Background
The life expectancy of the average American with diabetes has increased, but the quality of health and functioning during those extra years are unknown. We aimed to investigate the net effect of recent trends in diabetes incidence, disability, and mortality on the average age of disability onset and the number of healthy and disabled years lived by adults with and without diabetes in the USA. We assessed whether disability expanded or was compressed in the population with diabetes and compared the findings with those for the population without diabetes in two consecutive US birth cohorts aged 50–70 years.

Methods
In this prospective longitudinal analysis, we analysed data for two cohorts of US adults aged 50–70 years from the Health and Retirement Study, including 1367 people with diabetes and 11 414 without diabetes. We assessed incident disability, remission from disability, and mortality between population-based cohort 1 (born 1931–41, follow-up 1992–2002) and cohort 2 (born 1942–47, follow up 2002–12). Disability was defined by mobility loss, difficulty with one or more instrumental activities of daily living, and difficulty with one or more activities of daily living. We entered age-specific probabilities representing the two birth cohorts into a five-state Markov model to estimate the number of years of disabled and disability-free life and life-years lost by age 70 years.

Findings
In people with diabetes, compared with cohort 1 (n=1067), cohort 2 (n=300) had more disability-free and total years of life, later onset of disability, and fewer disabled years. Simulations of the Markov models suggest that in men with diabetes aged 50 years, this difference between cohorts amounted to a 0·8–2·3 year delay in disability across the three metrics (mobility, 63·0 [95% CI 62·3–63·6] to 64·8 [63·6–65·7], p=0·01; instrumental activities of daily living, 63·5 [63·0–64·0] to 64·3 [63·0–65·3], p=0·24; activities of daily living, 62·7 [62·1–63·3] to 65·0 [63·5–65·9], p<0·0001) and 1·3 fewer life-years lost (ie, fewer remaining life-years up to age 70 years; from 2·8 [2·5–3·2] to 1·5 [1·3–1·9]; p<0·0001 for all three measures of disability). Among women with diabetes aged 50 years, this difference between cohorts amounted to a 1·1–2·3 year delay in disability across the three metrics (mobility, 61·3 [95% CI 60·5–62·1] to 63·2 [61·5–64·5], p=0·0416; instrumental activities of daily living, 63·0 [62·4–63·7] to 64·1 [62·7–65·2], p=0·16; activities of daily living, 62·3 [61·6–63·0] to 64·6 [63·1–65·6], p<0·0001) and 0·8 fewer life-years lost by age 70 years (1·9 [1·7–2·2] to 1·1 [0·9–1·5]; p<0·0001 for all three measures of disability). Parallel improvements were gained between cohorts of adults without diabetes (cohort 1, n=8687; cohort 2, n=2727); within both cohorts, those without diabetes had significantly more disability-free years than those with diabetes (p<0·0001 for all comparisons).

Interpretation
Irrespective of diabetes status, US adults saw a compression of disability and gains in disability-free life-years. The decrease in disability onset due to primary prevention of diabetes could play an important part in achieving longer disability-free life-years.