Editor’s note: David Drachman is director of marketing research at University HealthSystem Consortium, an Oak Brook, Ill.-based alliance of 80 academic health centers.
In health care as in other fields, managers have learned that a satisfied customer is not necessarily a loyal customer, and that customer (patient) loyalty is crucial to their organization’s bottom line (for example, see MacStrivic, 1991). Although many health care institutions collect ample data on patient satisfaction, specific attempts to measure patient loyalty are less frequently encountered. Furthermore, many health care institutions are reluctant to impose yet another set of survey questions on their patient populations. This article describes an approach for inferring patient loyalty from patient satisfaction data of the sort already collected by many health care organizations.
Are loyal patients always satisfied patients?
There are many dramatic illustrations of the fact that a satisfactory experience with a company doesn’t guarantee loyalty to that company (Jones and Sasser, 1995). A highly satisfied customer or patient may nevertheless "defect" to a competitor. Not all satisfied patients are loyal, and from this fact it is easy to assume that satisfaction is a precondition for loyalty. However, some patients who are very dissatisfied with a specific hospital episode may nevertheless remain loyal to the hospital for a variety of reasons, such as past positive experiences at the hospital, a sense of allegiance to the caregivers that treated them, familiarity, lack of a viable alternative, and so on.
Another approach
Let’s look at how we might use this fact to infer loyalty from the standard satisfaction data that many hospitals routinely collect. Data from the University HealthSystem Consortium (UHC) patient satisfaction database will be used to illustrate the approach. This database includes over 11,000 adult inpatient surveys collected from patients of 27 academic health centers (AHCs) that belong to UHC. The survey instrument we are using is provided by the Boston-based Picker Institute. The majority of the survey questions ask the patients to report on, rather than rate, their experiences with hospital care. For example, one of the questions asks:
"When you had important questions to ask a nurse, did you get answers you could understand?"
The possible response categories are:
__ Yes, always
__ Yes, sometimes
__ No
__ Didn’t have questions
If a patient answers this question "Yes, sometimes" or "No" then the patient is considered to have a problem with this aspect of care. Each of the items on the survey is similarly scored as either a problem or not a problem for each patient. Using 28 report-type items from the Picker adult inpatient survey questionnaire, we can create a total problem score that ranges from 0 to 28 for each patient surveyed. Figure 1 shows the distribution of problem scores observed at the participating AHCs.
Like many patient surveys, the Picker Institute surveys have a question that asks patients if they would recommend the hospital to others:
"Would you recommend this hospital to your family and friends?"
__ Yes, definitely
__ Yes, probably
__ No
As we would expect, the response to this question is strongly related to the number of problems reported. Patients who are highly satisfied with their stay are likely to "definitely recommend" the hospital (see Figure 2). This recommendation may stem partly from the positive experience and/or partly from a feeling of loyalty to the hospital. Conversely, patients who experience a number of problems with their care are much less likely to recommend the hospital to others.
We can use the exceptions to this pattern to provide insights into patient loyalty. For example, what about a patient who reports numerous problems with the care received and nevertheless does recommend the hospital to others? If a patient recommends the hospital in spite of a bad experience, this strongly suggests that there is some sort of bond between the patient and the hospital that this bad experience was not able to overcome.
Overall, about 25 percent of the patients reported experiencing 10 or more problems with their care. For each hospital in our database, we can create a rough "loyalty index" by dividing the number of patients who report 10 or more problems and who nevertheless "definitely recommend" the hospital by the total number of patients who report 10 or more problems. How do the various hospitals compare on this loyalty index? Figure 3 shows wide variations. Furthermore, from these results we can construct a loyalty "norm" for adult inpatients at academic health centers of 40 percent. That is, an average of 40 percent of dissatisfied patients would nevertheless definitely recommend the hospital to others.
It is interesting to note that the 27 hospitals rank very differently on this loyalty index than they do on an item that measures overall satisfaction - in fact, the correlation between the two sets of ranks is only .38.
Profiling the loyal patient
Are there certain patient characteristics that are associated with loyalty? We can profile our loyal patients (in terms of background characteristics) to see what distinguishes them from the non-loyal patients. Figure 4 shows a profile of loyal patients for one of the hospitals in the study.
This chart indicates that patients loyal to this particular hospital are more likely to be older, male, less well-educated, with Medicare or Medicaid coverage. Bear in mind that these results are based on dissatisfied patients only (those reporting 10 or more problems with their care), so differences in levels of satisfaction are not a likely explanation of these results.
Conclusions
By looking at the relationship between patient problem reports and willingness to recommend the hospital, we can make some inferences about patient loyalty from patient satisfaction data. If the same patient satisfaction survey is in use at a group of hospitals, then a comparative benchmark for the loyalty measure can be created.
The "loyalty index" described here is clearly a proxy, not a pure measure, of patient loyalty. Some of the patients defined here as "loyal" may simply have been inconsistent in their responses to the survey. However, the strong relationships found between the loyalty index and patient demographic factors cast doubt on this as the only cause. A direct measure of loyalty would be preferable, but this would impose an additional burden on the patients and staff. In the absence of such data, conclusions can still be drawn from patient satisfaction data that is readily available from many health care organizations.
The loyalty link between customer and company is always fragile, and can be easily broken if the customer becomes aware of a new alternative. Follow-up interviews with customers identified as loyal from satisfaction survey data can help to pinpoint the nature of this link and show how it may be strengthened.
References
MacStrivic, R.S. Beyond patient satisfaction: building patient loyalty. Health Administration Press, 1991.
Jones, T.O., and Sasser, W.E. "Why satisfied customers defect." Harvard Business Review, November-December, 1995, 88-99.