Socioeconomic Status and Health Behavior in Nepal

Thesis (Master's)--University of Washington, 2016-07

Background: Socioeconomic status (SES) is an important determinant of health outcomes, but more research is needed to improve our understanding of the relationship between SES, behavior and health, particularly in low- and middle-income countries (LMICs). Efforts to improve our understanding are limited by the difficulty of measuring economic status in low-resource settings where income may be seasonal and informal employment is common. The debate over how to accurately measure economic wellbeing is pertinent in Nepal, where more than 25% of the population lives in poverty and subsistence farming is an important source of livelihood. Additionally, studies suggest a growing burden of noncommunicable disease (NCDs) in Nepal, many of which are impacted by behavioral risk factors. Objectives: This study aimed to construct an asset-based wealth index to estimate household wealth; and to develop and test prediction models for tobacco use and physical activity in a community-based sample of adults in Dhulikhel, central Nepal. Methods: We conducted a cross-sectional study using data from 863 adult participants of the Dhulikhel Heart Study in Dhulikhel, Nepal. Household characteristics, individual sociodemographic characteristics and individual health behaviors were assessed using standard questionnaires completed during in-home interviews. The Demographic and Health Surveys (DHS) wealth index model was used to construct an asset-based measure of household wealth. The wealth index used information collected in the DHS, including access to utilities and infrastructure (e.g. source of drinking water), durable asset ownership, and housing characteristics (e.g. number of rooms for sleeping), to produce a measure household wealth. The wealth index was constructed using principal components analysis (PCA) of these measures. Backwards stepwise logistic regression was used to develop and test prediction models for tobacco use and physical activity using the developed wealth index and other SES variables. Tobacco use was categorized as ever (lifetime) or never. Participants were categorized as those who met the WHO guidelines for recommended level of physical activity (600 MET-minutes per week) and those who did not. Area under the Receiver Operating Characteristic curve (AUC) was used to assess the performance of the predictive models; an average AUC of 0.70 was considered acceptable. Results: Of 863 participants included in this study, 59% were female. The average age was 40.6 years and nearly a quarter of participants were in the highest quintile of household wealth. On average, study participants had 6.7 years of formal education; approximately one-third of the study population had no formal education. The first component of the PCA, designated as the wealth index, found that households with the following characteristics had higher loadings: use of liquid petroleum gas (LPG) as fuel for cooking; had a toilet that flushed to piped sewerage system; had drinking water piped into the dwelling; owned a TV; owned a nonmobile phone; owned a refrigerator; owned a table; owned a chair; owned a sofa; owned a cupboard; owned a computer; owned a clock; owned a fan; owned a bike or rickshaw; owned a motorcycle or scooter; had internet; had a bank account; had cement floors; and had a cement roof. The wealth index accounted for 17% of the variability across all wealth indicators. Approximately 32% of participants reported lifetime tobacco use, and 40% of participants did not meet the recommendation for physical activity. After adjustment for sociodemographic characteristics, males were found to have significantly higher odds of tobacco use (OR=6.22, 95%C CI: 3.7-10.45, p<0.001) compared to females. No significant differences in physical activity were seen by sex. The prediction model for tobacco use included sex, age, and education; wealth was not a significant covariate in this model. The average AUC associated with the performance of the model was 0.829. The prediction model for physical activity included age, education, wealth, ethnicity and work status in the past twelve months. The average AUC associated with the performance of the model was 0.649 which is below the cut-off of 0.70 traditionally used for evaluating such models.