|
Activity Number:
|
295
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 11, 2004 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Health Policy Statistics
|
| Abstract - #300995 |
|
Title:
|
Comparison of Fixed and Random-effects Methods for Predicting Cancer Incidence in Iowa Counties Using SEER Data
|
|
Author(s):
|
Jeff M. Allen*+ and Lirong Zhao
|
|
Companies:
|
ACT and University of Iowa and University of Texas M.D. Anderson Cancer Center
|
|
Address:
|
Department of Biostatistics, , ,
|
|
Keywords:
|
cancer incidence ; prediction ; SEER ; model
|
|
Abstract:
|
Predictions of disease incidence occupy an important place in health care and their role is increasing with time. The aim of this research is to compare two different modeling methods on predicting the incidence of breast and colorectal cancer in 99 counties of Iowa based on cancer registry data (SEER) from 1973 to 1999. Two fixed effect regression models and two random coefficient regression models were used to predict the incidence rate in each county for each cancer site. The model performance in terms of predictive ability was evaluated using actual cancer incidence rates from 1999 by Friedman's test. Random coefficients models performed better for predicting colorectal cancer (P = 0.0272), while no significant differences were detected among the models for predicting breast cancer (P = 0.8853). Meanwhile, random coefficient models led to more stable prediction because of the "regression towards to mean" feature in both cancer types. Cancer incidence can be reasonably predicted by statistical models. The obtained predicted results are important in setting priorities for cancer control strategies.
|
