Psychometric Properties of Drinking Motives Questionnaire-Revised (DMQ-R) in Spanish Adolescents

The Drinking Motives Questionnaire-Revised (DMQ-R; Cooper, 1994) is the most widely used questionnaire to assess drinking motives. This research aimed to adapt and study the validity and reliability of the Spanish DMQ-R and its short form in a sample of adolescents. We assessed 504 drinkers at Time 1 (T1, Mage = 14.15, SD = 0.60, 52.38% female), of whom 238 participated in a 1-year follow-up (T2, Mage = 15.05, SD = 0.54, 58.82% female). They completed the DMQ-R at T1, the alcohol use measure at T1 and T2, and the alcohol-related problems questionnaire at T2. We performed Confirmatory Factor Analyses (CFAs) to test the structure validity of the questionnaire, Cronbach’s alphas to test the internal consistencies of the scales, and path analyses to test the concurrent and predictive validity of motives on alcohol outcomes. CFA indicated that the short form of the 4-factor model best fitted the data. Cronbach’s alphas were .70 or higher. Direct effects of the path analysis showed that social motives cross-sectionally predicted alcohol use, while coping motives prospectively predicted alcohol-related problems. Indirect effects showed that social motives prospectively predicted alcohol use and problems. To conclude, the DMQ-R short form appears to be reliable and valid to assess drinking motives among Spanish adolescents.

Psychometric properties of Drinking Motives Questionnaire-Revised (DMQ-R) in

Spanish adolescents
Alcohol is the most used drug, and age of initiation is around 13-14 years (National Plan of Drugs, 2013). Alcohol drinking during adolescence constitutes the main risk factor to disability-adjusted life years worldwide (DALYs) (Gore et al., 2011), and has also been associated with several health problems in adulthood (McCambridge, McAlaney, & Rowe, 2011). Thus, understanding adolescents' motivations to drink may help prevent and reduce its use (Cox & Klinger, 1988). Accordingly, the main aim of the present study is to examine the most widely used instrument to assess drinking motives (Kuntsche, Knibbe, Gmel, & Engels, 2005), Drinking Motive Questionnaire Revised (DMQ-R, Cooper, 1994), in Spanish adolescents.
Inspired by Cox and Klinger's (1988) Motivational Model of Alcohol Use (MMAU), which postulates that people drink to obtain valuable outcomes, Cooper (1994) proposed four drinking motives based on the (a) type of reinforcement desired (positive or negative reinforcement), and (b) the source of reinforcement (internal or external). Crossing these two dimensions results in four distinct drinking motives: social motives (external, positive) refer to drinking to facilitate social relationships; enhancement motives (internal, positive) refer to use of alcohol to increase positive affect; conformity motives (external, negative) relate to drinking in order to form part of a group of people; finally, coping motives (internal, negative) relate to drinking to manage negative affect.
Based on this theoretical model, Cooper (1994) adapted to adolescents several items from the DMQ (Cooper, Russell, Skinner, & Windle, 1992), which assess social, enhancement and social drinking motives in adults, and added the conformity motives SPANISH DMQ-R 5 scale. The resulting questionnaire, the DMQ-R (Cooper, 1994) contained 20 items, five per scale, and shows adequate psychometric properties for assessing drinking motives in adolescents from North America (Cooper, 1994;Kuntsche et al., 2008) and Europe (Kuntsche, Knibbe, Gmel, & Engels, 2006a). Another modified version of the DMQ, which is also available in Spanish, is the Modified DMQ-R (Mezquita et al., 2011). This version subdivides coping motives into coping-with-anxiety and coping-withdepression (see Supplementary Material 1, SM1 for a detailed review of the DMQ versions). However, like the DMQ (Cooper et al., 1992), the Modified DMQ-R was developed for assessing drinking motives in adults. (2009) (Németh et al., 2011), as far as we know, no other study has explored their psychometric properties in Spanish adolescents.
There are only a handful of prospective studies on motives and alcohol use in adolescents, which have shown a less clear pattern of associations. They have found that social, enhancement, and coping motives predict alcohol use and misuse (Bradizza, Reifman, & Cooper, 1999;Cooper et al., 2008). However, one research work has indicated that when controlling for the correlations between drinking motives, only social motives prospectively predict alcohol outcomes (Schelleman-Offermans, Kuntsche, & Knibbe, 2011). The fact that there are only a small number of prospective studies, and that they have been carried out in North American and Swiss samples, are reasons for recommending additional prospective studies in adolescents in other sociocultural contexts.
As a result, the present study aims to adapt and study the validity and reliability of the Spanish DMQ-R in a sample of adolescents. Specifically, we study the structure validity of the DMQ-R. We hypothesised that a 4-factor solution of the DMQ-R would fit better our data than: a) one general model of drinking motives; b) a 2-factor model that compares positive and negative reinforcement (enhancement combined with social; coping combined with conformity); c) a 2-factor model that compares the internal vs.
the external source (social combined with conformity; enhancement combined with coping); and d) a 3-factor model with social and enhancement loading on a single factor  (Cooper, 1994). We also tested if a short form of the 4-factor model (DMQ-R SF), previously validated in European adolescents (Kuntsche & Kuntsche, 2009;Németh et al., 2011), better fitted our data than the original DMQ-R. Based on previous adaptation studies of the questionnaire in Europe (Kuntsche et al., 2006a;Kuntsche & Kuntsche, 2009), we also expected gender invariance for the four-factor structure.
We aimed to study the reliability of the scales. We expected them all to be .70, or higher, based on the standard Cronbach's alphas cut-off (Kaplan & Saccuzzo, 2009).
Finally, we aimed to study the concurrent and predictive validity of the scales of motives to predict alcohol outcomes. We predicted that social, enhancement and coping motives would relate to alcohol consumption, while coping and conformity motives would associate with alcohol-related problems (Cooper, 1004;Kuntsche et al., 2006a;Kuntsche & Kuntsche, 2009;Mazzardis et al., 2010). According to recent studies in young adults (Mezquita, Ibáñez, Moya, Villa, & Ortet, 2014;Studer et al., 2014), we expected enhancement and social motives to relate to weekend alcohol use, while coping motives to relate to drinking on weekdays (see SM2 to consult the hypothesised path analysis).

Procedure
In order to translate and adapt the DMQ-R (Cooper, 1994), various experts on drinking motives wrote the Spanish DMQ-R items from the original English scale. A back-translation was then carried out by a native English speaker and was compared to the original questionnaire. This process resulted in a Spanish version that was considered comparable to the original (see SM3).
Afterwards, we contacted the psychologists from six high schools in Castellón.
Parents gave their informed consent for their children to participate in the study. All the study procedures obtained the approval of the Ethics' Committee. During the assessment sessions, trained psychologists followed the standard instructions: handed out the scales, guaranteed confidentiality and encouraged participants to provide sincere answers. All the attending students voluntarily completed the questionnaires in the classroom and did not receive any compensation for their participation. One year later, we conducted the same assessment process. At Time 1 (T1), students answered the Spanish adaptation of the DMQ-R (Cooper, 1994), together with the drinking quantity questions. One year later (Time 2, T2), they completed the alcohol intake measures again and AUDIT (Babor, Higgins-Biddle, Saunders, & Monteiro, 2001).
Drinking quantities were assessed in two open-ended questions. Adolescents responded how many drinks (beers, glasses of wine, liquors and spirits) they drank on weekdays (Monday to Thursday) and at weekends (Friday to Sunday). Then we calculated the number of Standard Drink Units (SDU) after taking into account that beers and glasses of wine are the equivalent to one SDU, while liquors and spirits are the equivalent to two SDUs. In Spain, an SDU is the equivalent to 10 g of alcohol (Rodríguez-Martos, Gual, & Llopis, 1999).
We included the three items from the dependence symptoms scale (e.g., morning drinking) and the four items from the harmful alcohol use scale (e.g., alcohol-related

Missing data imputation
The missing values in the sample used to analyse the cross-sectional data (T1, N=504) were .23%, or lower, of the total data. Similarly, among those who participated in both time frames (N=238), we found only .21% of missing values. For these reasons, we followed a person mean imputation approach on each scale (Bentler, 2006). 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65 We conducted Confirmatory Factor Analyses (CFA) to compare which of the six models better fitted our data. We performed a path analysis to test the concurrent and predictive validity of each motive scale on alcohol outcomes with the factor solution that showed the best fit in the CFAs. We performed multi-group analyses to study the gender invariance of the structure, and the moderation effect of gender in the relation between motives on alcohol outcomes. These analyses were carried out with the EQS 6 software. In addition, with SPSS 21, we calculated Cronbach's alpha to test the internal consistency of the scales, and the means and SDs to explore the rank order at the mean endorsement levels of the four motives.

Data analyses
In all the SEM, we employed Satorra-Bentler's robust method, as our data was non-normally distributed (Satorra & Bentler, 2001). To consider that a model has an excellent fit, the χ 2 S-B must be non significant. However, this is uncommon in CFA. So the use of other fit indices to compare competing models is interesting: normed chi- In the path analysis, we considered the Lagrange multiplier (LM) and Wald tests suggestions to either include or remove the additional paths to those hypothesised in the model. We obtained the direct effects and indirect effects. Lastly, we tested if there were differences in the magnitude of the paths between motives and weekday SDUs and weekend SDUs by constraining the paths to be equal (Byrne, 2006;Bentler, 2006 relative goodness of fit between models was analysed via the scaled S-Bχ 2 difference test (Satorra & Bentler, 2001) using the "sbdiff" software.

Confirming the factor structure
The CFA results revealed that a 20-item 4-factor model of the DMQ-R showed acceptable fit indices. This factor solution significantly better fitted our data than the 1factor model (χ 2 S-B diff (6) = 97.80, p < .001); the two-factor models (positive vs.  Table 1). Factor loadings were high, except for items "to be sociable" and "because your friends pressure you to drink" (see Table 2). We also compared the 20-item 4factor model with the 12-item 4-factor model (short form, SF) and found that the second showed adequate fit indices and a significantly better fitted the data than the former (χ 2 S-B diff (118) = 264.37, p < .001). All the factor loadings of the 12-item solution on their corresponding factors were high (see Table 2).

Factorial invariance of the DMQ-R SF across gender groups
The fit indices of the multi-group analysis were adequate (see Table 1). The addition of cross-gender equivalence constraints for factor loading (χ 2 S-B diff (8) = 14.32, p > .05), variances (χ 2 S-B diff (4) = 4.17, p > .05) and covariances (χ 2 S-B diff (6) = 7.68, p > .05) did not result in a significant degradation in fit, which suggests that the structure of the DMQ-R SF was invariable between boys and girls. In both, long and short versions of the questionnaire, Cronbach's alphas were .70 or higher (DMQ-R: from 75 to .84; DMQ-R SF: from .70 to .85) (see Table 2). The highest endorsement of drinking motives was on the social motives scale, followed by enhancement, coping and conformity motives (see Table 2).

Path analysis
When we tested the hypothesised model (see SM2), the fit indices came close to the adjustment (see Table 1). However, when we included the path suggested by the LM test (social on weekday SDU T1), and when were removed the non-significant paths suggested by the Wald test (enhancement on weekend SDUs, coping on weekday SDUs, conformity on alcohol-related problems), the fit indices were excellent (see Table 1). Direct effects showed that social motives predicted weekend and weekday SDUs at T1, while coping motives predicted alcohol-related problems at T2 (see Figure   1).
When we constrained the path of social motives on weekday SDUs to be equal to the path of social motives on weekend SDUs, the differences in fit (χ 2 S-B diff (1) = 56.99, p < .001) indicated that social motives were significantly more strongly related to weekend SDUs than to weekday SDUs at T1.
The differences in fit between the multi-group analysis of the final path model (see Table 1), and the model that included the paths constrained to be equal between boys and girls, were not significant (χ 2 S-B diff (7) = 9.65, p > .05). This indicates that the model was invariant between groups.

Discussion
The present research aimed to adapt and study the psychometric properties of DMQ-R (Cooper, 1994) in Spanish adolescents. According to the hypothesis, the results showed that a 20-item 4-factor solution better fits our data than a 20-item 1-factor, a 20item 2-factor and a 20-item 3-factor solution, as found with the original DMQ-R (Cooper, 1994). However, even when the fit indices were considered acceptable and were similar to those found in previous studies (Kuntsche, et al., 2006a), some of the items showed low factor loadings (items 2 and 5). When we removed these items presented the same rank order at the mean endorsement levels of the four motives and patterns of interfactor correlations (e.g., higher correlations between socialenhancement and lower correlations between conformity-enhancement) than in previous studies conducted in adolescents from North America and Europe (Cooper, 1994;Kuntsche & Kuntsche, 2009;Kuntsche et al., 2008;Mazzardis et al., 2010;Németh et al., 2012), which confers validity to our results. In reference to the concurrent validity of the motive scales to predict alcohol use, social motives were the only ones that cross-sectionally and directly predicted alcohol drinking. Although they were related to weekday SDUs, the strongest relationships were found with weekend SDUs. However unlike our hypothesis, the specific associations between enhancement and weekend SDUs, and between coping and weekday SDUs, were not significant. This lack of significance is probably due to a confounding effect of the social motives. Since social and enhancement correlated highly and positively (see Figure 1 and SM4 for the Table of Correlations), the significant effect of the social motives on weekend SDUs may overshadow the influence of the enhancement motives (Studer et al., 2014). We can apply the same explanation to the non-significant association between coping and weekday drinking as the correlations between the coping and social motives were also high.
When we looked at the validity of motives to prospectively predict alcohol use, social motives were the only ones that indirectly predicted weekend SDUs, and the weekday SDUs, but to a lesser extent. Previous studies with Dutch adolescents obtained similar results, where social motives were the single predictor of total weekly consumption and also heavy episodic drinking at the 1-year follow-up (Schelleman-Offermans et al., 2011). Thus it seems that among European adolescents, social motives are the most relevant drinking motives to prospectively predict alcohol use.
In relation to the associations of motives with alcohol-related problems, no significant associations were found between conformity and alcohol-related problems.
The present study has several limitations. The first is the sample attrition between both assessments (52.78%). This is partly because some participants at T2 were of legal age to drop out of high school. High school psychologists also informed about a high level of mobility in that year. In relation to this, the attrition may lead to bias; for example, we found that the participants who dropped out of the study scored higher in alcohol use, motives and age than those who continued at T2. However, previous studies have shown that attrition has a minor impact on the associations between predictors and outcomes (Wolke et al., 2009). Secondly, we did not assess drinking motives in the second wave, so it was not possible to explore the test-retest reliability of the scales, and also the reciprocal influences between alcohol and motives.
Finally, we did not include measures of binge drinking or heavy drinking that are highly informative, especially in the case of enhancement motives (Cooper, 1994;Kuntsche et al., 2008).  Table   Table 2 Results of the CFAs, interfactor correlations, Cronbach's alphas, and means (SD)