Human mortality at extreme ages: Data from the NLTCS and linked medicare records

An analysis using the 1982-1999 National Long-term Care Survey (NLTCS) linked to Medicare vital statistics data 1982-2003 focused on deaths at ages 85+ where deviations from the Gompertz mortality function are often observed. To model the complex mortality patterns observed at late age, standard mortality functions were generalized with i) a location parameter, ii) a fixed heterogeneity distribution, and iii) parameters expressing heterogeneity in the individuals' rate of aging. The data are consistent with not only an upper bound to mortality rates but also declines in the yearly hazard rates at ages 100+. To determine if these patterns could be due to errors in age reporting, effects of plausible age misreporting patterns were simulated. Biases due to age misreporting at ages 95 to 115 did not materially change this pattern.

Duke Scholars

Author Kenneth G. Manton Trinity College of Arts & Sciences

Author Igor Akushevich Social Science Research Institute

Author Alexander Kulminski Social Science Research Institute