- Posts: 122
PROMIS, NIHToolbox and NeuroQoL Patient-Reported Outcome scoring uses "Expected A Priori" (EAP) pattern response methods to estimate latent trait scores (Bock & Mislevy, 1982).
Our scoring methods (except our "look-up" table scoring) use any available non-missing responses for an individual respondent to predict the most accurate score possible. This practice of handling missing responses is based on the premise that missing responses occur at random and are not correlated with the construct being measured. This assumption allows us to estimate a score when, at a minimum, 1 non-missing response is present. The lack of measurement precision due to the presence of one or more missing responses will be reflected in the elevated standard error of scores.
At this time, we do not recommend that HealthMeasures users use "look-up" table scoring when data on some item responses is missing. Instead, we encourage users with missing data to utilize the free HealthMeasures Scoring Service, which is uses EAP pattern response scoring methods.
Bock, R. D., & Mislevy, R. J. (1982). Adaptive EAP estimation of ability in a microcomputer environment. Applied psychological measurement, 6(4), 431-444.
I plan to use HealthMeasures Patient-Reported Outcomes in my research, but the sponsor wants to know how missing data is handled in your scoring algorithm (e.g., AssessmentCenter API, NIHToolbox iPad Apps, REDCap). Help?