New Estimates of Mortality Trajectories at Extreme Old Ages
Date: 09/27/2012 (Thu)
Time: 3:30pm- 5:00pm
Location: Seminar will be held on-site: Social Sciences 111
Organizer: Ken Land
Meeting Schedule: Login or email the organizer to schedule a meeting.
All meetings will be held in the same location as the seminar unless otherwise noted.
*** - All meetings (unless otherwise noted) will be in 221-B Social Sciences ***
10:00am - Jacob Moorad
10:30am - Seth Sanders
11:00am - OPEN
11:30am - Jake Fisher
12:00pm - Lunch, Jenny Tung (Faculty Commons)
1:00pm - Igor Akushevich
1:30pm - Konstantin Arbeev, Alexander Kulminski
2:00pm - DuPRI students
2:45pm - Giovanna Merli
3:15pm - preparation for seminar
3:30pm - Seminar Presentation (3:30pm to 5:00pm)
6:00pm - Eric Stallard, Anatoliy Yashin (dinner meeting)
Additional Comments: ABSTRACT: A growing number of persons living beyond age 85 underscore the need for accurate measurement and modeling of mortality at advanced ages. This is also very important issue for making correct forecasts of population aging and related demands for medical services and social support. Earlier studies indicate that exponential growth of mortality with age (Gompertz law) is followed by a period of mortality deceleration with slower rate of growth. This study challenges earlier conclusions with new data and estimates. In this study we used U.S. cohort survival data for people born in the same calendar year. For this purpose we obtained data from the U.S. Social Security Administration Death Master File to estimate hazard rates for 15 single-year extinct birth cohorts born in 1881-1895. We found that mortality deceleration is far less pronounced when it is measured for shorter monthly age intervals rather than for traditional annual intervals. To find out why does it happen we have made a simulation study and found that traditional measures of hazard rate (like the Nelson-Aalen hazard rate estimate) underestimate mortality force at extreme old ages (underestimation bias) when death rates are exceptionally high. We also found that mortality deceleration is far less pronounced when datasets with higher data quality (age reporting) are analyzed. Mortality modeling found that the Gompertz model demonstrates better goodness-of-fit in age interval 88-106 years compared to logistic (Kannisto) model when used for data of good quality. Study of mortality among other mammalian species (mice and rats) also found no mortality deceleration at advanced ages. It appears that the earlier reports of mortality deceleration for ages below 106 years may be a result of age exaggeration and the use of biased estimates of hazard rate.