To conduct an EMA study, researchers prepare survey questions and plan the scheduling of data collection points in advance. Then an administrator registers survey questions in advance in an EMA portal web application and establishes the schedule for survey questions. Furthermore, a participation schedule is set up separately for each participant (subscription). According to the schedules determined previously, the scheduler on the service uses Firebase Cloud Messaging (FCM) and Access Point Names (APNs) to automatically send alarms regarding responses from a participant. In some cases, an alarm can be sent out several times for just one scheduled question. To encourage participants to respond actively during the subscription period, gift-icons can be sent as an incentive. As for the gift-icons, various goods can be provided through a separate company, and depending on the case, interoperating with other companies is also possible.

Various response data collected in this process can be analyzed by using a data analysis tool. The mobile EMA service can extract the data in comma-separated values (CSV) format from such various data, thereby providing users with opportunities to analyze such data by using a tool familiar to them or any specialized tool. Furthermore, a stable service can be provided through cloud networking, and if necessary, it can be easily scaled up/scaled out. The structure of the mobile EMA service is presented in Figure 1.

The process of the mobile EMA service is summarized as follows. After an administrator, usually a researcher prepares the survey questions and gets ready to distribute them to participants. The administrator sends the application URL to a participant. The participant downloads the application to her/his smartphone and sends a request to be registered to the administration. Then the administrator allows the participant to register, and the study participant registers with a username and password. The administrator sets up subscription dates and time schedules for each participant. According to the date and time determined, a push notification is sent to the corresponding participant, and the participant receiving the push notification can answer the survey questions immediately, which is an advantage in comparison to the conventional paper survey method. When the participant completes the survey, she/he submits the responses to the administrator. If the survey questions are not answered on time, the researcher can send a push alarm separately to the corresponding participant as a reminder to participate in the survey. The data sent to the administrator are stored on a cloud server and exported in CSV or Excel format for further analysis (Figure 2).

In the case of using the mobile EMA service, the subscription dates can be set differently for each participant, and the push notifications can be defined according to the corresponding subscription dates. Therefore, the researcher can obtain the participants' momentary data, for example, momentary affect and stress that the participant experiences at a given moment.

The conventional paper survey method required as much efforts as the number of survey participants in terms of organizing the question items and collecting the data. Questions on paper entail a reliability issue for the responses of participants (e.g., responses are all composed at the same time, response is not composed on time, and participants may omit a question).

Meanwhile, in mobile EMA design, researchers can easily create the corresponding survey items, compose the desired questions using various formats and setup an explanation for each question. This will help greatly in saving resources and time, thereby helping the researchers perform activities more efficiently. When comparing digitalized survey method to the conventional paper survey method, a participant can manage and respond to the survey questions by using mobile devices easily. The administrator can receive the survey responses from each participant without laboring to transfer the data collected on papers to Excel one by one. Furthermore, the response results can be obtained as raw data in an Excel file. When multiple participants are targeted, the registration and management of participants are easy in comparison to the paper survey method. When a response request signal is received through a smartphone, a participant responds immediately by recording data, such as current activity, environment, affect, stress, and so forth. The administrator can distribute survey questions with a minimal effort since functions such as sending alarm notifications for response requests to each participants are automated (Figure 3).

A convenient sample was drawn from police officers who were recruited from the police stations in Gyeonggi Province of the Republic of Korea in 2018. The study was approved by the Institutional Review Board of Yonsei University (No. Y-2018-0035). Our research team installed a mobile application into the participants' smartphones, and participants were requested to answer a set of questions of a total of 25 items assessing their momentary affect (13 items; 9 for positive affect and 4 for negative affect) and stress (8 items), including their shift of work, what they do, whom they are with, and where they are at the moment of responding to the questions. Most participants spent 2 to 3 minutes to answer those questions. The momentary affect and stress questions [1516] were translated into Korean language using translations and back-translations and obtained content validity from a group of experts in the related fields. We also measured the participants' perception of the data quality and technical challenges in using this mobile application in terms of usability and these items were derived from the previous study utilizing EMA and modified for this study [17].

We asked multiple repeated responses, four times a day (08:00 to 9:00, 13:00 to 14:00, 17:00 to 16:00, and 21:00 to 22:00) for seven consecutive days, which yielded 28 data points per person. The application sets alarms at semirandom interval with 60-minute blocks. For example, if the notification alarm for a morning survey is sent at 8:30 to the participant, which is randomly set by the EMA application, she or he is supposed to respond by 9:30, within 60 minutes.

The participants were 97 police officers, and the mean age was 37.7 ± 10.8 and ranged from 22 to 60 years old. Ninety-two persons (94.8%) were males. About two-thirds (66.0%) were college graduates or higher. The mean completion rate was 60.0%, ranging from 3.5% to 100% per person. For affect, the means of positive and negative affect were 18.34 ± 4.57 of total score of 28 and 19.09 ± 9.75 of total score of 63, respectively. The mean of momentary stress was 7.92 ± 5.76 of total score of 40 (Table 1).

We also assessed the usability of the mobile EMA application from the participants. Eighty participants responded that the mobile application correctly measured their affect (4.34 ± 0.83) and stress (4.48 ± 0.62) on a 5-point Likert scale. They also felt that the mobile application was easy to use (4.05 ± 1.04), and they were comfortable to answer the questions in the application (4.11 ± 0.89) installed on their smartphones (Table 2).