First, thank you for selecting our article and giving us an opportunity to reply to the commentary. We particularly appreciate the praise offered by O'Connor and Shewchuk that the quality and impact of our article would be as sound as those published by Ware and Davies; Bowers, Swan, and Koehler; Cleary and McNeil; and Taylor and Cronin.
We agree with O'Connor and Shewchuk that even though there have been many studies on patient satisfaction, most of them do not employ a theory that would lead to a valid and reliable research study. Without a theoretical background, the study outcomes are only context specific and, thus, they are useful in that environment at that particular time. To advance our knowledge, we need to conduct our research based on a solid theory or a clearly defined framework.
We started this study with our daily experience that we do not process multiple information by a simple mathematical model, rather we make our decision based in a more complicated manner-noncompensatory or nonlinear. The ideas, theories, and models were already developed by Kahneman and Tversky,1 Brannick and Brannick,2 Einhorn,3,4 and Ganzach.5-7 However, few studies have applied these models to a large data set to empirically verify the models. We are fortunate to have access to a carefully collected large data set to test these models.
Our first analysis found that patients do not mathematically average out their attribute reactions to form their overall satisfaction. They arrive at their overall satisfaction level with the noncompensatory model. Our second analysis found that three attributes (nursing care, compassion to family/friends, and pleasantness of surroundings) show a diminishing marginal return function for both overall satisfaction and behavioral intentions. Our third analysis found that there is an interaction effect between overall satisfaction levels and attribute reactions (discharge process and nursing care) on behavioral intentions. These findings show that patients do not process attribute reactions with a simple mathematical model. Thus, O'Connor and Shewchuk wrote in the commentary, "health care management practitioners often make inferences that inform strategic decisions. This may create problems because the resulting practice decisions are often based on evidence that is weighted beyond its merits." Therefore, health care managers need to avoid this pitfall of assuming that an attribute with a larger coefficient is always more important to improve. Another concern for patient satisfaction researchers should be the development of the patient satisfaction questionnaire. Psychometric properties of the instrument are not always examined when a survey questionnaire is developed. Our questionnaire was first developed by experienced patient satisfaction researchers, and then tested with selected patient populations. To further confirm the convergent and discriminant validity of the questionnaire, we used a confirmatory factor analysis. We also examined the reliability of the questionnaire by using Cronbach's alpha coefficient. Without assessing the psychometric properties of the instrument, any further analysis of the data would not provide any meaningful results. We made every effort to assure the psychometric properties of the questionnaire used in our study.
As we discussed in the article, it is important to replicate this study not only with hospital inpatient care but also in other health care settings. Health care organizations are forming an integrated delivery system in which they deliver a continuum of health care services. As a system, it is critical to ensure that patients will have a positive experience throughout the system. Decision makers will be better equipped with the results of additional studies to know how they can select and improve attributes of patients' health care experiences and increase patient satisfaction levels with more efficient strategies.
REFERENCES