gms | German Medical Science

MAINZ//2011: 56. GMDS-Jahrestagung und 6. DGEpi-Jahrestagung

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V.
Deutsche Gesellschaft für Epidemiologie e. V.

26. - 29.09.2011 in Mainz

Comparison of methods for the estimation of prevalence in a dynamic cohort

Meeting Abstract

Search Medline for

  • Tania Schink - BIPS – Universität Bremen, Bremen
  • Edeltraut Garbe - BIPS – Universität Bremen, Bremen

Mainz//2011. 56. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds), 6. Jahrestagung der Deutschen Gesellschaft für Epidemiologie (DGEpi). Mainz, 26.-29.09.2011. Düsseldorf: German Medical Science GMS Publishing House; 2011. Doc11gmds244

doi: 10.3205/11gmds244, urn:nbn:de:0183-11gmds2441

Published: September 20, 2011

© 2011 Schink et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by-nc-nd/3.0/deed.en). You are free: to Share – to copy, distribute and transmit the work, provided the original author and source are credited.


Outline

Text

Introduction: Prevalence of drug use as well as prevalence of morbidities are frequently used measures in drug utilization and burden of disease studies. They are also important for the assessment of the public health impact of observed drug risks. Calculation of prevalence is straightforward in cross-sectional studies. However, estimation of prevalence is not as simple in the dynamic cohorts of claims databases where morbidity is only indicated by a claim made at a specific date.

Methods: Several methods for estimation of prevalence in dynamic cohorts will be presented and their (implicit) assumptions will be discussed and compared:

1. Point prevalence at a fixed date, i.e. number of drug users at a certain date divided by the number of cohort members at this date.

2. Period prevalence 1, i.e. numbers of drug users during a fixed period divided by the number of cohort members which were insured during this whole period or died during this period.

3. Period prevalence 2, i.e. the number of members with a certain condition divided by the number of cohort members at this date where the numerator is determined not only by looking at a certain date but by using a fixed time period before this date.

4. Person time approach, i.e. number of drug users in a certain period divided by the person time accrued in this period.

The application of the methods will be demonstrated using data of the German Pharmacoepidemiological Research Database based on claims data from over 14 million insurants. Prevalences for several scenarios (e.g. the prevalence of a short chronic co-morbidity, prevalence of an acute short-term comorbidity, prevalenve of drug use for a chronic disease, prevalence of drug use for a short-term disease) will be calculated and compared.

Results: All four methods make (implicit) assumptions on the length of “exposure” and the dynamics of the cohort. This results in differences both in the numerator and the denominator of the prevalence estimator and hence also in - sometimes considerable - differences in the estimated prevalence. The (implicit) assumptions additionally lead to differences in characteristics of cohort members taken into account for prevalence estimation.

Discussion: There is no single best method for the estimation of prevalence in a dynamic cohort. The method to use has to be chosen based on the characteristics of the “exposure” (e.g. short-term or long-term) and the dynamics of the cohort, especially the dynamics of the exposed cohort.