ABSTRACT
New Statistical Methods for Estimation of Prevalence Incidence and
Mortality Based on Pharmacoepidemiological and Other Health-related Databases
Henrik Støvring,
National Centre for Register-Based Research, Aarhus University
When measuring the use of health care services in a population, the basic measures are prevalence and incidence.
Traditionally these measures have been costly to obtain, since they require information on
individual treatment status, either for the population or a sample of it. With the advent of population-based automated event registries, it
has now become possible to obtain information on individual event histories at a very small cost. Typically these databases do however
not contain direct information on individual treatment or disease status, as they only record
the time of certain events related to treatment. Traditionally this problem is circumvented by use of ad hoc
decision rules for determining treatment status.
Inspired by the waiting time approach conceived by Hallas et al., we have developed a maximum likelihood based
method. The method is based on the first events in a time window for each individual and avoids the ad hoc decision
rules. In the presentation I will demonstrate how the developed method alllows estimation of incidence, prevalence,
and mortality among treated. The method is applied to data on use of antidiabetic medications and
compared to traditional methods. In the presentation I will briefly touch on simulation results and diagnostic procedures.
Return to
2002-2003 Schedule of Presentations |
Return
to Home Page
|
