EMA_Normalization of Alcohol Drinking in Young Adults

Study design

The proposed project is a combination of (1) a time-based system-triggered EMA, which will collect real-time exposure to alcohol advertisements and marketing, and (2) a longitudinal RCT of incentive strategies survey in current alcohol users. The participants will first complete a baseline questionnaire and then participate in EMA via a smartphone application (App) for 21 consecutive days. This assessment length was designed to capture participants' exposure to alcohol marketing for a long period of time. The EMA will include questions about self-reported exposure to and /or responses to alcohol marketing. One month after completing the EMA, participants will be contacted to complete a telephone follow-up survey. (The one-month follow-up shall understand participants' perception and receptivity of alcohol marketing and drinking behavior due to alcohol marketing exposure and alcohol consumption.)

Subjects

The investigators will recruit 360 participants with the following inclusion criteria: (1) aged between 18 and 35 years, (2) own a mobile smartphone with internet access, (3) drank any alcohol in past 30 days (current drinkers), (4) will be in Hong Kong during the 3-week EMA study period, and (5) able to read and write in Chinese.

Procedures

Recruitment strategy

A mass email containing the study information will be sent to all undergraduate and postgraduate HKU students and staff for recruitment with up to three email reminders describing the research. Also, online recruitment advertisements will be posted in different online discussion boards that Hong Kong university students usually visit. Voluntary participants can directly contact our recruitment staff via telephone, email, or submitting the online application form for a subsequent briefing meeting. Our pilot study has shown that these methods are feasible to recruit a large number of interested participants. After participants registered online, they will be re-directed to a web link for eligibility screening. Research assistant shall contact eligible participants for instructions on the consent form, baseline questionnaire and guidance on installing the EMA app and further usage.

EMA App

After the participant gives consent and completes the baseline questionnaire, a trained research assistant will install an EMA App on the participant's smartphone, and give instructions on using the App. The App will be developed for both iOS and Android platforms to administer the regular EMAs automatically. To protect the privacy of the participants, no personal information (e.g., names and telephone numbers) will be collected by the App. Only a 5-digit case number will be used for identification. Unless otherwise specified, the case number will be the last five digits of the subject's phone number. Once the App is successfully installed, the user will be asked to choose a start date for the 21 consecutive days study period within the coming three weeks and to set up a lunch hour between 11:00 am and 4:00pm in these study days to undertake the regularly prompting EMA.

On each study day, the App will regularly prompt the participants to complete a questionnaire about their recent exposure to alcohol advertisements in different channels every 5 hours for a total of three times that approximately match the time of lunch, dinner and before sleep at night, and alcohol consumption. Existing meta-analysis on the compliance rate of EMA studies found that among nonclinical studies with participants consists of the youth population, the highest average compliance rate was observed in studies that prompted the questionnaire 2-3 times daily (91.7%) compared with more frequent prompts of 4-5 times (77.4%) or 6+ times (75.0%)45. This meta-analysis also showed that the 3-week length of EMA yielded the highest compliance rate (84.0%) compared with 1-week (75.8%) and 2-week (76.8%) length. The 21-day time period was also the average number of days (SD = 28; range of 2 to 175) of assessment based on previous systematic reviews examining EMA studies on substance abuse46.

Moreover, as alcohol consumption tends to occur late at night, it is also important to capture the drinking behavior within the full range of participants' waking hours. Therefore, the first EMA at each afternoon shall record the hang-over effects from the previous night's drinking47 and the last EMA would capture the drinking behavior that mostly would happen48. For example, if a participant gets up late from hang-over and sets the lunch hour to 1:00 pm, he/she will be asked to complete an EMA survey at 1:00 pm, 6:00 pm and 11:00 pm on that day. Each EMA survey will take about 1 minute to complete (the last EMA of each day may last 2 minutes). If the user cannot respond to the EMA survey prompt immediately, he/she could respond to the EMA prompt within 30 mins in two further reminders at 15 mins and 5 mins. Each prompt not responded to for 30 minutes will be regarded as a non-response. All the above data, including the date and time of completing the EMA survey and responses to the prompts, will be sent to our server immediately for further analysis.

Randomization, allocation concealment, and blinding

All eligible participants will be randomly assigned to control or intervention groups using a random sequence generated by an online sampling randomizer. Before the participants' consent, the group allocation was concealed for both participants and interviewers. All participants and interviewers at baseline and 1-month follow-up are not blinded.

Follow-up

To minimize the survey effect of EMA on the study outcomes, a telephone follow-up to assess the study outcomes will be conducted 1 month after the completion of the EMA. The survey included the outcomes of drinking normalization, alcohol consumption level, and their satisfaction level regarding the baseline survey, mobile App, EMA surveys, follow-up survey, and the overall rating of the study.

Data analyses

Data analysis will be conducted using STATA (version 15). All significance tests were two-sided with a 5% level of significance. The internal consistency of the constructs for perceived popularity, perceived social approval and positive expectancy will be assessed with Cronbach's alpha. In each construct, the investigators will sum all items to build the score for that attribute in each participant. The investigators will create a composite measure of receptivity by averaging or summing the 5 items of receptivity documented in the EMAs in each participant.

For the 1st research question, independent t-test/ Mann-Whitney U test will be used to test the difference of compliance rate for EMA survey between groups. Data will present as mean ± standard deviation (SD).

For 2nd to 6th questions, the investigators will use multiple linear regression model to assess the exposure frequency to alcohol marketing on the three attributes of participants' normalization of drinking (the 2nd research question), and alcohol consumption (the 3rd research question). Multiple linear regression model will also be used to assess the association between receptivity to alcohol marketing and drinking normalization (the 5th research question). Multilevel modeling will be used to answer the second question to explore the influence of exposure frequency to alcohol marketing on receptivity to alcohol marketing (the 4th research question), and the association between receptivity to alcohol marketing and alcohol consumption (the 6th research question).

Exploratory factor analysis (EFA) will be adopted to explore factor structure of perceived popularity, perceived social approval, and positive expectancy (the 7th research question). The investigators will use Bartlett test of sphericity and Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy to do the correlation test. If the P value of bartlett test less than 0.05 and KMO value large than 0.5, factors will be used for further analysis. In principal component analysis, factors with eigenroots greater than 1 will retain. Varimax will be used to factor rotation. Items with factor coefficients of greater than or equal to 0.5 were required for interpretation of factor structures. Inter-item reliability was assessed by Cronbach's α. Accordingly, Cronbach's α of greater than 0.7 was considered to be an acceptable reliability coefficient of the scale.

To answer the 8th research question, structural equation modeling will be adopted to examine the pathways of exposure frequency and receptivity to alcohol marketing, perceived popularity, perceived social approval, and alcohol consumption level. The model will be adjusted for participants' age and sex. For within-subject effects, the investigators will lag the data to prevent occurrence effect if the participants were exposed to alcohol marketing when purchasing alcohol. The investigators will include the exposure to alcohol marketing reported in the index reports (t0) to predict the subsequent report of alcohol consumption. The elapsed time between index (t0) and subsequent reports will be included as a level-1 (within-subject) covariate. Autocorrelation will be detected and controlled for in the model. Bootstrapping with size 10,000 will be used to estimate the 95% confidence interval of the parameters. After checking data distribution normality, several fit indices will be served as the judgement of the fit of the model. A non-significant chi-square (X2) is desired, but the test is sensitive to erroneously reject a correct model when the sample size is not large enough.57 The Comparative Fit Index (CFI) and the Non-Normal Fit Index (also known as the Tucker-Lewis Index), ranging from 0.00 to 1.00, with higher values representing better fit, will be adopted.58 The standardized root-mean-square residual will be used to detect model misspecifications, with values below 0.08 considered desirable.58 The indirect effect of the model will be checked with Sobel test. Similar model will be conducted to detect between-subject effects.