Subject Selection
The cross-sectional was conducted between July and November of 2018 in Tabriz, Iran with 387 current smokers. To recruit the participants, 10 parks and coffee shops in Tabriz’s 10 districts were randomly selected. The sample consisted of the youth who ranged in age from 18 to 30 years, reported being smokers, were not taking any medications for psychiatric disorder, voluntarily agreed to participate in the study, and signed the consent, representing a non-random sample. Data were collected through a self-administered questionnaire. Cigarette packs with health warning messages, which had been prepared by the research team, were presented to the participants before completing the questionnaires. The Ethics Committee of Tabriz University of Medical Sciences had approved the study.
Instrumentation
A panel of Seven (7) health psychologists and eight (8) health education and promotion experts examined the content validity of the instrument by rating each questionnaire item for relevance, clarity, representativeness, and essentiality. The means of the content validity index (CVI) and the content validity ratio (CVR) were 0.87 and 0.95, respectively, attesting to the content validity of the instrument. The instrument was pilot-tested with 50 young smokers to examine its utility.
Demographics
The demographic variables included age, gender, living arrangement (with parents, personal home, dormitory), marital status (single or married), employment status (full-time, part-time, unemployed), and the highest level of education. Additionally, history of hookah use, alcohol use, drugs abuse, smoking among friends and family members, the number of cigarettes smoked per day, smoking behavior after waking up and the first bidder of cigarette smoking were measured.
Message Processing Route
For evaluating processing routes of peripheral and central, based on the ELM, we included two most influential factors of “motivation” and “ability”. The extent of motivation was determined by the attitude towards the message, personal relevance, and the need for cognition. Additionally, individuals' ability for elaboration was operationalized by “distractions” and “knowledge”. In our study, the midpoint of the sum of motivation and ability was used to categorize the processing route into the peripheral (less than 2016.65) or central (greater than 2016.65).
The attitude towards the message was measured by 12 items developed by the researchers; for example, “pictures motivated me to reduce my daily number of cigarettes smoked”. Additionally, we developed 3-item scale to assess perceived relevance; for example, “in my opinion, the pictures on the cigarette packet talked about my health conditions.” The need for cognition was measured by the 6-item version of Cacioppo and Petty’s (1982) scale that was proposed by Lins de Holanda Coelho et al. (2018) (24, 25); for example, “I would prefer complex to simple problems”. A 5-point Likert-type scaling (1 = extremely uncharacteristic of me, 5 = extremely characteristic of me) was used. Reliability coefficients for the attitude towards the message (α = 0.92), perceived relevance (α = 0.82), and need for cognition (α = 0.71), attested to the internal consistency of the scale scores.
Additionally, ability was measured by knowledge and distractions, utilizing two scales that were developed by the research team. Specifically, an 8-item scale was used to measure the knowledge about the potential negative consequences of smoking cigarettes; for example, “smoking can cause lung cancer.” Responses were coded as 0 = no/don’t know or 1 = yes. A 4-item scale was used to gauge distractions, utilizing a 4-point Likert-type scaling (1 = never, 4 = always); for example, “presence of people around me caused to lose my focus on pictures and smoking outcomes.” The reliability coefficients for the knowledge and distractions were 0.67 and 0.62, respectively.
Cognitive Variables
Perceived severity. To measure the seriousness of smoking risks, Harris’s four-item scale of perceptions of personal risk about smoking and health was employed (26); for example, “smokers live shorter lives than non-smokers” and “smoking increases your chance of getting lung cancer”. The reliability coefficient for the scale was 0.75.
Sensation-Seeking. A published 8-item questionnaire was used to assess sensation-seeking behavior (27); for example, “I would like to explore strange places”. The estimated reliability coefficient was 0.82.
Psychological dependence. A four-item scale, derived from Autonomy Over Smoking scale (28) was used to measure psychological dependence; for example, “I rely on smoking to focus my attention" and "I rely on smoking to take my mind off being bored.” The reliability coefficient for the scale was 0.80.
Smoking abstinence self-efficacy. A 12-item instrument (SASEQ) was used to assess self-efficacy (29); for example, ‘‘you feel very sad, are you confident that you will not smoke?’’ The reliability coefficient for this scale was 0.80.
Positive attitude toward smoking. A 9- item researcher-made instrument was used to gauge participants’ attitude toward smoking; for example, “smoking makes me look attractive” and “smoking makes me feel independent.” The reliability coefficient for the scale was 0.67.
Cognition reaction. A 5- item researcher-made scale, which was based on a published study (29) was used to gauge the cognitive response that participants felt after seeing the images; for example, “I felt scared after seeing my pictures.” The reliability coefficient for the scale was 0.76.
With the exception of ability, we used a 5-point Likert-type scaling (1 = the lowest, 5 = the highest) to measure the abovementioned scales. For the purpose of the SEM, all were standardized, ranging from 0 to 100.
Health Warning Messages on Cigarette Packs
The research team developed four (4) health warning messages to stick on cigarette packs that were pictorial in nature, because they are known to be more effective than are the text warnings (29). Given that health warning messages that emphasize the physical consequences of a threat are helpful in informing and encouraging people to engage in preventive health behavior (30), we used images related to smoking-related (e.g., cancers of the respiratory system, mouth, and teeth, Buerger's disease).
Data Analysis
To analyze the data, the Statistical Package for the Social Sciences (SPSS), version 23, and Mplus software, version 6, were used. Descriptive statistics, Mean (SD) and frequency (%), were used to summarize the data. The respondents’ responses to the questionnaire items were used to measure each scale score. The normality of all distributions was examined by skew and kurtosis indices. A series of Chi-square Test of Independence was performed to examine the simple associations between the processing route and demographic characteristics. To compare the cognitive constructs scores in processing routes, a series of independent sample t-test was applied. The significance level for all analyses was set, a priori, at 0.05.
Applying the Mplus software, we performed structural equation modeling (SEM), with maximum likelihood estimation, to test the hypothesized model for cognitive predictors of the processing route in full sample (Model A) and gender groups (Model B). The SEM included model specification, identification, estimation, testing, and modification. The first step focused on the conceptual model regarding the hypotheses. The second step consisted of the model fit process, wherein the number of input and output parameters were suitably chosen. The maximum likelihood estimation was performed in the third step. The fit indices were assessed in the fourth step. In the fifth step, the modification indices were used to modify the model. Model fit measures were attained to judge how well the proposed model captured the covariances between all measures. Since the quality of fitted models is influenced by the sample size, multiple model fit indices were estimated, which were as follows: χ2 (p > 0.05), χ2/degrees of freedom < 5, the Root Mean Square Error of Approximation (RMSEA ≤ 0.08), the Standardized Root Mean Square Residual (SRMR ≤ 0.05), the Comparative Fit Index (CFI ≥ 0.90), and Tucker-Lewis index (TLI ≥ 0.90) (31).