Calculating significance of differences in death rates
Share with others:
The Standardized Mortality Ratio represents a comparison of actual deaths of a smaller group to the expected death rate of a larger group. In the data below, the SMR represents actual deaths from particular diseases in each county as compared with expected national rates for the same category or categories of disease.
Figuring whether these increases over the national average are statistically significant requires calculation of confidence intervals -- a statistical determination of how reliable the number is.
In the confidence intervals below, the rates show the actual SMR, calculated as a ratio of actual death rates divided by expected death rates. In the parenthesis is the statistical range in which the rate would fall 95 percent of the time. The number is considered significantly higher than national rates if the confidence-interval range does not include the whole number 1.0. If the range does include 1.0, then it is possible there is no difference in death rates.
One example is Beaver County's lung cancer ratio of 1.03, which might suggest a rate 3 percent higher than the national rate. However the confidence interval range of 0.97-1.09 includes 1.0. That means, despite the appearance of a high rate, statistically it could equal the national average.
The formula used for calculating the confidence interval is as follows:
95% Confidence Interval = (Actual deaths/expected deaths) +/- 1.96 x [(square root of actual deaths)/expected deaths].
Below, the SMR for the disease category in each county is listed along with the confidence interval for age-adjusted county death rates, 2000-2008.
They reveal that all 14 counties have heart-disease mortality rates exceeding the national average; 12 of 14 have respiratory disease mortality rates exceeding the national average; three of 14 have lung-cancer mortality rates exceeding the national average; while 13 of 14 have a combined mortality rate for all three diseases in excess of combined national expected rates for the three.
First Published December 12, 2010 12:00 am
