In recent years, health outcomes tools have proliferated. Equating is a psychometric method that allows different measures of the same domain to be meaningfully compared to each other, even if administered to completely separate groups. The purpose of this project is to obtain evidence-based guidelines for equating health outcomes items using item-response theory. The study is designed to test the practical impact of the following equating design variables: sample size, number of common linking items, content representativeness of the common linking items, distributional properties of the common linking items, and type of item-response-theory model employed. The project is a secondary analysis of previous work in developing the Health of Seniors Survey. Because the parent study made extremely conservative equating design decisions, it will serve as a proxy """"""""gold-standard"""""""" for evaluating suboptimal equating settings. Design sensitivities will be demonstrated by the degree to which variance in the scores on health status instruments and health care utilization variables is accounted for by the equated scores. Equatings will be performed both under optimal and suboptimal conditions. Equatings under optimal conditions are referred to as """"""""reference equatings."""""""" For these, the responses of all participants (3,358) will be used, along with the entire set of 71 content-balanced common items from the parent study. Subsamples of persons and subsets of items for equatings defined as suboptimal will be selected. These datasets will be designed to """"""""push the limits"""""""" of the equatings so that the impact on the resulting scores can be evaluated. For each dataset, the Pearson product-moment correlation will be calculated between equated scores under each suboptimal condition and those obtained in the reference equatings. To test whether there is a systematic bias in the scores, the root mean squared error (RMSE) for each equating will be calculated.
| Cook, Karon F; Taylor, Patrick W; Dodd, Barbara G et al. (2007) Evidence-based practice for equating health status items: sample size and IRT model. J Appl Meas 8:175-89 |
