Widowhood and mortality risk in a community sample of the elderly: a prospective study.

Mortality risk during early bereavement was examined in a community sample of 1046 married elderly persons 65 years and over, followed from 1982 to 1988. Cox' regression models with time-dependent covariates were computed to estimate mortality risk, while controlling for pre-widowhood sociodemographic and health-related variables. Elderly young-old (65-74) and old-old men (> or = 75) showed slightly elevated age-adjusted relative risks (RR) during the first 6 months of widowhood (RR = 1.69; 95% CI: 0.86-3.31 and RR = 1.79; 95% CI: 0.44-7.28 respectively). These RRs increased slightly after adjustment for pre-widowhood control variables. The age-adjusted RR during early widowhood for young-old women was 2.87 (% CI: 0.81-2.42), which increased to 3.86 (95% CI: 1.11-13.45) after adjustment for sociodemographic and health-related variables. This analysis stresses the usefulness of Cox' regression models with time-dependent covariates to calculate mortality risk for variable periods after onset of widowhood adjusted for pre-widowhood characteristics. However, the power of the study was limited, resulting in mostly insignificant risk estimates and wide confidence intervals.