Nonlinear Modeling and Predicting of Clinical Laboratory Testing

• Activity Number: 69
• Type: Contributed
• Date/Time: Monday, August 12, 2002 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology*
• Author(s): Cheng Wang*+ and Jianming He and Eileen Koski and R. Peter Mallon
• Affiliation(s): Quest Diagnostics and Quest Diagnostics and Quest Diagnostics and Quest Diagnostics
• Keywords: Clinical epidemiology ; Clinical Laboratory Testing ; Non linear modeling ; Pattern and trends recognition

Abstract

Clinical laboratory testing varies seasonally in terms of testing volume and average levels. To monitor the testing trends, testing volume and average value of Lyme disease antibody have been collected for a two-year time period through a clinical laboratory data warehouse stated elsewhere. Non-linear model has been applied to the clinical laboratory data by using a newly developed procedure from a Statistical Analytical System (SAS, Version 8.02, SAS Institute) (Df=4, X2=215.2, P< 0.0001). As new tests come out, the ordering trends usually follow certain patterns. Using the same approach as above, ordering testing pattern of H. pylori IgG has been modeled with seven years of data for the purpose of facilitating prediction of the tests ordering trends (Df=4, X2=248.8, P< 0.0001). The comparisons between different laboratory tests by using the same methodology have been performed to test and adjust the established models.