This cross-sectional survey was conducted among university students in Guangzhou, China, from March 1, 2016, to April 20, 2016. A multistage stratified cluster sampling method was used. Selection criteria included all full-time students of universities located in Guangzhou admitted from 2013 to 2015 but excluded students whose majors were sports-related.

Universities in Guangzhou were divided into first and second class according to the Education Examinations Authority of Guangdong Province [31]. Two of the first-class universities and three of the second-class universities were selected using purposive sampling. Student majors from the universities were divided into five categories (natural science, agricultural science, medical science, humanities and social science, and engineering and technology science) according to China’s National Classification and Code Disciplines [32]. For each major category, we recruited at least 3 classes from each grade (students admitted in 2015, 2014, and 2013) separately. Contact persons were recruited from the selected universities. After training on study purpose, procedure, data collection, and quality control, they served as helpers to approach different classes and collect data. These contact persons distributed the questionnaires among their classmates, collected the answered questionnaires, and performed a preliminary verification of the quality of all answered questions. Financial reimbursement of 25 yuan (US $4) per hour was provided to them as compensation for their time.

According to UTAUT, a commonly used theory for identifying determinants of intention to use health technologies [29], direct determinants of intention are performance expectancy (PE, the degree to which using a technology will provide benefits in performing certain activities), effort expectancy (EE, the degree of ease associated with the use of the technology), and social influence (SI, the degree to which an individual perceives that important others believe he or she should use the system). Associations between PE, EE, SI, and intention are moderated by variables such as age and voluntariness of use. As PA app use is voluntary and age range does not vary much among university students, moderators were replaced by BMI (Figure 1).

The questionnaire was anonymous and self-administered. Background variables including university, major, grade, sex, ethnicity, being the only child or not, hometown type, monthly living expenses (using 1000 yuan as a cutoff based on living expenses of university students in Guangzhou), and BMI were collected. BMI was calculated by self-reported height and weight, using 18.5 and 24 kg/m² as cutoffs. BMI between 18.5 and 24 was defined as normal and BMI outside the range as abnormal (ie, BMI <18.5 means underweight and BMI ≥24 means overweight or obese, as defined by China’s Obesity Working Group). We measured intention to use PA apps with one question: Will you use PA apps in the coming 6 months? The experience of having used PA apps before was defined as having used PA apps at least once in the last 6 months. The questionnaire took about 5 to 8 minutes to finish.

PE, EE, and SI were measured with items generated by the research team according to literature and qualitative interviews. An item pool of questions used to measure UTAUT according to literature review was formed, and items were screened by the research group. We interviewed some university students to understand their thoughts on exercise app use according to UTAUT and collect their suggestions on the screened UTAUT items. A revised version of the UTAUT-related questionnaire measuring determinants PE, EE, and SI was developed after the research group discussion.

The determinants were assessed using multi-item measures scored by summing relevant item scores. Each item was scored on a 5-point Likert-type scale (1 = extremely disagree, 5 = extremely agree). Using a principal component analysis, factors were identified for the PE, EE, and SI scales, explaining 80.7%, 82.2%, and 76.3% of total variance, respectively. Cronbach alphas were 0.88, 0.89, and 0.68, respectively.

Univariate logistic regressions were performed to measure the associations between background variables, experience using PA apps in the last 6 months, and intention to use PA apps in the coming 6 months. Background variables with P<.10 in the univariate analyses were selected by a multivariate model using a stepwise method except for experience using PA apps in the last 6 months. Both univariate and multivariate analyses (adjusting for significant background variables and experience using PA apps in the last 6 months) were performed to calculate the association between the UTAUT construct and intention to use PA apps in the coming 6 months.

To identify the interaction effects of BMI on the associations between UTAUT constructs and intention to use PA apps, we performed multivariate logistic regression models, adjusting for significant background variables.

Among the students (from 55 classes) contacted, the response rate was 94.1% (1603/1704), and the effective response rate was 85.7% (1461/1704). Of all participants, 64.8% (947/1461) were from first-tier universities, 50.3% (735/1461) were male, 61.5% (899/1461) were not the only child, 48.9% (714/1461) were from a town or rural area, 67.8% (991/1461) had over 1000-yuan (US $150) monthly living expenses, and 33.3% (487/1461) had a BMI beyond the normal range (24.1%, 352/1461, were lower than the normal range while 9.2%, 135/1461, were overweight or obese; Table 1).

Of the participants, 41.8% (611/1461) intended to use PA apps. In univariate analysis, all background variables except for students’ grades and ethnicity were significantly associated with the intention to use PA apps (Table 1). In the multivariate analysis, students who were female (odds ratio [OR] 1.36, 95% CI 1.10-1.68) and from a capital city or municipality (OR 1.43, 95% CI 1.11-1.86) with monthly living expenses were over 1000 yuan (OR 1.44, 95% CI 1.14-1.83) and BMI beyond the normal range (OR 0.71, 95% CI 0.56-0.89) were more likely than others to intend to use PA apps (Table 1). Meanwhile, compared with those not having used PA apps in the last 6 months, participants having used apps in the last 6 months were more likely to intend to use PA apps in the coming 6 months (OR 4.16, P<.001; Table 1).

In the univariate analyses, all 3 UTAUT-related scales were significantly associated with intention to use PA apps (OR_u 1.22-1.49, P<.001). Such associations remained significant after adjusting for significant background variables (gender, hometown type, monthly living expenses, and BMI), and experience of using PA apps in the last 6 months (OR_a 1.10-1.31, P<.001).

In the associations between each UTAUT-related item and the studied outcome, all 8 items were significantly associated with the intention to use PA apps (OR_u 2.16-2.97; Table 2).

One out of the three models considered presented statistically significant interaction effects: interaction between the BMI and the performance expectancy scale (beta 0.10, P<.001; Table 3). Higher scores on the PE scale (x-axis) were associated with higher log odds for intention to use PA apps (y-axis), but the strength of associations depended on the BMI, as seen by the slopes of the straight lines (Figure 2). The significant moderating effect indicated stronger associations between the PE scale and the intention to use PA apps among those whose BMI was beyond normal range (BMI≥24), as compared with those whose BMI was within normal range.

To the best of our knowledge, this study is among the first to examine university students’ intention to use PA apps based on a new technology use–related model in China. Overall, university students showed a relatively high level (611/1461, 41.8%) of intention to use PA apps for increasing PA. Moreover, the number of participants having the intention to use PA apps in the coming 6 months was higher than the number of participants having used PA apps in the last 6 months (611 vs 537), which indicated that the intention to use PA apps of university students in Guangzhou showed an upward trend. This relatively strong intention suggests that future efforts to increase PA among university students with PA apps will be promising.

Of all the participants, 24.1% (352/1461) had a BMI lower than the normal range and 9.2% (135/1461) were overweight or obese according to their BMI. The rates of lower body weight and overweight or obese were both higher than the results of research conducted in Henan Province (low body weight: 14.7%; overweight or obese: 8.5%) [33] and northern Anhui Province (low body weight: 12.6%; overweight or obese: 8.3%) [34] among university students. But the rate of overweight or obesity in our study was a little lower than that found in one study conducted on university students in Zhejiang (low body weight: 8.2%; overweight or obese: 10.9%) [35]. Although there are regional differences in BMI, our results still indicated that the situation of university students’ BMI in Guangzhou being beyond the normal range was relatively serious. The physical health status of university students in Guangzhou is thus in urgent need of improvement.

In our study, PE, EE, and SI were all positively related to PA app use intention after adjusting for significant background variables. UTAUT can therefore be used to develop implementation interventions to increase the use of apps designed for improving PA among university students. SI showed a relatively strong association with the intention to use PA apps among the three scales with OR 1.49 (95% CI 1.39-1.60). This may be due to the community lifestyle of university students in China. University students live on campus, spend most of their time in class or in the dormitory, have close contact with their peers (classmates or roommates), and are easily influenced by their peers [36]. A survey conducted in European universities reached a similar conclusion that students’ alcohol use behavior was affected by their peers’ alcohol use behavior [37]. Social influence mainly refers to the influence from surrounding people and environment, which can explain why social influence plays a more important role in affecting the intention to use PA apps. To promote university students’ intention to use PA apps, schools or society may be able to achieve their goals with the aid of peer influence.

UTAUT-based interventions seem to be useful among all students regardless of their BMI. However, the strength of associations between UTAUT and PA app use intention may depend on other contextual factors. UTAUT-based interventions may have a better chance of success if they pay more attention to university students whose BMI are beyond the normal range. Such findings suggest that to improve physical activities among university students via app use, for students with BMI beyond normal range interventions enhancing PE (eg, peers sharing of benefits) could be considered, but for students with normal or low BMI intervention targeting PE alone may be less effective.

This study has some limitations. First, the observational design may not establish a causal relationship between independent variables and outcome. Although we collected data of use intention in the coming 6 months and use experience in the last 6 months, which guaranteed the time sequence, participants’ intention might still be based mainly on the time point of the survey. It was assumed that cognitive situations would be quite stable in the coming 6 months.

Second, although anonymity and privacy were guaranteed, a reporting bias due to social desirability and self-expectation may still exist. For example, as university students, the participants might give a high score to the EE scale. There were at least two items in each scale to try to avoid this problem, and Cronbach alpha was high for each scale.

Finally, we cannot assume that the results can be extrapolated widely without further research. A purposive sampling method was used to recruit participants, which may weaken the external validity of our sample. In addition, our study involved only one city and this city’s economic level is higher than most other cities in China, which might affect participants’ acceptability of new technology and health consciousness.

This study, based on a theoretical approach to technology use, indicated which factors will need to be addressed to design an effective implementation intervention for the use of PA apps to increase PA among university students. Our findings indicated that university students’ intention to use PA apps was influenced by UTAUT-related constructs, but potential moderating effects of BMI should be kept in mind when UTAUT-based interventions are being developed. Different intervention strategies should be considered for students within and beyond the normal range of BMI.