The 131st Annual Meeting (November 15-19, 2003) of APHA

Abstract #62867

Stuart A Gansky, DrPH, Oral Epidemiology & Dental Public Health, University of California, San Francisco, 3333 California St, Suite 495, San Francisco, CA 94143-1361, 415-502-8094, sgansky@itsa.ucsf.edu

Recently, knowledge discovery and data mining (KDD) methods have started to be used in some health research studies. KDD risk prediction methods (such as logistic regression, classification and regression trees, and artificial neural networks) are compared and contrasted graphically and numerically in terms of accuracy and interpretability. The illustrative example is a California statewide oral health needs assessment (Shiboski et al, 2003) of 2649 preschool children residing in 271 unique ZIP codes and attending 84 preschools. Individual (eg race/ethnicity and feeding practices), family (eg socioeconomic status) and neighborhood (eg preschool type, Census information, and area health care provider supply) level characteristics are utilized to identify children at high risk for early childhood caries (tooth decay in preschoolers). Particular attention is given to the strengths and weaknesses of the KDD methods. If used properly, KDD methods can be useful tools in public health data analyses. Support: US DHHS/NIH/NIDCR U54 DE14251

Learning Objectives:

• Learn pros and cons of data mining methods

Keywords: Risk Factors, Health Disparities

Presenting author's disclosure statement:
Organization/institution whose products or services will be discussed: SAS Institute Inc. Salford Systems
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.