EMA and Content Analysis for Alcohol Marketing in Hong Kong In-school Adolescents

Study design

The proposed project is a combination of a time-based system-triggered ecological momentary assessment (EMA), which will collect real-time exposure to alcohol marketing and alcohol consumption, and a longitudinal survey. Consented participants will first complete a baseline questionnaire, followed by completing 3 regularly prompted EMA surveys via smartphone application (APP) every day for 21 consecutive days. The EMA will include questions of self-report exposure, brand capital (marketing receptivity and brand recognition), and alcohol consumption. One month after the EMA period, a telephone follow-up will be conducted to understand participants' drinking attitude and alcohol consumption.

Setting and subjects

The inclusion criteria will include (1) aged 13 to 18, (2) studying in secondary schools, (3) own a mobile smartphone with internet access, (4) stay in Hong Kong during the EMA study period, and (5) being able to read and write Chinese. To increase statistical power in analyzing alcohol consumption, we will recruit one-third (about 220) of the participants who are ever drinkers. To avoid seasonal variation in the ad exposure and drinking pattern (i.e. more ads and drinking in festivals), the recruitment will be evenly distributed within a year.

Sample size determination

The sample size calculation is based on the multilevel modelling of the estimated effect of alcohol marketing exposure on alcohol consumption and drinking attitude, and the mediation effect of brand capital (Figure 1 in Appendix 1). The R package simsem, MASS and lavaan were used for sample size calculation with Monte Carlo simulation run for 1000 times. Parameters of the multilevel modeling used for simulation was based on our pilot study, which showed that 210 participants will be needed to reach the threshold of 80% power, the comparative fit index and the Tucker Lewis Index larger than 0.90, the standardized root mean square residual less than 0.08 and the root mean square error of approximation less than 0.05.32 We conservatively assume that one in three students will report any exposure to alcohol marketing, therefore we need 636 (210/0.33) students to join the study. According to our first pilot study, the dropout rate of the EMA was 3.9% (2/51). Hence, 661 participants will be needed.

Recruitment procedures

The investigators will conduct both offline and online recruitment. The investigators have collaborated with the Hong Kong Council on Smoking and Health (COSH) to organize the leadership training (the Smokefree Teen project) annually since 2015. The investigators will promote the study to the secondary school students (n=200) participating in this training camp. The investigators will also send invitation letters to the 451 government and subsidized secondary schools. After receiving the schools' consent to join the study, our research assistant will contact these schools to plan the recruitment methods, including the distribution of recruitment posters and leaflets to the students with the help from the cooperators, organization of talks or recruitment booths to recruit participants in those schools. In addition, online recruitment advertisements will be disseminated in online discussion boards and social media (e.g., Facebook, Instagram) that Hong Kong adolescents usually visit. Volunteer participants can directly contact our research staff via telephone, email or submit an online application form via a Quick Response (QR) code. First, all volunteer participants will receive a web link for eligibility screening. Then our research assistant will contact eligible students via a phone call, and invite them for a face-to-face or virtual meeting via WhatsApp or Zoom. If they choose virtual meeting, the research assistant will mail a package including an invitation letter, consent form, and a pre-stamped return envelope, and request them to send back the signed consent form with the participant's and parent's signature. Afterwards, the research assistant will assist them in completing the baseline questionnaire, app installation and using the app in the virtual meeting. If participants choose the face-to-face meeting, the research assistant will complete all study procedures in the meeting, and request the participants to send their parent's signed consent after the meeting. Our second pilot study found that both virtual and face-to-face meeting are feasible in secondary school students.

EMA App

An EMA app will be developed in both iPhone operating system (IOS) and Android platform to administer the regular prompts automatically. Our research assistant will assist participants to install the EMA App in their own smartphone after obtaining study consent. To protect the privacy of participants, no personal information will be collected via the app. Participants only need to impute a 5-digit case number (such as last 5 digits of phone number) as an identifier. Participants will use the App to complete 3 time-based EMA each day for 21 consecutive days. Two systematic reviews on the length of the study period of EMA in the underage showed that compliance can be optimal when participants were prompted to complete about 3 EMA per day within 1 week or more. Once the App is installed and logged in, the participants will be asked to choose a start date for the 21-day EMA study. To facilitate the EMA on school days when they cannot use smartphone all the time, participants can create custom schedule of the 3 EMA prompts. The time gap between any two prompts must be longer than 3 hours to capture more information within the participants' waking hours. The first, second, and third prompt must range from 11:00 am and 1:00 pm, 4:00 pm to 6:00 pm, and 9:00 pm to 11:00 pm, respectively. All data will be uploaded to our server at home office immediately after each EMA completion for further analysis. Only the project administrators can access the datasets in the server with login name and password.

Follow-up

To reduce reactive effect from EMA, all participants will be followed up via telephone 1 month after completing the EMA survey. The survey includes drinking attitude and alcohol consumption. Participants are also requested to rate their satisfaction on EMA App, frequency and content of the EMA survey by a 5-point Likert scale at the follow up.

Consent

Participation in the study is voluntary. The recruitment staff will first explain the project's details to the potential subjects who agree to join the study before seeking consent. The potential subjects will be assured that they can withdraw from the study anytime without any prejudice, and all the information will be kept confidential and results will be reported in an aggregate format. For participants who choose to complete the recruitment procedures virtually, a research assistant will mail them a package including an invitation letter, consent form, and a pre-stamped return envelope, and request that they send back the signed consent form with the theirs and their parent's signature. For participants who choose to complete the recruitment procedures face-to-face, the research assistant will obtain their signed consent in the meeting, and request that they send their parent's signed consent after the meeting.

Statistical analysis

In the 1st research question, ordered logistic regression will be used to examine the association between alcohol marketing exposure from EMA and drinking attitudes from the follow-up survey. In the 2nd research question, descriptive statistics about the themes of alcohol marketing content will be computed. Structural equation modeling will be conducted to examine the research question 3a and 3b. In research question 3a, we will use alcohol marketing exposure and brand capital from EMA, and drinking attitude from 1-month follow-up in the model. In research question 3b, We will use alcohol marketing exposure, brand capital, and alcohol consumption from the EMA in the model. To avoid reversal causality between alcohol marketing exposure and alcohol consumption in research question 3b, the lagged alcohol consumption will be regressed on the alcohol marketing exposure in the previous EMA. In models for 3a and 3b, autocorrelation will be detected and controlled for in this model. 95% confidence interval will be estimated by using bootstrapping with size 10,000. Chi-square (χ2), root mean square error of approximation (RMSEA), comparative Fit Index (CFI), and Tucker-Lewis Index (TLI) will be used to evaluate model fit. Rules of thumb guidelines are that p-value for χ2>0.05, RMSEA≤.05, CFI≥0.95, and TLI≥0.95 represent a good fitting model. All data analysis will be conducted by using R (version 1.3.1073). All significant tests were two-sided with a 5% level of significance.