Abstract

Classification trees and its application to cervix cancer screening in the Belgian Health Interview Survey 1997

Hens, N., Bruckers, L., Arbyn, M., Aerts, M., and Molenberghs, G.

Objectives: To outline an evidence-based health policy, one is often interested in the profiles of persons who are at risk to obtain certain diseases or who do not respond to prevention programs as e.g. cervix cancer screening via smears.

Methods: Statistical modeling can provide a tool to discover such profiles. In this paper the method of classification trees is described. The use of classification trees has advantages but also limitations with respect to their application in the survey domain. A closer look on the handling of missing data and weighting in this context will be given.

Material: The Belgian Health Interview Survey (HIS) was conducted in 1997. The Belgian communities are responsible for cervix cancer screening as a part of the preventional health care.

Results: There are no strong conclusions to be drawn with respect to the objective of determining a typical profile for women that underwent a screening. The application of the methods to the HIS data however provides insights and incitements for further investigation.

  Last update: March 25, 2003  - Contact : S. Malali