Back to browse results
Epidemiological Modelling of Tobacco Use to Find Its Correlates among Adolescents in India: Comparative Appraisal of Conventional and Hierarchical Models
Authors: Ashish Datt Upadhyay, Sada Nand Dwivedi, Vishnubhatla Sreenivas, and Anju Dhawan
Source: Demography India, 47(1): 61-70
Topic(s): Tobacco use
Country: Asia
Published: JAN 2018
Abstract: Tobacco use is the known risk factor of some of the major causes of death like cancer, respiratory diseases and heart diseases. In case of hierarchical structure of data, to obtain more accurate analytical results, multilevel regression model is often preferred in epidemiological and public health research. Aim of this study was to find factors associated with tobacco use among adolescents and appraisal of comparative results under conventional and multilevel logistic models. For this, available data on 37033 adolescents (age 15–19) years under National Family Health Survey-3 (2005–2006) were used. The performance of the models was assessed by the log likelihood values, area under the ROC curve, AIC and BIC. In this study, socio-demographic variables which are likely to be positively associated with the chance of tobacco use among adolescents were low education, employment, caste and religion other than Non-SC/ST, low wealth index, advancing age and male gender. Adolescents who belong to tobacco using household were more likely to use tobacco than their counterparts. Among the state level variables, adolescents belonging to states where prevalence of education of 10th standard and above was below 30.8%, or prevalence of tobacco users was above 34.6%, were more likely to use tobacco than belonging to otherwise. This study revealed that low education of adolescents and higher prevalence of tobacco use in the state are important factors to encourage adolescents regarding tobacco use. On other hand, although associated factors under multilevel model also remained similar, the model performance parameters showed efficacy of multilevel model in comparison to conventional logistic model. Introduction