The study applies household production theory to investigate how socioeconomic characteristics and behavioural choices affect health status of an individual (here related to cardiovascular diseases). Structural equation modelling (SEM) is used as an estimation approach, which allows to model health inputs as endogenous variables, to test the hypothesized linkages among them (e.g., between smoking and weight) and to investigate their contribution to a production of a particular health state. Moreover, direct effects of exogenous personal characteristics on health can be disentangled from their indirect impact via lifestyle choices. To improve its measurement properties, health status is modelled via SEM as a latent variable. The results indicate a multidimensionality of cardiovascular health, which is more adequately represented by a two-factor latent variable model rather than by a single-factor model. Moreover, the specified two health dimensions (hypertension and lipids) are (partly) differently explained by the model variables. While higher weight negatively and significantly impacts both health dimensions, a diet rich in F&V (Fruit & Vegetable) has stronger contribution to a lower hypertension risk. At the same time, higher educational level showed to be rather related to lower measurements of blood fat. A special focus in the model is given to the (in)direct impact of education and income. Both variables showed to be related to less frequent smoking and engaging in sedentary leisure time activities. However, their impact on diet was rather indirect through their positive contribution to nutrition knowledge.