• Ron D Hays, Julie Brown, Lorraine U Brown, Karen L Spritzer, and James J Crall.
• RAND Health Program, Santa Monica, California, USA. drhays@ucla.edu
• Med Care. 2006 Nov 1; 44 (11 Suppl 3): S60-8.

BackgroundClassical test theory and item response theory methods can provide useful and potentially different insights into the performance of items in a survey designed to elicit parental perceptions of dental care delivered to children in publicly funded programs.ObjectivesWe sought to illustrate the use of both classical test theory and item response theory to evaluate survey instruments.Methods: Using 2 years of cross-sectional data collected from enrollees in dental plans in 2001 and 2002, we studied families with children between ages 4 to 18 who were enrolled in 1 of 5 dental plans for 12 months or longer. The 2001 survey yielded a total of 2536 usable surveys and the 2002 survey yielded 2232 useable surveys (50% and 46% response rate, respectively) for a total sample size of 4036 children who used the plan for most or all of their care.MeasuresThe beta version of the CAHPS(R) dental care survey instrument includes 2 global rating items (dental care, dental plan) and multi-item scales assessing getting needed care, getting care quickly, communication with dental providers, office staff, and customer service.ResultsItem missing data rates were low. Item-scale correlations for hypothesized scales (corrected for overlap) tended to exceed correlations of items with other scales. Classical test theory analyses identified 5 of 10 communication items that did not perform well. Internal consistency reliability estimates for the scales ranged from 0.73 to 0.86. Item response theory painted a more promising picture than classical test theory for the 2 communication items that assessed access to an interpreter when needed.ConclusionsThe beta CAHPS(R) dental survey performed well and the revised instrument is recommended for future studies. Classical test theory and item response theory can provide complementary information about survey items.

21 keywords...

hide…