BACKGROUND: Variations in 'avoidable' mortality may reflect variations in the quality of care, but they may also be due to variations in incidence or severity of diseases. We studied the association between regional variations in 'avoidable' mortality and variations in disease incidence. For a selection of conditions we also analysed whether the proportion of in-hospital deaths can explain the regional variations in incidence-adjusted mortality. METHODS: Relative risks for mortality, incidence, incidence-adjusted mortality and in-hospital mortality (1984-1994) were calculated by log-linear regression. Linear regression was used to examine the relationship between mortality and incidence on the one hand, and between incidence-adjusted mortality and in-hospital mortality on the other. RESULTS: Significant regional mortality variations were found for cervical cancer, cancer of the testis, hypertensive and cerebrovascular disease, influenza/pneumonia, cholecystitis/lithiasis, perinatal causes and congenital cardiovascular anomalies. Regional mortality differences in general were only partly accounted for by incidence variations. The only exception was cervical cancer, which no longer showed significant variations after adjustment for incidence. The contribution of inhospital mortality variations to total cause-specific mortality variations varied between conditions: the highest percentage of explained variance was found for mortality from CVA (60.1%) and appendicitis (29.2%). CONCLUSIONS: Incidence data are a worthy addition to studies on 'avoidable' mortality. It is to be expected that the incidence-adjusted mortality rates are more sensitive for quality-of-care variations than the 'crude' mortality variations. Nevertheless, further research at the individual level is needed to identify possible deficiencies in health care delivery.
BACKGROUND: Variations in 'avoidable' mortality may reflect variations in the quality of care, but they may also be due to variations in incidence or severity of diseases. We studied the association between regional variations in 'avoidable' mortality and variations in disease incidence. For a selection of conditions we also analysed whether the proportion of in-hospital deaths can explain the regional variations in incidence-adjusted mortality. METHODS: Relative risks for mortality, incidence, incidence-adjusted mortality and in-hospital mortality (1984-1994) were calculated by log-linear regression. Linear regression was used to examine the relationship between mortality and incidence on the one hand, and between incidence-adjusted mortality and in-hospital mortality on the other. RESULTS: Significant regional mortality variations were found for cervical cancer, cancer of the testis, hypertensive and cerebrovascular disease, influenza/pneumonia, cholecystitis/lithiasis, perinatal causes and congenital cardiovascular anomalies. Regional mortality differences in general were only partly accounted for by incidence variations. The only exception was cervical cancer, which no longer showed significant variations after adjustment for incidence. The contribution of inhospital mortality variations to total cause-specific mortality variations varied between conditions: the highest percentage of explained variance was found for mortality from CVA (60.1%) and appendicitis (29.2%). CONCLUSIONS: Incidence data are a worthy addition to studies on 'avoidable' mortality. It is to be expected that the incidence-adjusted mortality rates are more sensitive for quality-of-care variations than the 'crude' mortality variations. Nevertheless, further research at the individual level is needed to identify possible deficiencies in health care delivery.