The goal of this project is to develop statistical methods to facilitate comparative effectiveness research in regard to pediatric screening. As a result of a growing number of professional recommendations and state-sponsored programs, pediatricians are increasingly using parent-report instruments to screen for developmental and behavioral disorders in primary care. Despite years of research, the evidence supporting many of the most frequently used instruments suffers from methodological problems including low sample size, verification bias (e.g., when the diagnostic status is verified only among children who screen positive), and spectrum bias (e.g., when research is based on a sample that does not match the spectrum of patients for whom the instrument is intended). High quality research is urgently needed to validate and compare the effectiveness of pediatric screening instruments.
A central challenge to such research results from the fact that available diagnostic reference tests are expensive to administer and have imperfect reliability. The problem of expense makes verification of a large primary care sample impractical; the lack of a true gold standard biases results. Fortunately, solutions have been proposed. Two-phase study designs, in which a large population receives screening tests in phase I and selected participants receive full diagnostic testing in phase II, are commonly used to increase efficiency. Latent class analysis (LCA) has been used to account for imperfect reference tests, and it has more recently been used to improve the selection of participants for phase II testing, thus increasing the precision of results while controlling study costs.
Although promising, these methods are not fully developed, and they are unfamiliar to pediatric researchers. The specific aims of this study are to apply past developments in LCA methods to research in the comparative effectiveness of screening instruments by:
- refining structural models used to compare and validate screening instruments in 2-phase designs; and
- developing guidance for creating optimal selection strategies and for conducting power analyses.
Results will be disseminated to the pediatric community through submissions to pediatric research conferences and a submitted manuscript. The new methods will also be central to an application for NIH funding to investigate autism screening instruments. This project furthers T2 translational pediatric research by expanding research on pediatric screening instruments to larger patient populations.