What are the implications of our results for comparisons of health care organizations based on patient satisfaction data? On the one hand, our results can be interpreted as justifying the need to account for certain differences among health care organizations. This is likely to be particularly true for patient demographic characteristics because sound public policy dictates that health care organizations should not be given incentives to deny care to patients on the basis of their clinical or demographic profiles. Efforts to account for such differences in patient satisfaction analyses would appear to be similar to initiatives comparing the clinical outcomes (eg, patient mortality) of providers after adjustments are made for patient risk factors that are believed to be outside the control of clinicians. To assess the practical implications of such adjustments, we compared rankings of VHA hospitals on the basis of patient satisfaction scores before and after adjusting for the 3 patient characteristics that we found to be consistently associated with satisfaction scores: age, health status, and race. For both inpatient medicine and inpatient surgery, the adjustment procedure had the effect of moving ~10% of the hospitals at least 2 quartiles in the rankings. For outpatient care, the adjustment procedure had relatively little effect on hospital rankings. This analysis suggests that the effect of adjustment may be variable but also not insubstantial depending on the health care service or setting.
From this perspective, resources should be devoted to developing more sophisticated models for adjusting patient satisfaction results. For example, if the ability of certain demographic and clinical variables to predict patient satisfaction scores derives from their relationship to expectations, then more direct and refined measurement of expectations will make more precise adjustment possible.
However, the use of adjustment models for patient satisfaction data arguably undermines the true goals of customer feedback. Berwick,23 for example, suggests that a key to the success of world-class organizations is their ability to deliver what feels like individualized products and services to their customers. This ability is based on a profound understanding of the variation in needs and expectations among subgroups of customers. Instead of a "one size fits all" or "that's our policy" approach, these organizations practice what Berwick calls "mass customization." They have ready on the shelf the "5 sizes that fit 75%" and can readily identify the right size for any given customer at the point of service.
From this perspective, the adjustment of satisfaction data is a potential barrier to the customization of health care services. "Leveling the playing field" with regard to variables like race and health status assumes these factors are "out of the control" of the service provider when in fact these variables may define the strata for mass customization. Although a health care organization cannot (or least should not) control the age, gender, health status, or race of the patients who walk through its doors, the organization is not powerless with regard to the way that it treats those different groups. In Berwick's words,23 to adjust for these factors "is not getting closer to the needs of customers. It is ignoring them." This argument has particular force when applied to race, given the obvious concern that there be no racial barriers to quality health care services.
Perhaps one way to resolve these competing viewpoints is to distinguish between short-term and long-term uses of patient satisfaction data. In the short term, it must be recognized that patient satisfaction data are being made publicly available and that, for better or for worse, consumers are making membership and purchasing decisions on the basis of those data with real financial consequences for health care organizations. With this in mind, careful consideration must be given to the fairness of comparative data or rankings based on patient satisfaction scores. For factors that affect satisfaction rankings but are not likely to be readily within-but not necessarily beyond-the control of health care organizations, some steps may need to be taken to account for differences among organizations in these factors.