INTRODUCTION

Smoking-related diseases have been the primary causes of premature death and disability in Malaysia for the last three decades^1,2. Annually, an estimated 20000 deaths are from smoking-related diseases², with approximately a third of the burden of diseases being related to smoking¹. This rate is expected to increase if the smoking prevalence among Malaysian adults persists².

It is known that smoking is a learned behavior that usually begins during adolescence^3-5. The National Health and Morbidity Survey 2015 in Malaysia found nearly 70% of smokers began smoking before the age of 18 years². The earlier adolescents start smoking, the more likely they will become smokers in adulthood⁶. Earlier initiation of smoking is also linked to higher risk of smoking-related diseases, such as cancer and cardiovascular diseases^6,7. Nevertheless, should they quit smoking, they can gradually regain their health⁸. However, relapse of nicotine addiction after stopping makes quitting difficult⁹. Usually, long-term smokers are not easily persuaded by the benefits of quitting smoking, especially when they have yet to be afflicted by smoking-related diseases^10,11. Thus, reducing the incidence of smoking initiation and increasing smoking cessation among youths is the only way to significantly reduce the prevalence of young smokers and address the health problems of smoking among the Malaysian population².

Behavioral modification models are frequently used successfully to reduce smoking initiation and improve smoking cessation among teenagers. One such model, the Transtheoretical Model (TTM)^12,13 assesses a person’s willingness to engage in a new and better behavior. It also includes methods for guiding the individual through the process of change. The TTM consists of five essential constructs: stages of change, the process of change, decisional balance, self-efficacy, and temptation, that describe the willingness to change and its advancement via a series of phases^12,13.

The struggle model, a fundamental process of making a decision connected with specific health behaviors^12,13, is reflected in the decisional balance construct. The perceived benefits (pros) and perceived barriers (cons) associated with smoking behavior are the decisional balance. Perceived benefits of smoking include helping in coping with stress and irritation and being a pleasurable activity. Therefore, it increases the attraction of smoking, despite the perceived barriers, including its health risks and being an irritation to others. Many studies have found that when a person progresses through the stages of behavioral change, the perceived benefits grow and the barriers decrease. The expectancy theory, which contends that a person’s relative course of action is influenced by the level of his expectation of rewards or failures, is also the foundation of the DBI¹⁴. The more critical the information considered before making a decision, the more successful the commitment to that decision and the more steadfast the adherence to that decision¹⁵. Two studies, one by Velicer et al.¹⁶ and another by Spencer et al.¹⁷ demonstrated excellent predictive ability of the DBI. Pallonen et al.¹⁸ developed a short version of the DBI for adolescents, which consisted of 12 items divided into three categories: smoking cons (six items), social pros (three items), and coping pros (three items). Each of these items is assessed on a 5-point Likert scale, with 1=least important and 5=most significant. The DBI has been validated in many countries, including the United States¹⁹, Bulgaria²⁰, and several Asian nations^21-23. However, a Malay language version of the DBI has not been developed and validated. Given the high incidence of smoking initiation among adolescents²⁴ and prevalence of smoking among adults in Malaysia²⁵, there is a pressing need for a suitable questionnaire on smoking behavior changes. This study therefore aims to establish the validity and reliability of this scale among Malaysian adolescents.

METHODS

We used the approach by Wild et al.²⁶ to validate the DBI in the Malay language, which consists of translation, followed by assessment of content and face validity, and finally of construct validity (Figure 1).

Figure 1

Schematic diagram of methodology to validate MDBI

Data management and analysis

The data were cleaned before any statistical analysis was carried out. Descriptive statistics was used to describe the characteristics of the respondents. We calculated item-level Content Validity Index (I-CVI), scale-level content validity index (SCVI), and kappa statistic for agreement to assess content validity based on the experts’ assessment of the questionnaire. Exploratory factor analysis was performed to determine the construct validity of MDBI. The ideal number of factors was determined using eigenvalue above 1 as criterion. Parallel analysis was used to determine the number of domains in MDBI (Figure 2) using Varimax rotation and factor loadings of ≥0.3 as the criterion for item inclusion. Kaiser Mayer-Olkin and Bartlett’s tests were used to assess adequacy of the data. Determination of the number of domains was by examination of the meeting point of the variance generated by SPSS and FACTOR.

Figure 2

Eigen value and parallel score analysis to determine the number of domains in MDBI

Following this, confirmatory factor analysis (CFA) was performed to validate the structure elucidated by EFA. Model fit was evaluated using multiple fit indices, namely relative chi-squared, goodness of fit index (GFI), comparative fit index (CFI), root mean square error of approximation (RMSEA), non-normed fit index (NNFI), and normed fit index (NFI). Construct validity and discriminant validity in CFA were measured by the average variance and the construct reliability²⁷. Reliability of the MDBI was assessed by examining the total correlation and the impact of removing each item. SPSS software was used to run the EFA and reliability analysis, FACTOR freeware (version 12.01.02; Lorenzo-Seva & Ferrando, 2006)²⁸ to determine the number of domains that needed to be extracted, and SPSS AMOS software to perform the CFA. All statistical analyses were performed at the 95% significance level.

RESULTS

Table 1 shows the content validity results for the second session after improvements were made to the questionnaire based on the comments of the content experts. The results showed that five items obtained a score of 3 or 4 from all content experts, while the remaining seven items obtained a score of 3 or 4 from only five of the experts. The CVI coefficients ranged from 0.30 to 1.00, while modified kappa varied from 0.816 to 1.00. The SCVI/UA was 0.4167, and the SCVI/Ave exceeded the set cut-off value of 0.90.

A total of 662 students responded, giving a high response rate of 98.2%. Of the 662 respondents, 3 out of 5 were male (n=389), and almost 70% (n=463) were from rural secondary schools. Most of the respondents were of Malay descent (86.3%, n=571), followed by Chinese (10.6%, n=70), and the remainder were of Indian and other ethnicities. Approximately three-quarters of the respondents were not current smokers (Table 2). Exploratory factor analysis based on eigenvalues above 1 and parallel analysis with FACTOR identified three domains. The total variance explained by the three domains was 65.57%. The first domain consisted of six items (accounting for 30.8% of the total variance), the second domain had three items (17.7%), while the third domain had three items (17.1%) (Table 3).

Figure 3 shows results of the CFA, which showed correlation coefficients between 0.51 and 0.81 between the items and the latent social pros domain and a comparable coefficient value in the coping pros domain. In the cons domain, two items had low correlation coefficients, but both were retained in the questionnaire at the suggestion of the content specialists. The RMSEA value was 0.061 (0.08), the relative chi-squared was 2.245 (5.00), while the CFI, ILI, and GFI values exceeded 0.90, indicating a good fit of the model. Table 4 shows that the average variance explained was >0.50 for the coping pros and social pros domains, but only 0.468 for the cons domain.

Figure 3

Schematic diagram of MDBI validation methodology

DISCUSSION

This study aimed to assess the psychometric features of the MDBI among secondary school students. The findings suggest that the translated DBI is a suitable and valid questionnaire for assessing smoking versus non-smoking decision-making among Malaysian school-aged teenagers. CVI and kappa for agreement demonstrated the MBDI content’s validity was at an acceptable level. Furthermore, as performed by Pallonen et al.¹⁸, EFA and parallel analysis showed that MDBI is divided into three domains. In addition, the communalities in each domain are comparable to one another. The first version’s explained variation was greater than the prescribed 50% level. The results of the CFA corroborated the conclusions of EFA and parallel analysis. In addition to the domains’ reliability, which ranged from 0.753 to 0.857, the item-to-subtotal correlation coefficients for the subscales were greater than 0.45, indicating a satisfactory level of reliability.

The results also showed that the questionnaire could distinguish between different domains in the MBDI and measure various aspects of the disadvantages of smoking (e.g. concerns for smoking effects on health). However, the results contradicted the findings of Chen et al.²³, who found only two domains in the Chinese version of the DBI for Taiwanese adolescents, and Hoeppner et al.¹⁹, who discovered four factors (two positive and two negative) among African American adolescents. This discrepancy could be due to the different characteristics of the respondents in this study. For example, the study by Hoeppner et al.¹⁹ involved only females, versus the present research which consists of both males and females in a 3:2 ratio.

The three domains accounted for 65.4% of the variance. This is higher than the 55.4% previously report by Khazaee-Pool et al.²² and Velicer et al.¹⁶. In the Velicer et al.¹⁶ analysis, the solution of two components (i.e. advantages and disadvantages) accounted for 41% of the observed variance, which is higher than in the study of Pallonen et al.¹⁸ where three-factor solutions (social benefits, overcoming advantages, and drawbacks) accounted for half of the variance. Hoeppner et al.¹⁹ found a four-factor solution (two pro factors and two cons factors) that explained 45% of the variance, which was lower than in Chen et al.²³ (74.5%). This could be explained by more items in the two DBI domains found in the results of the study among adolescents in Taiwan.

CFA analysis of the MDBI revealed a positive link between social and coping pros, with a correlation of 0.61 in this study. This is consistent with other studies using the same instrument. The Plummer et al.²⁹ correlation coefficient was virtually the same at 0.59. The poor correlation between social coping and the cons of smoking is also consistent with earlier research.

The Cronbach alpha coefficients for the smoking cons scale, the social pros scale, and the coping pros scale were 0.867, 0.754, and 0.753, respectively (Table 5). These reliability estimates are similar to the coefficients obtained in the study for these same scales, which were 0.80 for smoking, 0.787 for social pros, and 0.832 for coping pros by Khazaee-Pool et al.²². These reliability estimates are consistent with the alpha obtained in the Plummer et al.²⁹ study for the same scales, in which the alphas for social pros, coping pros and cons were 0.68, 0.79 and 0.86, respectively.

CONCLUSIONS

Overall, the findings suggest that the Malay version of the DBI is a reliable and valid scale for assessing adolescents’ consideration of whether to smoke or not. The MDBI should be further tested on adolescents in various sociodemographic and geographical settings to confirm its applicability in the general Malaysian adolescent population.