Functional Assessment of Cancer Therapy (FACT-G): non-response to individual questions

The purpose of this investigation is to examine methods of scoring the FACT-G when there is non-response to individual questions. Using completed questionnaires from 350 patients, random and non-random missing responses where simulated. Seven methods of scoring the FACT-G are compared on the basis of accuracy (bias and precision) of both population estimates and prediction of individual scores. Substituting the mean of the completed items in the subscale when more than 50% are completed is generally the most unbiased and precise approach. Case deletion is the worst approach and results in clinically significant bias when the missing responses were non-random and a lack of precision when the rate of non-response was high.