Background: A self-report questionnaire is the most widely used method to assess (in)patients' satisfaction with (hospital) care. However, problems like nonresponse, missing values, and skewed score distributions may threaten the representativeness, validity, and reliability of results. We investigated which of alternative item-response formats maximizes desired outcomes.
Design: Five formats were compared on the basis of sample characteristics, psychometric properties at the scale and item levels, and patients' opinions of the questionnaire.
Subjects: Consecutively discharged patients (n=784) were sampled, of which a representative (sex, age, length of hospital stay) subsample of 514 (65%) responded.
Measures: A 54-item satisfaction questionnaire addressing 12 aspects of care was used. Patients responded using either a 10-step evaluation scale ranging from "very poor" to "excellent" (E10), a 5-step evaluation scale ranging from "poor" to "excellent" (E5), or a 5-step satisfaction scale ranging from "dissatisfied" to "very satisfied" (S5). The 5-step scales were administered with response options presented as either boxed scale steps to be marked or words to be circled.
Results: E5 scales yielded lower means than S5 scales. However, at the item level, the S5 scale showed better construct validity, more variability, and less peaked score distributions. Circling words yielded fewer missing item scores than marking boxes. The E5 scale showed more desirable score distributions than the E10 scale, but construct validity and reliability were lower.
Conclusions: No large differences among formats were found. However, if individual items are important carriers of information, a (5-step) satisfaction response scale, with response options presented in words next to each item, appears to be the optimal format.