Modification of a Smoking Motivation Questionnaire for Chinese Medical Students

Smoking Intervention methods by physicians have been recognized potentially as a key factor in the prevention, reduction, and cessation of tobacco use and diseases related to smoking (Hum et al., 2011; Nunes et al., 2013). The physicians’ smoking habits, attitudes towards smoking, and advice to patients have tremendously impacted the anti-smoking campaign ( Araya et al., 2012). However, in 16 countries, nearly half of practitioners smoke frequently (Pipe et al., 2009). Although medical students know effective cessation-counseling (Kusma et al., 2010; Sreeramareddy et al., 2010), many medical students are smokers, (Warren et al., 2008) (only 3 sites surveyed have smoking rates less than 5% in 48 countries). Therefore, understanding the motivations of the smoking medical students is important for tobacco control. Reviews of the pharmacological actions of nicotine have determined an association with smoking psychology


Introduction
Smoking Intervention methods by physicians have been recognized potentially as a key factor in the prevention, reduction, and cessation of tobacco use and diseases related to smoking (Hum et al., 2011;Nunes et al., 2013). The physicians' smoking habits, attitudes towards smoking, and advice to patients have tremendously impacted the anti-smoking campaign ( Araya et al., 2012). However, in 16 countries, nearly half of practitioners smoke frequently (Pipe et al., 2009). Although medical students know effective cessation-counseling (Kusma et al., 2010;Sreeramareddy et al., 2010), many medical students are smokers, (Warren et al., 2008) (only 3 sites surveyed have smoking rates less than 5% in 48 countries). Therefore, understanding the motivations of the smoking medical students is important for tobacco control.
Reviews of the pharmacological actions of nicotine have determined an association with smoking psychology and physical dependence, and its influence on the biochemical and physiological functions of the brain (Litvin et al., 2010;Philip et al., 2013). However, motivations of initiating smoking are varied and multidimensional, which are upon individual differences. Exposing smokers to either external cues (e.g., pictures or smell of cigarette) or internal cues (e.g., negative affect induction) can increase urge to smoke and other behavioral and physiological responses and the two cues did not interact (Litvin and Brandon, 2010). Most smoking is beginning from adolescence. Approximately 80% of adult smokers begin smoking before 18 years (Philip et al., 2013). Studies on students of Nigeria, India, and Turkey indicate that being male and having parents or friends who smoke were more likely to initiate smoking (Golbasi et al., 2011;Muttappallymyalil et al., 2012;Odukoya et al., 2013;Ozturk et al., 2011). A study with 3, 706 undergraduate students from seven universities in England, Wales, and Northern Ireland found that smoker was low income, or those fathers had at least a bachelor degree, and binge drinkers (Ansari et al., 2012). While smoker was less likely to be healthy students, or those ate more than portions of fruit or vegetables, had never taken illicit drugs (Ansari and Stock, 2012). The medical education and the health risks knowledge of smoking could decrease the prevalence of smoking in adolescence, as the prevalence of smoking was significant higher in non-medical female students than medical female students in Saudi Arabia (Azhar et al., 2012). There is a high shisha smoking among Malaysia medical students because they believed that it does not contains nicotine, carbon monoxide, and tcan not lead to lung cancer, dental problems, and cardiovascular diseases. Moreover, having parents, siblings, and friends smokers of shisha, family problems, problems with friends, financial problems, and university life were all found to significantly associated with smoking status among medical students (Al-Naggar et al., 2012). Some participants also reported that the cardinal motive to smoke was the relief of negative moods such as anxiety, sadness, and stress (Al- Naggar et al., 2011;Spielberger et al., 1982). These factors are even more relevant to medical students considering the process for getting into and thriving in medical school. Thus, the smoking motivation of medical students still needs to be further investigated.
Russell's Smoking Motivation Questionnaire (RSMQ) constructed by Russell et al. (1974), containing 34 items (Russell et al., 1974), was an effective mental scale to evaluate the motives of smoking, and it was first used in West and Russell's study in 1985 as a 20-item scale (West et al., 1985). Recently, the RMSQ combined with the Reason for Smoking Scale (RSS) was translated into several countries (Berlin et al., 2003;Souza et al., 2009) due to its stable factor structure, internal consistency and temporal stability. Now there was an adapted Chinese version.
The Russell Reason for Smoking Questionnaire (RRSQ) has been popular instrument to evaluate risk factors for smoking across China. Yet, it has not been used in medical students to explore the factors of tobacco use, and has not been convinced whether it was appropriate due to the particularity of the medics that they grasped medical knowledge which may disturb their answering. The present study seeks to evaluate the validity and reliability of the subscales constructed from items in the RRSQ among medical students.

Study design and Participants
A cross-sectional survey was conducted in Xuzhou medical college in 2012. One thousand and one hundred fifty sophomore, junior and senior students were selected in this survey using cluster sampling. According to the standardized definition given by WHO, a current smoker was defined as a smoker who used tobacco on one or more days in last 30 days prior to the survey and who smoked more than 100 cigarettes in the past year (Global Youth Tabacco Survey Collaborative, 2002). The study was approved by Xuzhou medical college Ethics committee.

Questionnaire
The RRSQ (1999), derived from the RMSQ, contains 24 items which cover a variety of smoking risk factors. The 24 items were grouped into two dimensions, Pharmacology and Social Psychology, and obtained 8 oblique factors: Psychological image, Hand-mouth, Indulgent, Sedative, Stimulation, Addictive, Automatic and Supplementary Scale. Social psychology was identified to separate the psychological image, Handmouth and Indulgent factors from others such as pharmacology. Items were scored from 'not at all or uncertain' ('0') to 'very much so' ('3') and scores on the last two factors plus supplementary scale yielded a total dependence score. It was possible to be dependent for 6-point increase of total dependence score and likely to be addictive for 20-point increase.

Procedure
Monitors in 7 faculties helped to release the questionnaires after centralized cultivation, using the same leading words. Two weeks later, 50 students were sampled down to rewrite the RRSQ to evaluate the retest reliability through e-mails they had put on the scales.

Analysis
All analyses were performed using the SPSS version 16.0 and MPLUS version 6.1. Mean score of each item and subscale were obtained.
Construct validity, determining whether an instrument measures a construct as intended, was evaluated by confirmatory factor analysis (CFA), using a mean-adjusted WLS estimator (WLSMV) which was recommended for the analysis of small sample sizes less than 250 and variables with categories less than 5 through MPLUS version 6.1 (Beauducel et al., 2006;Rhemtulla et al., 2012). The degree of model fit with data was assessed by absolute fit indices, including the model Chi-square statistic (c 2 ), root mean square error of approximation (RMSEA) and incremental fit indices consisting of Tucker-Lewis index (TLI) and comparative fit index (CFI). A 90% confidence interval of REMSA both with p value of close fit was also included. The value of c 2 was not absolute since the Chi-Square statistic lacked power and this may not discriminate between good fitting models and poor fitting models when the sample size was small (Kenny et al., 2003). In general, a value of CFI ≥ 0.95 and a cut-off criterion of TLI ≥ 0.95 were presently recognized as indicator of good fit (Hu et al., 1999). Another significant fit index was RMSEA, value of which less than 0.08 would signify reasonable model fit (MacCallum et al., 1996).
In regard to convergent validity, spearman correlation coefficient between the scores of subscales and total scale, and those between subscales was judged using SPSS. Average variance extracted (AVE), measuring the amount of variance that was captured by the construct in relation to the amount of variance due to measurement error, was another index of convergent validity. Its recommended acceptable threshold was 0.5 or higher (Bagozzi et al., 1988;Fornell et al., 1981). (The computation formula is as following).   (1) The quality of individual item and reliability of measurement were assessed by single-item reliability and subscale reliability, all of which were evaluated by the results from CFA. According to classic definition of reliability, the real variation of the.and its recommended value was 0.2 or higher (Jöreskog., 1971). q is the residual variance.
(2) Composite reliability (CR) reflecting the internal consistency of each subscale, was calculated as following when the measurement error was uncorrelated (Jöreskog., 1971). The meaning of 1 was the standardized factor loading and that of q was the residual variance. The recommended value of CR was 0.6 or higher (Bagozzi and Yi, 1988;Fornell and Larcker, 1981). Whether the spearman correlation coefficient between the items and their own score were higher than the correlation coefficient between these items, and remainder subscales was also assessed to evaluate whether each subscale represented a separate domain through SPSS. (3)

Results
Of 1, 150 students who participated in the study, 1128 (557 men and 571 women) were included in the analysis and 22 were excluded because of their empty and incomplete questionnaires. Among those medical students, 111 students were identified as a current smoker. All of these smokers were male, and 27, 34, and 50 were sophomore, junior, and senior students, respectively.

Validity
Through the result of first model, model fitting was found to be satisfactory (Table 1). However, the factor loading of first-order factor supplementary was higher than 1 and the correlation coefficient between factor 'supplementary' and factor 'addictive' was 0.99, meaning multi-colinearity existed between Supplementary and Addictive. With West and Russell's point of view (West and Russell, 1985), therefore, Supplementary and Addictive were converged to the same factor Addictive.
After correction of the scales the c 2 was 351.65, TLI was 0.96, CFI was 0.96, and REMSA was0.064. The 90% confidence interval of REMSA was 0.048 to 0.077, and test for goodness of fit showed a p-value of 0.077) ( Table  1).
The mean score of each item and subscale of corrected scale are shown in Table 2. Model fit results showed that first-order factors Psychological Image, Hand-Mouth Indulgent, Sedative, Automatic, Stimulation and Addictive, as the indicators of second-order factors Pharmacology and Social psychology, had the high factor loadings in this model (0.82, 0.91, 0.92, 0.95, 0.98, 0.91, 0.97 separately). The factor loading of each item was from 0.61 to 0.87 which was significant.
The spearman correlation coefficients for content validity between factors and total scale ranged from 0.44 to 0.90. The AVE values met the criterion of 0.5, ranging from 0.66 to 0.78 (Table 3).

Reliability
Considering the results above, single-item reliability ranged from 0.30 to 0.76 for all subscales. The results with regard to composite reliability showed all coefficients were greater than 0.6. Each of correlation coefficients between items and their own subscales was greater than that between this item and the remainder subscale, showing all subscales represent separate domains (Table 4).

Discussion
Although smoking questionnaires have been used widely for studying motivation of using tobacco, few of them were customized for medical students. Our study assessed the reliability and validity of the pilot questionnaire designed specifically for this population. In this study, approximately 10% of medical students smoked, who were all males. This finding of high prevalence of smoking among male medical students is consistent with the literature (Warren et al., 2008).
Our results also showed that multi-colinearity emerged in the original Russell reason for smoking questionnaire (RMSQ) when it was used in medical students. The strong correlation between factor Addictive and Supplementary may be attributable to the medic's prior understanding of addiction reflected by the items. In addition, Addictive and Supplementary can be treated as the same factor (West and Russell, 1985). Due to their high correlation, the last model of RRSQ had 7 factors instead of 8 identified in the initial version. With the current version, the goodness of model fit was satisfactory due to the significant CFI, TLI and REMSA. The correlation between factors and average variance extracted suggest that the last model had a good convergent validity. Each factor could account for more than 50% variance captured by the construct. Acceptable single-item reliability, composite reliability and higher correlation of item-own confirmed the quality and reliability of the measure and its items.
To our knowledge, this is the first study conducted to assess the reliability and validity of the standard smoking questionnaire RRSQ within medical students this special population. Nevertheless, our study had some limitations. First, Flynn (Flynn et al., 2001) suggested that a ratio of five responses per parameters was required to obtain reliable estimates based on 120 participants. Our study had slightly lower sample size, though the scale seemed to be a reliable measure of smoking motivation according to the item-single reliability, composite reliability and validity.
Second, the study was conducted in one medical college and all the participants are male It might not be generalized to other medical colleges in national across More medical colleges to assess the reliability and validity of the scale used to ensure similar gender distribution are warranted. Third, smoking motivation, such as personal relationship and personal resource needed to address (Bowen et al., 2012). The test-retest reliability should be examined for these scales using the second questionnaire. Due to the low response rate , this validation study is not avaiable.
In conclusion, our results indicate moderate to high reliability and validity of the RRSQ among medical students. Further investigation is warranted to validate this tool and examine its generalizability.