Fig. 2
From: The healthcare costs of increased body mass index–evidence from The Trøndelag Health Study
Effect and 95% confidence intervals of BMI on healthcare costs using different analytical approaches, for males, with adjustment for age (top panel), and fully-adjusted models (bottom panel). In the fully adjusted analyses the OLS, GLM, 2PM, lagged IV-analyses, and the offspring IV-analyses, estimates were adjusted for age, education, smoking status, marital status, resident in rural or urban area, income and country of birth. In the offspring IV-analyses we also adjusted for offspring age and sex, and in the GRS based IV-analyses we only adjusted for age (since a prerequisite for these models is that there is no need for further adjustment because our genes are randomly assigned at conception, and therefore assumed to be uncorrelated with confounding factors)