November 30, 2021 | News | Congratulations
Marie Böhnstedt Obtained Her Doctorate
On November 30, 2021, Marie Böhnstedt from the Joint Research Laboratory of Statistical Demography successfully defended her doctoral thesis on “Statistical Methods for Frailty Models: Studies on Old-Age Mortality and Recurrent Events” at Leiden University Medical Center (LUMC).
In her dissertation Marie Böhnstedt presents statistical methodology for two frailty models used for studying old-age mortality and recurrent events, respectively.
The risks of experiencing events like repeated infections or death naturally differ between individuals. While some of these differences can be explained by observable characteristics, such as gender or health markers, frailty models also allow for unobserved heterogeneity between individuals.
The first part of the thesis studies the slowing down of human death rates at advanced ages, known as mortality deceleration, in a frailty model. Here, the apparent deceleration results from selection, as survivors to increasingly higher ages tend to be individuals with lower and lower mortality risks. However, the statistical analysis of this phenomenon is complicated by the scarcity of observations at the oldest ages. The thesis proposes a new approach for model selection which outperforms existing techniques in detecting mortality deceleration. It also investigates how data limitations – for example, if only information on survivors beyond some high age is available – affect the performance of the statistical methods.
The second part of the thesis focuses on a frailty model for studying recurrent events in the presence of the competing event death. The model allows that an individual’s recurrence process and mortality may be interdependent. For instance, a higher risk of repeated infections may be accompanied by a higher risk of death. This, again, implies that individuals who survive longer are a selected group whose infection risks differ from those who die earlier. The thesis extends existing methodology to two common situations of incomplete observations. The first situation arises if it is only known how often an individual experienced the recurrent event in given time intervals, but not when the events occurred exactly. In the second situation, individuals enter the sample only some time after becoming at risk of the events, a typical setting in studies of disease events and mortality in an elderly population.
The dissertation was supervised by Jutta Gampe (MPIDR) and Hein Putter, Leiden University Medical Center (LUMC).