Abstract
Background/Objectives: Patients' reports on their satisfaction with the care received often have been seen as a key quality indicator of hospital performance. However, the potential effect of different approaches to its measurement has not been adequately assessed in the health care setting. This study therefore aimed to methodologically compare two different response formats in patient satisfaction questionnaires-Likert scales and Item-Specific Response Options (ISRO)-within a Greek public hospital context. The aim was to comparatively explore resulting item- and scale-level score values, ceiling effects, acquiescence bias, and psychometric properties, including reliability and validity. Methods: An overall sample of 400 hospitalized patients at a National Health Service general university hospital was randomly assigned to two groups during February-March 2025. One group completed a Likert-scale questionnaire and the other a questionnaire, with the same content, that employed an ISRO format instead. The questionnaire items covered two aspects of the hospital experience, these being the satisfaction with doctors/nurses as well as the organization and planning of care. Statistical analysis involved Kolmogorov-Smirnov tests for normality, descriptive statistics, chi-square and Fisher's exact test, t-tests, Mann-Whitney tests, ceiling effects, regressions, Cronbach's alpha coefficients, and confirmatory factor analysis (CFA), with measures of composite reliability and average variance extracted and model fit indices. Results: Our analysis identified differences in the distributions of patient responses for many items, including variations in median values and the proportion of positive answers. ISRO items tended to produce higher ratings for nursing care and overall satisfaction, whereas Likert items yielded higher scores in organizational aspects. However, the magnitude of these differences was generally small. Regression analysis, adjusting for length of stay, confirmed statistically significant but modest differences in scale scores between formats. Neither format was superior in terms of ceiling effects, whereas no consistent evidence of acquiescence bias was found. Psychometric testing showed that Likert scales had somewhat higher internal consistency reliability and convergent validity, while ISRO exhibited a better model fit in CFA. Conclusions: The item response format seems to affect reported satisfaction scores, yet the impact is rather limited in practical terms for decision-making. Since neither format is consistently superior, the choice between them should depend on study aims, respondent burden, and the intended use of satisfaction scores by policy makers. Moreover, concerns about acquiescence bias may have been overstated in the health care context. Future research should extend these comparisons with other instruments and larger and more diverse samples, as well as employ complementary methods to clarify how response format affects patient satisfaction measurement.