Retest reliability of self-reported function, self-care, and disease history

Exposures and outcomes frequently are measured by self-reports in epidemiologic studies. However, compared with objective data based on physiologic or laboratory tests, self-reports may suffer from lower accuracy and reliability. In addition, few reports examine reliability in population subgroups, such as black adults.

The authors examined the retest reliability of common self-reports concerning self-care, function, and chronic conditions in a random subsample of 92 middle-aged black subjects from a larger cohort of 998 subjects in St. Louis, Missouri. Subjects completed in-home interviews between September 2000 and July 2001.

Function and self-care measures included 7 basic activities of daily living, 8 instrumental activities of daily living, and 9 items based on the Nagi physical performance scale. Chronic conditions included a list of 11 common diseases and conditions and the 7 items of the Rose Angina protocol. Item level agreement was measured by kappa and scale level agreement was measured by intraclass correlation coefficients.

Function and self-care items demonstrated highly variable agreement with 7 items failing to reach even moderate (kappa = 0.40) levels of agreement. Scale reliability was better, and intraclass correlation coefficients ranged from 0.75 to 0.95. Self-reported chronic conditions all achieved at least moderate agreement, except for angina based on the Rose protocol.

This study provides evidence that summary measures of function and a number of chronic conditions and diseases are reliable based on self-reports from urban black adults.