Indicators of "avoidable" mortality in health administrative areas in Sweden 1974-1985.

When comparing health administrative areas in Sweden the variation in death rates for 13 suggested indicators of avoidable mortality was analysed for the time periods 1974-79 and 1980-85. For most indicators the variation was significant. The level of systematic (nonrandom) variance differed between the sexes for certain conditions. According to the theory of mass-significance it would be appropriate to use a sharpened significance test of high SMRs. For certain health policy indicators it was possible to detect areas accounting for nearly all the deviation above the national standard using a sharpened test (p < 0.001). For most medical care indicators, however, the high death rates had a lower level of significance. When sensitivity is given priority an ordinary test (p < 0.05) should be preferred. This means, however, dealing with potential false positive warning signals. Epidemiological surveillance of indicators of avoidable mortality should be the starting point for in-depth studies.