The data consist of 213 pairs of subjects鈥攁ged 90 years and older鈥攁nd proxies. The relationship between SRH and PRH was analyzed by chi-square test and Cohen鈥檚 kappa. Logistic regression analysis was used to find out the factors that are associated with the congruence between health ratings. The association between SRH and PRH with mortality was studied using Cox proportional hazard models.
The subjects rated their health more negatively than the proxies. Kappa value indicated only slight congruence between SRH and PRH, and they also predicted mortality differently. Good self-reported functional ability was associated with congruence between SRH and PRH.
The results imply that the evaluation processes of SRH and PRH differ, and the measures are not directly interchangeable. Both measures are useful health indicators in very old age but SRH cannot be replaced by PRH in analyses.