A rank statistic for assessing the amount of variation explained by risk factors in epidemiologic studies

A statistic, Q, based on the ranks of the estimated probabilities of disease is proposed for assessing the effectiveness of regression models used with dichotomous dependent variables in epidemiologic studies of risk factors for chronic diseases. The pitfalls of R2 are discussed, and the proposed statistic is compared with R2 utilizing 8.6-year incidence data from the national cooperative Pooling Project and 15-year mortality data from the Chicago Peoples Gas Company Study. Based on the risk factors, systolic blood pressure, serum cholesterol, and smoking status for middle-aged males, the proposed statistic, Q, attains from 27-44% of its maximum value for the endpoint, the first major coronary event, and from 35-46% of the maximum for death from the cardiovascular diseases.