Title: The Apriori Algorithm as an Engine for Computerized Adaptive Assessment

Author: Niels Smits

Affiliation:

Abstract:
Both in medical research and clinical practice, questionnaire-based
assessments are increasingly used to obtain information about patients. Such
information is often used to make clinical predictions, classifying
respondents into categories, such as `at risk' and `not at risk' for
pathology. An important consideration in designing questionnaires is to
minimize respondent burden, and computerized assessment provides an
opportunity to make test administration efficient. For questionnaires meeting
certain measurement properties, adaptive testing algorithms have been
developed which not only allow for early stopping, but also for the dynamic
selection of items to minimize testing time. By contrast, questionnaires not
meeting such standards, such as those assessing symptomatology, do not allow
for this methodology, and therefore alternative methods are needed. In the
current study it is illustrated how the apriori algorithm, commonly used in
market basket analysis, may be utilized as a method for computerized adaptive
testing.