The aim of this study is to identify patterns of sleep difficulty in older women, to investigate whether sleep difficulty is an indicator for poorer survival, and to determine whether sleep difficulty modifies the association between disease and death. Data were from the Australian Longitudinal Study on Women's Health, a 15-year longitudinal cohort study, with 10 721 women aged 70–75 years at baseline. Repeated-measures latent class analysis identified four classes of persistent sleep difficulty: troubled sleepers (N = 2429, 22.7%); early wakers (N = 3083, 28.8%); trouble falling asleep (N = 1767, 16.5%); and untroubled sleepers (N = 3442, 32.1%). Sleep difficulty was an indicator for mortality. Compared with untroubled sleepers, hazard ratios and 95% confidence intervals for troubled sleepers, early wakers, and troubled falling asleep were 1.12 (1.03, 1.23), 0.81 (0.75, 0.91) and 0.89 (0.79, 1.00), respectively. Sleep difficulty may modify the prognosis of women with chronic diseases. Hazard ratios (and 95% confidence intervals) for having three or more diseases (compared with 0 diseases) were enhanced for untroubled sleepers, early wakers and trouble falling asleep [hazard ratio = 1.86 (1.55, 2.22), 1.91 (1.56, 2.35) and 1.98 (1.47, 2.66), respectively], and reduced for troubled sleepers [hazard ratio = 1.57 (1.24, 1.98)]. Sleep difficulty in older women is more complex than the presence or absence of sleep difficulty, and should be considered when assessing the risk of death associated with disease.