Measuring horizontal inequity in Belgian health care using a Gaussian random effects two part count data model.

We estimate the determinants of utilisation of physician and hospital services in Belgium using a one- and two-part panel count data model, and a one- and two-part pooled count data model. We conclude that the two-part panel count data model is most appropriate as it controls for unobserved heterogeneity and allows for a two-part decision-making process. The estimates of the determinants of utilisation of health care are then used to calculate indices of horizontal inequity. We find that inequity for general practitioner and hospital services is stable across time and in favour of low-income individuals, in the sense that, overall, they consume more than one would expect on the basis of their need, albeit the indices for hospital care are not significant. Horizontal equity applies to specialist care in all years, but from 1999 onwards, some evidence (although not statistically significant) of pro-rich inequity is found. Copyright 2004 John Wiley & Sons, Ltd.