In The Netherlands, as in many other countries, important geographical variation in mortality from conditions amenable to medical intervention exists. Associations with a number of simple medical care supply characteristics (general practitioner density, hospital bed density, and percentage of regional hospital beds located in university and small hospitals) are generally weak and inconsistent, both before and after controlling for possible confounding factors. We explored one of the possible reasons for this lack of consistency, which is the time dependency of the relationship between medical care supply and avoidable mortality. A comparison of associations in four time periods (1950-54, 1960-64, 1970-74 and 1980-84) shows that the percentage of variance in regional mortality levels which can be "explained" by the medical care supply variables has changed over time. Although the patterns of change differ little from what one would expect on the basis of the time of introduction of medical care innovations, the exact nature of the associations is puzzling. Apart from some expected negative associations between mortality and the presence of university hospitals, we also found a few unexpected positive associations with general practitioner density. Possible explanations for these findings are discussed, and it is concluded that further study is necessary to reveal the causes of a higher or lower mortality level for conditions considered to be amenable to medical intervention.