of Beijing to explore the effectiveness of the mobile app WeChat on weight loss behavior. Participants who were willing to use our official WeChat account were enrolled in a WeChat group and received 6 months of interventions for weight loss, and those who were not willing to use the account were in the control group. Before the interventions began, the Shunyi District Government released an official weight loss document to all 134 government agencies and enterprises in Shunyi District. At least 60% of the staff of government agencies were required to participate in this activity. The government agencies included the Shunyi Branch of the Beijing Municipal Public Security Bureau, the Shunyi Court, the Shunyi People’s Procuratorate, the Shunyi Branch of the Beijing Administration for Industry and Commerce in the Urban Management Law Enforcement Bureau, the Shunyi State Administration of the Taxation of China, the Beijing Shunyi Local Taxation Bureau, and the Food and Drug Administration of Beijing Shunyi District. At least 30% of the staff of other units were required to participate as well.

Interventions were given to the WeChat group through the WeChat app and routine publicity, such as the slogan “take the stairs and lose weight,” was given to the control group. Participants were asked to report their demographic characteristics such as gender, age, educational level, and telephone number online when they registered with our official WeChat account. Additionally, on average, two weight managers per agency were trained to obtain participants’ data on height, weight, and waist circumference before and after the interventions were initiated for both groups. The demographic characteristics of participants in the control group were collected by weight managers.

Participants were enrolled if they (1) were from one of any of the 134 government agencies and enterprises in Shunyi District, (2) were 18 years of age or older, and (3) wanted to lose weight. Participants willing to lose weight via WeChat were enrolled as members of the WeChat group and those who were willing to lose weight but did not want to use WeChat were in the control group. Pregnant women and those whose health conditions were not suitable for weight loss were excluded from the study.

All participants gave verbal consent to participate in the study and they were entitled to withdraw from the study whenever they wished and for whatever reason. The official weight loss document also addressed all the individuals who enrolled or who dropped out of our study voluntarily. People in the WeChat group would scan a QR code, below which was the phrase “enrolled or dropped out voluntarily” and for people in the control group, before they enrolled, weight managers in each agency told them everything about our study and asked them if they were willing to enroll. From the perspective of each participant’s private information, an agreement was made between the Shunyi Center for Disease Prevention and Control (SYCDC) and a technology company that provided technological support to protect all private information. Only two people from the technology company and two people from the SYCDC knew all the information. Private information such as a telephone number or identification number was deleted when the data were analyzed.

The SYCDC developed an official WeChat account, known as the “Health Education in Shunyi District, Beijing,” with the technological support of a specialized information technology company (Figure 1). Before becoming members of the official WeChat account, users were to register and provide information on their age, gender, educational level, and telephone number. Then, they were provided and could log in with a WeChat identification and password. All participants were asked to follow the instructions set forth in the official account after registration. Participants could read new messages and review the message history of all content published in the official account.

The official WeChat account consisted of the following six components: introduction, weight loss process, weight loss unit rankings, weight loss school, activity area, and awards (Figure 1). Each component is described in detail subsequently.

The weight loss process was designed to provide feedback on weight, diet, and exercise during the intervention process to motivate participants to lose weight. The duration of 12 types of activities, including walking, running, cycling, playing, skipping, swimming, sit-ups, push-ups, plank flat support, doing housework, workshops, and climbing stairs, were reported daily (Figure 1). According to a diet pagoda that was based on the Chinese Dietary Guidelines set forth by the Chinese Nutrition Society, each type of food was given a reasonable intake range; therefore, participants could know whether the amount of food they ate was excessive, moderate, or inadequate (Figure 1). Weight data were reported each week. When an individual reported data on diet, exercise, or weight, his or her physical condition would be assessed and the results would be provided immediately.

The weight loss unit rankings component aimed to rank the total score per unit, including the number of participants and the total amount of weight lost, to motivate and encourage participants to compete with one another (Figure 1). The weight loss school component consisted of a variety of materials on weight loss including micro videos and popular science knowledge (Figure 1) in addition to five experts (two from the Peking University Third Hospital, one from Peking University People’s Hospital, one from Peking University Health Science Center, and one from the Beijing Center for Disease Prevention and Control) who formed a consulting group to address questions from participants (Figure 1). The activity area component included a microcommunity (Figure 1) and other types of activities, where participants could communicate with one another. Finally, there was a rewards component. Participants received scores during the intervention and, to win a competition, participants had to lose weight and actively participate in various activities.

At the end of the program, winners and losers were determined on the basis of their cumulative scores (Figure 1). The scores could be from interactions, feedback information, reading articles, and so on. Different types of activities resulted in different scores. For example, if an individual registered with our official WeChat account, he or she would earn 10 points; reading articles earned two points per day and feedback information on weight, exercise, and diet were worth five points each. The top 60 WeChat active participants per month and the top 50 at the end of the project could win a prize.

A total of 210 messages were sent (one message every other day, on average) to participants in the intervention group during the period. The messages were read more than 247,000 times and were sent to other WeChat accounts more than 6500 times. In total, 3620 participants communicated with others in “microcommunity discussions” and there were more than 20,000 posts.

Participants in the intervention group could ask weight loss-related questions on our official WeChat account expert consultation page at any time during the 6-month period. Experts were consulted online more than 14,576 times and 28,000 questions were asked, among which 270 representative questions were responded to in detail. In addition, more than 8200 people regularly received feedback on their weight, diet, and exercise data and conducted self-assessments via WeChat. Examples of the content of WeChat messages can be found in Table 1.

Data was analyzed using SPSS version 18.0. Mean and standard deviation were used to present continuous variables for normal distribution or median and quartile for nonnormal distribution. Frequency and percentages were employed to express categorical variables. The chi-square test for categorical variables was used to compare parameters between the control group and the WeChat group, or t test for continuous variables with normal distribution. To control for confounding factors and to explore the effects of WeChat on weight loss, propensity score methods with a multinominal logistic regression was utilized; the propensity score replaced all single covariates to adjust the effectiveness on weight loss. A P value less than .05 was considered statistically significant.

At the beginning, a total of 15,818 participants were enrolled, including 12,296 participants in the WeChat group and 3522 in the control group. Data were collected on a total of 15,523 participants, including 12,002 in the WeChat group and 3521 in the control group at the baseline. After 6 months of interventions, data were collected on a total of 15,310 participants, including 11,843 in the WeChat group and 3467 in the control group at the baseline and after 6 months of interventions (Figure 2).

The data that were collected both at baseline and after 6 months of interventions were used. That is, a total of 15,310 participants were enrolled in this study. The mean age of the control group was 39.0 (SD 9.5) years and that of the WeChat group was 35.1 (SD 8.5) years. Participants in the WeChat group were younger than those in the control group, and the difference was significant. Participants in the WeChat group were largely females (66.53%, 7879/11,843) and largely had university/college degrees or above (91.85%, 10,878/11,843). In the control group, nearly half the participants were female (40.47%, 1403/3467) and more participants had university/college degrees or above. Baseline demographics including age, gender, and educational level between the two groups were not balanced (Table 2).

WeChat activeness was represented by WeChat cumulative scores, which were related to actual official WeChat account use. Higher scores indicated that an individual was more active in using WeChat to lose weight. Scores less than or equal to 50 were regarded as inactive, and scores of 50 or more were active. A significant number of participants in the WeChat group were inactive (83.18%, 9852/11,843) and only 16.8% (1991/11,843) were active. The differences between the control group and the WeChat group on demographics were all statistically significant (Table 2).

The weight and waist circumference in the control group decreased by mean 1.78 (SD 2.96) kg and mean 2.39 (SD 3.91) cm, respectively, whereas in the WeChat group, weight and waist circumference decreased by mean 2.09 (SD 3.43) kg and mean 2.74 (SD 4.48) cm. A stratified analysis was performed to show the mean weight loss of the two groups. It also showed that for males, a decrease in weight loss was statistically significant, which indicated the 6-month WeChat interventions were effective for weight loss; however, for females, weight loss changes were not statistically significant (Table 3).

Propensity score methods are increasingly used to control for confounding factors in many medical studies. In our study, demographics including age, gender, and educational level between the two groups were not balanced at the baseline; thus, propensity score methods were used to control for them. With a group variable (the WeChat group and the control group) as the dependent variable and two demographic characteristics (age and educational level) as covariates, a binomial logistic regression analysis was used to estimate the propensity score based on gender.

To estimate WeChat effectiveness on weight loss, multinominal logistic regression used to test the parallel lines of the ordinal logistic regression that did not meet the criteria. The weight loss outcome was categorized as follows: weight gain, weight unchanged, weight loss from 0 to1 kilograms, weight loss from 1 to 2 kilograms, and weight loss more than 2 kilograms. The reference category for the dependent variable was the classification of weight gain. The propensity score was used as a continuous covariate replacing all single covariates with the independent variable, WeChat active group, in the multinominal logistic regression to estimate the adjusted WeChat effectiveness on weight loss.

For males, in the control group, 52.37% (1081/2064) of participants lost 0 to 1 kg, 35.72% (1220/3415) in the inactive group, and 41.17% (226/549) in the active group lost more than 2 kg. For females, 38.92% (546/1403) of participants in the control group and 31.30% (2015/6437) in the inactive group lost 0 to 1 kg and 33.36% (481/1442) in the active group lost more than 2 kg (Table 4).

For males, the results of the multinominal logistic regression showed that when controlling for confounding factors, compared with the classification of weight gain, the WeChat group (with both active and inactive subgroups) had higher probability of maintaining weight, weight loss from 1 to 2 kg, or weight loss of more than 2 kg than the control group. However, the control group had higher probability of weight loss from 0 to 1 kg. For females, the difference between maintaining weight and weight gain was not statistically significant. As for weight loss of more than 1 kg, the WeChat inactive group had lower probability than the control group. As for weight loss from 0 to 1 kg, the WeChat group (with both active and inactive subgroups) had lower probability than the control group (Table 5).

This study had some limitations. First, this study was a nonrandomized trial and although propensity score methods were used to control for confounding factors, the conclusions were limited. Second, weight loss behavior is a multifactorial phenomenon in that adherence [33], holidays, certain seasons, and important festivals [35], such as New Year’s Eve celebrations, may influence weight loss. Therefore, in the future, there should be greater focus on how to improve adherence to weight loss behaviors, and future studies over a longer period of time are necessary. Additionally, there was selection bias in the study.

This weight loss intervention campaign based on an official WeChat account focused on an occupation-based population in Shunyi District was found to be effective for males. The more active male participants were in using WeChat, the more weight they lost. There might be no effects or even negative effects on weight loss for females. Future research should focus on how to improve adherence to the WeChat weight loss interventions and to protect personal privacy, especially for females.