Lack of stress-health relationships: a base rate problem?

The lack of relationship between psychological or social stressors and various health indices is common. This paper presents stress and health data collected from nurses at 10 different hospitals and from food processors at eight different sites. Correlations computed in samples constructed so that the number of people affected by a particular illness is approximately equal to those not so affected were consistently higher than similar correlations computed for the entire sample of 1,934 respondents. Because correlations are affected by the base rate of the phenomenon one is studying, it is suggested that other indices such as simple percentages in affected and unaffected groups are more appropriate.