The relationship between smoking status and monthly medical expenditure in a Japanese population; a gamma regression approach
*Yoshitaka Murakami, Shiga University of Medical Science
Keywords: gamma regression, skewed distribution, monthly medical expenditure, smoking, Japan
Purpose: It is well known that smoking affects long-term health outcomes concerning disability, mortality and costs. Whether smoking affects short-term outcomes, such as monthly medical costs, is less obvious. We examined the relationship between smoking status and monthly medical expenditure in a Japanese population. Method: The study was conducted in Shiga Prefecture from April 2000 to March 2006. A comprehensive dataset was used, including individualized medical expenditure and related health measures. We analyzed participants aged 40 to 80 years (2,393,424 person-months) to investigate the relationship between smoking status at baseline, and 72 months’ worth of medical expenditure that followed. Smoking status was categorized as ‘never’, ‘quit’ and ‘current’. A gamma regression model was used to account for skewed distributions of cost data, and a generalized estimation equation method was applied to deal with autocorrelations between monthly costs. Finally, a sex-specific analysis was conducted to adjust for systolic blood pressure, diabetes, and age during the observation period. Results: The cost ratios (95% confidence intervals) of total medical expenditure against smoking status, compared with ‘never smokers’, was high in women: quit: 1.36 (1.05–1.77), current: 1.20 (1.11–1.29), but not in men (quit: 0.95 (0.89–1.02), current: 0.99 (0.94–1.04). These patterns were similar for the inpatient costs: women; quit: 1.36 (1.05–1.76), current: 1.07 (0.98–1.16), men; quit: 0.99 (0.92–1.06), current: 1.02 (0.97–1.06) and outpatient costs: men; quit: 1.00 (0.91–1.08), current: 0.96 (0.89–1.02), women; quit: 1.09 (0.94–1.26), current: 1.08 (1.01–1.16). Discussion: Our study found that current female smokers, and female smokers who had quit, were associated with higher monthly costs than females who had never smoked. The advantage of our study is that we used a novel statistical approach to handle the short-term variations and cost time trends. Further investigation is needed (especially for men) because the health profile of current smokers, and those who have never smoked, appears to be similar to that of the older Japanese population.
Important Dates & Deadlines
- October 9 - 11, 2013