Abstract

Some clinicians and some epidemiologists appear to be under the illusion that techniques available for the diagnosis of internal diseases such as lung cancer have improved so much that autopsies are not necessary on the bodies of most people who die. Partly for this reason, partly for economic reasons and partly because clinicians fear litigation if autopsy shows that they treated patients for the wrong disease, autopsy rates have been falling in most developed countries. The object of the present review was to ascertain how much reliance can reasonably be put on clinical diagnoses made and death certificates completed in the absence of autopsy data. In the case of lung cancer, high rates of false positive and false negative diagnoses are universally prevalent, with biases influencing these rates, so that smokers are more likely to be appropriately investigated for lung cancer, and false negatives are commoner in nonsmokers. All investigators who have compared clinicalbased and autopsy-based death certificates have concluded that higher autopsy rates are necessary and the results of one study suggest that a high autopsy rate in a hospital leads to improvements in the accuracy of clinical diagnoses. The extent to which diagnoses on death certificates that are dependent solely on clinical data are seriously inaccurate for internal diseases such as lung cancer should engender caution in all who use mortality data to guide public health policies and to identify and quantify environmental risks to healths.