The purpose of the study is to investigate dynamics across provinces and time causing Spanish pharmaceutical expenditure to increase. Adjustment for residual correlation across time periods and residual variance instability is shown to be essential. Presence of parametric instability is captured using linear time trends in the coefficients. Control for endogenous spatial spillover across provinces is proved to be highly important. Finally, the results are shown to be robust toward inclusion of control for fixed autonomous community effects. The study adds to previous knowledge by demonstrating the fallacy of simpler traditional approaches and stresses the necessity of not only adjusting for each of these features, but integrating them into a unified framework when analysing small-area health care data.