Diabetes, employment and behavioural risk factors in China : marginal structural models versus fixed effects models

A diabetes diagnosis can motivate its recipients to reduce their health risks by changing lifestyles but can adversely affect their economic activity. We investigate the effect of a diabetes diagnosis on employment status and behavioural risk-factors taking into account their potentially intertwined relationships. Longitudinal data from the China Health and Nutrition Survey covering the years 1997 to 2011 are used to estimate the effect of a diabetes diagnosis on employment probabilities, alcohol consumption, smoking cessation, body mass index, physical activity and hypertension. To deal with potential confounding, two complementary statistical techniques - marginal structural and fixed effects models - are applied. The marginal structural and fixed effects models generate similar results despite their different underlying assumptions. Both strategies find patterns distinct for males and females, suggesting a decrease in employment probabilities after the diagnosis for women but not for men. Further, few improvements and even further deterioration of behavioural risk factors are found for women, while for men these risk factors either improve or remain the same. These results suggest differences in the impact of diabetes between sexes in China and highlight the potential of reducing behavioural risk factors for women to narrow these inequities.