Activity Number:
|
215
|
Type:
|
Contributed
|
Date/Time:
|
Tuesday, August 13, 2002 : 12:00 PM to 1:50 PM
|
Sponsor:
|
Section on Statistics in Epidemiology*
|
Abstract - #300665 |
Title:
|
A Decision Tree for Tuberculosis Contact Investigation
|
Author(s):
|
Shenghui Tang*+ and Lynn Gerald and William Bailey
|
Affiliation(s):
|
University of Alabama, Birmingham and University of Alabama, Birmingham and University of Alabama, Birmingham
|
Address:
|
WTI 153, 1530 3RD AVE S, Birmingham, Alabama, 35294-3300, USA
|
Keywords:
|
Logistic Regression ; Tuberculosis Contact ; Classification and Regression Tree ; Decision Tree ; TB Background Rate
|
Abstract:
|
A tree-base model was developed and tested to predict a positive tuberculin skin test among contacts to an active TB case. The Classification and Regression Tree (CART) algorithm was used in analysis. The data set contained data collected on 292 consecutive cases and their 2941 contacts seen by the ADPH from 1/1998 - 10/1998. The decision tree was developed with 80% sensitivity and 42% specificity; it was tested with 90% sensitivity and 22% specificity, using prospectively collected data from 366 new TB cases and their 3162 contacts. Use of the decision tree would allow us to decrease the number of contacts investigated by 20% for the test sample thus greatly reducing the costs associated with contact investigation. The tree-base model was compared with a logistic regression model.
|