Item response theory in high-stakes pharmacy assessments.

OUR SITUATION

Classical test theory (CTT) and item response theory (IRT) are two measurement models used to evaluate results from examinations, questionnaires, and instruments. To illustrate the benefits of IRT, we compared how results from multiple-choice tests can be interpreted using CTT and IRT.

METHODOLOGICAL LITERATURE REVIEW

IRT encompasses a collection of statistical models that estimate the probability of providing a correct response for a test item. The models are non-linear and generate item characteristic curves that illustrate the relationship between the examinee's ability level and whether they answered the item correctly. Several models can be used to estimate parameters such as item difficulty, discrimination, and guessing. In addition, IRT can generate item and test information functions to illustrate the accuracy of ability estimates.

OUR RECOMMENDATIONS AND THEIR APPLICATIONS

Researchers interested in IRT should gather the necessary resources early in the research process and collaborate with those experienced in quantitative and advanced statistical models. Researchers should confirm IRT is the optimal choice and select the model ideal for their needs. Once data are acquired, confirm model assumptions are met and model fit is appropriate. Lastly, researchers should consider disseminating the findings with accompanying visuals.

POTENTIAL IMPACT

IRT can be a valuable approach in assessment design and evaluation. Potential opportunities include supporting the design of computer adaptive tests, creating equivalent test forms that evaluate a range of examinee abilities, and evaluating whether items perform differently for examinee sub-groups. Further, IRT can have noteworthy visuals such as test information and functions.