Attribute Importance Ratings: Are We Learning Anything We Don’t Already Know?

Attribute importance is a key information need for marketers. An understanding of attribute importance can help explain physician prescribing, it can help identify competitive advantages and disadvantages and it can provide guidance on what needs to be communicated in advertising and promotion. Attribute importance can also be a basis for segmenting and targeting physicians.

The most common approach for assessing attribute importance in the pharmaceutical industry is to ask doctors directly. Doctors are smart people who routinely weigh alternative treatments and make rational prescribing decisions. It is reasonable to assume that they know the reasons for their choices and can describe the relative role each product attribute plays. It follows that the average of importance ratings from a representative sample of physicians experienced in treating a disease should be a good measure of what really matters in treating that disease.

But are physician ratings an accurate measure of how much influence each attribute has on their prescribing decision for a disease? A moment’s reflection suggests maybe not. Importance ratings ask for global estimates of importance but treatment decisions are patient specific. To produce ratings, physicians must weigh the importance of each attribute by the number of patients for which it is important, compare all of the attributes and then come up with a number that accurately represents each attribute’s relative importance. Moreover they usually have to do it while rushing through a lengthy interview and without knowing all of the attributes they will be rating.

A few years ago we ran a large multi-phase project that included an assessment of attribute importance across a wide range of specialties and diseases. The diseases were diverse, including Alzheimers, diabetes, COPD, schizophrenia, rheumatoid arthritis, overactive bladder and HIV. An initial qualitative phase identified a list of 26 attributes that physicians described as clear, relevant and capturing the key dimensions they use to evaluate brands. Subsequently we had 1100 primary care physicians and specialists evaluate the importance of these attributes in determining brand selection for their new start and second line patients in one of the 7 diseases that they treated frequently.

The analysis revealed ratings were amazingly similar across specialty and disease. Correlations of average ratings were uniformly high across disease, specialty and line of therapy. In fact, every correlation was greater than .6 and the average across diseases, specialties and lines of therapy was .87.

Take a minute and think what these high correlations mean. By knowing the relative importance that primary care physicians place on a condition like overactive bladder you can predict the importance rheumatologists place on the same attributes applied to rheumatoid arthritis, and you can do it very accurately. You can predict the importance of attributes for second line patients in diabetes from first line patients with Alzheimer’s disease. The implications are sobering. If there is so little variation in average importance ratings why are we asking for them in market research studies? What can we hope to learn that we do not already know?

While it is tempting to conclude that attribute importance ratings are waste of time and money, there may be a more plausible explanation for the above results. As noted earlier, physicians treat patients one at a time, not as a group. They make their prescribing decisions based on the characteristics and needs of a particular patient, not a group of patients with a common disease. It follows that if you want to measure what is important you need to do it at a patient level, not at a disease level.

Shifting from groups of patients with a common disease to specific patients makes a more complex study but pays off in the richness of information received. Finding the appropriate level of complexity is critical. At RG+A, we have found that a small number of “archetypal patients” each representing a different set of physician treatment goals can provide the richness researchers seek with importance ratings. Asking importance about individual patients also makes the task easier for physicians and encourages them to be specific. Limiting the ratings to a select group of patients representing different goals tells where differences exist without making the task too lengthy.

Would you like to hear more about what we have learned about using specific patients to get richer attribute importance information? Feel free to email me at bduncan@thinkRGA.com, and we can arrange a time to discuss it further.