Scoring methods and measurement properties of health status measures: An empirical comparison of summative and item response theory scoring for the SF-36 physical functioning scale

• Main Author: Chan, Kitty Suk-Lin
• Format: Dissertation
• Language: English
• Published: ProQuest Dissertations & Theses 01.01.2001
• Subjects: Comparative studies; Educational evaluation; Educational tests & measurements; Physical fitness; Psychological tests; Public health; Quantitative psychology
• ISBN: 9780493154985; 0493154981

Summative scoring, a popular method for scoring multi-item scales, has been criticized for producing scores with only ordinal measurement properties. Some researchers have suggested that item response theory (IRT) may be useful for improving the measurement level of multi-item scale scores. However, empirical evidence of its usefulness for scoring health status instruments has been limited. Using data from five independent studies (two asthma samples, one end-stage renal disease sample, one injury sample, and a random sample of the general U.S. population), the effect of IRT and summative scoring on the measurement precision of the 10-item SF-36 physical functioning scale was compared. The nature of score assignment for the two scoring methods was described. Measurement precision was examined using the standard error of measurement from classical testing theory and the standard error of the IRT scores. The influence of scoring method on score distribution and variance, sample size requirements and heteroscedasticity in linear regression models were also investigated. Using the data from the asthma subjects who participated in a clinical trial, the effect of scoring method on the responsiveness of the SF-36 physical functioning scale was compared with three standardized indices of change (i.e., effect size, standardized response mean, Guyatt's responsiveness index). The relationship of summative and IRT change scores with changes in other health variables were also examined through correlation and linear regression analyses. Study results suggest that item response theory offer useful techniques for investigating measurement precision and its approach to score assignment offer potential benefits for improving precision and instrument responsiveness. However, benefits from IRT scoring may not be generalizable to other health status instruments and experimental conditions. A priority for future research on the usefulness of item response theory for health status applications would be to determine the conditions under which item response theory would prove beneficial.