Abstract
Background/Aims In recent years, pharmaceutical companies have been requested by Food and Drug Administration to report item fit statistics when making decisions about item inclusion/exclusion of Health Outcome questionnaires. Clinical researchers, often with no training in item response theory and no IRT-specific software (e.g. WINSTEPS, BILOG, etc), are forced to outsource such tasks to academic institutions. In addition to monetary expenses, clinical researchers, not fully understanding concepts of IRT models, spend considerable time communicating with educational/psychological methodologists to consolidate clinical insights from results. The current paper aims to help circumvent such costly expenses and provide hands-on SAS resolutions to health outcome researchers. Because outcome research often uses questionnaire items with graded responses, the current paper provides SAS codes to conduct Partial Credit Model and generates fit statistics accordingly. Results of fit statistics will be compared with those from conventional psychometric methods (e.g., Cronbach’s Alpha, Item-total Correlations, etc) and thus help outcome researchers evaluate the contribution of IRT in survey item development process.
