Analysis of Medical Consumer Behavior

ABSTRACT - The three papers discussed below used sophisticated methodological approaches and advanced analytical techniques to assess consumer decision making. One paper reports that FDA is able to assess physicians' perception of advertising through an experimental research procedure. A second paper introduces consumer researchers to the Subjective Probability model to measure consumer beliefs. The third paper uses conjoint analysis to assess physicians' perception of prescription drugs. The study demonstrates that physicians are more likely to use noncomPensatorY than compensatory decision strategies.

Citation:

Michael B. Mazis (1984) ,"Analysis of Medical Consumer Behavior", in NA - Advances in Consumer Research Volume 11, eds. Thomas C. Kinnear, Provo, UT : Association for Consumer Research, Pages: 235-238.

Advances in Consumer Research Volume 11, 1984      Pages 235-238

ANALYSIS OF MEDICAL CONSUMER BEHAVIOR

Michael B. Mazis, The American University

ABSTRACT -

The three papers discussed below used sophisticated methodological approaches and advanced analytical techniques to assess consumer decision making. One paper reports that FDA is able to assess physicians' perception of advertising through an experimental research procedure. A second paper introduces consumer researchers to the Subjective Probability model to measure consumer beliefs. The third paper uses conjoint analysis to assess physicians' perception of prescription drugs. The study demonstrates that physicians are more likely to use noncomPensatorY than compensatory decision strategies.

PHYSICIANS' PERCEPTION OF ADVERTISING JUDGED DECEPTIVE BY THE FDA

The paper delivered by Oliver, Walbridge, and Rheinstein (1983) makes both methodological and theoretical contributions to the consumer behavior and public policy literature. The method of data collection consisted of surveying a systematic sample of physicians stratified by specialty and selected from the telephone directories of five large SMSA's. Two hundred and thirty physicians were selected for inclusion in the study and 655/o (150) of the contacted physicians agreed to participate.

Since most studies employ student subjects or use small samples, the external validity of advertising studies is frequently questioned by policy makers. The Oliver, Walbridge, and Rheinstein study surveys the appropriate target audienceC physicians -- and uses a large sample. As a result, policy makers will place a significant amount of confidence in the findings. The high level of external validity in this study is likely to result in the findings being useful to the policy makers at FDA and FTC.

The "mock journal" approach, using five articles and four prescription drug advertisements, also increased the study's external validity. Physicians tended to read the simulated magazine as they would a medical journal, such as Medical Economics or Medical World News. As a result, readership of the experimental and control advertisements is likely to approximate natural exposure conditions.

The most important contribution for advertising regulators is that the subjective judgment of public policy officials can be verified by empirical research. Government officials should have increased confidence that consumer behavior methodology can be useful in the assessment of deception. This study may lead to increased use of empirical research in deceptive advertising cases. Of course, both regulatory agencies and the courts have been gradually moving toward increased use of empirical data in the assessment of potentially misleading advertisements. This study should increase the confidence regulators have in the usefulness of consumer behavior research.

A second important contribution of the study is that the theoretical approach first proposed by Russo, Metcalf and Stephens (1981) can be used to assess potentially deceptive advertising. Thus, the Russo, et. al. approach has been "validated" using actual deceptive advertisements.

Oliver, Walbridge and Rheinstein discuss three methods of assessing deceptive advertising. The "n percent rule" first discussed by Gardner (1975) and Jacoby and Small (1975) states that if a certain specified percentage of consumers hold a false belief, then there is reason to believe that an advertiser has engaged in false advertising. Oliver, et. al. point out that an "after only" approach to deceptive advertising measurement offers no proof that an advertisement caused the false belief. The "n percent rule" may nevertheless be useful to advertising regulators. If an advertisement has been running for a long time period (such as for Listerine mouthwash), the "n percent rule" may be the only feasible research method. For long-running deceptive television advertising campaigns, "after only" designs may be useful for establishing the existence of false beliefs. A court would then face the difficult task of linking the false belief to a deceptive advertising campaign which has been run for a long period of time. Although covariation does not prove that the ad caused deception, the mere existence of false beliefs preceded a false advertising campaign would De useful to a court. The "n percent rule" could not be used to assess whether a new advertising campaign or a campaign for a new product was deceptive. Consumers may have been deceived, but the advertising would not have been aired long enough to establish a false belief.

While the "n percent rule" may be useful in measuring deceptive beliefs, causal inferences cannot be made using this approach. As a result, investigators have frequently used a second ap?roach to measure consumer beliefs before and after exposure to an allegedly misleading advertisement. In the case of long-running television advertising campaigns, beliefs before exposure to an experimental advertisement would already be extremely high as a result of substantial exposure. After experimental subjects were exposed to a deceptive commercial, the already high false beliefs would be unlikely to change. Therefore, the investigators would falsely conclude that the false advertisement was benign since beliefs did not change after exposure to advertisement. However, the "before and after approach" could be successfully employed if the allegedly misleading advertisement had not been previously seen by the target audience. In this case, initial beliefs would not be contaminated by previous exposure to the false advertisement.

A third approach, which was used by Oliver, et. al. and Russo, et. al., is the "after only with control" design. Using this procedure, two groups of subjects were both exposed to one of two advertisements. The first group was exposed to the allegedly misleading advertisement, while the second group saw a similar advertisement with a false material excised. Each group supplied their perception of the advertisement after exposure. The difference between perception of the experimental and control groups was used as a measure of the amount of deception produced by the false advertisement.

The "after only with control" approach can be used to measure deception in print advertising when subjects had little prior exposure to the deceptive advertisement. However, if a broadcast commercial were under investigation by a regulatory agency, the control group should be exposed to a similar commercial with the deceptive material excised. It would, of course, be extremely difficult to produce a control commercial which is equivalent to the experimental commercial. The "altered" control commercial might be significantly shorter than the misleading experimental commercial if false material is deleted. If deception is endemic to the advertising campaign, a new commercial would have to be created for the control group. The persuasive appeal of the experimental commercial and the control commercial should be difficult to match. Therefore, the "after only with control" design seems appropriate only when print advertising is the subject of regulatory investigation.

Also, the "after only with control" approach would be less appropriate for long-running advertising campaigns. If consumers had a high level of false beliefs before being exposed to the experimental and control commercials, one more exposure would be unlikely to produce any additional deception. As a result, a study may falsely conclude that an ad is not deceptive due to the fact that deception frequently cannot be assessed for long-running campaigns using an "after only with control group" design.

Therefore, none of three measurement approaches is useful for all circumstances; each technique is appropriate in differing situations. Researchers should carefully study the relevant issues before selecting a method for deception detection.

Another important issue raised by Oliver, et. al. is the use of communication effects or brand beliefs as the appropriate dependent variable. Oliver, et. al. state that they used "deceptive potential" as the criterion variable. Physicians were asked to indicate whether a series of statements were implied by the misleading or the non-misleading versions of the experimental advertisement. Public policy makers have generally followed this approach in assessing the misleadingness of on advertisement.

Other investigators, however, (Russo, et. al. 1981) and Jacoby and Small (1975) suggest that brand beliefs are a better criterion for deceptive advertising assessment. While use of brand belief s would be a "better" criterion variable, many regulators would be unhappy with this stringent standard. Litigators prefer "deceptive potential" as a standard since research may show that an advertisement has "deception potential" but may fail to show changes in brand beliefs. No experimental study can adequately measure changes in brand beliefs when long-running deceptive campaigns are tested.

Academic researchers should realize the limitations of their research tools. Experimental procedures are useful when the audience has not been previously exposed to tile allegedly misleading advertisement; however, when deceptive advertisements have been aired for a long period of time either surveys employing the "n percent" methodology or laboratory studies using content recall as a measure of "deception potential" as a criterion are appropriate.

AN ANALYSIS OF GENERIC PRESCRIPTION DRUG ATTITUDES USING TWO BEHAVIORAL INTENTION MODELS

The paper by Brinberg and Cummings (1983) makes a useful contribution to the consumer behavior literature by introducing a "new" behavioral intention model. Consumer behavior researchers have heretofore focused almost exclusively on using the Fishbein model (Fishbein and Ajzen 1975) to assess the factors leading to behavioral intention. However, Wyer and Goldberg's (1970) Subjective Probability Model may prove to be more useful to consumer researchers than the more commonly used Fishbein model.

Brinberg and Cummings also replicate their study across adult and college student samples. The practice of replicating results across samples is typically not followed in a consumer behavior literature. The authors' effort to com?are both student and non-student samples is a practice which should be followed by other consumer behavior researchers.

The results indicate that both college students and adults respond similarly on the topic of intention to purchase generic prescription drugs. Would college students and non-student samples yield the sa ne results for other product categories? The fact that both groups are relatively ignorant about generic prescription drugs tended to confirm the authors' hypothesis that the two groups would respond in a similar manner. If another subject were studied, such as purchasing a house or real estate, the two groups might respond differently.

In examining Fishbein's Theory of Reasoned Action (Fishbein and Ajzen 1975) and Wyer and Goldberg's Subjective Probability Model (1970), the authors do not compare the predictability of the two models. Since the Subjective Probability Model does not sum beliefs as does the Fishbein model, the authors hold that the predictability of the two models cannot be compared. However, the authors could contrast correlations between individual belief times evaluation scores and intention for the Fishbein model with the sum of the two conditional probabilities correlated with intention for the Subjective Probability Model. In this way, predictability of the two models could be contrasted on a belief-by-belief basis.

One question which should interest consumer researchers is: When is the Subjective Probability Model or the Fishbein model more useful? The Subjective Probability Model uses a direct measure of "psychological relevance" of a belief by examining the difference between the two conditional probabilities. '#hen the differences between two conditional probabilities is small, the belief is said to be psychologically irrelevant; when the difference between the two conditional probabilities is large, the belief is considered to be psychologically relevant. Therefore, the importance of a belief can be assessed directly. When direct measures of importance -are needed and relevant to the research study, the Subjective Probability Model is most appropriate. On the other hand, the Fishbein model does not directly assess belief importance. Indices of importance on the entire set of normative beliefs and on the entire set of attitudinal beliefs are determined through regression analysis. In the Fishbein model, the magnitude of the regression coefficients is used as a measure of the importance of the two components. However, no direct assessment of the importance of individual beliefs is made in the Fishbein model.

The Subjective Probability Model is likely to show higher correlation between belief and intention than the Fishbein model The Subjective Probability Model directly ties an individual's belief to behavioral intention by asking subjects about the perceived probability of performing or not performing a certain behavior given that a belief is true or not true. On the other hand, the Fishbein model merely asks respondents to judge whether it is likely or unlikely that a particular attribute is linked to a brand or act. This approach does not link belief as directly to behavioral intention since conditional probabilities are not measured. Therefore, if researchers desire a belief measure which is closely tied to intention, then the Subjective Probability Model would seem more useful than the Fishbein model.

On the other hand, the Fishbein model supplies information which is unavailable in the Subjective Probability Model. An evaluation measure, which judges the goodness or badness of tile attribute, is assessed through the Fishbein model. Therefore, additional diagnostic information is provided to the researcher. An assessment of the subjective norm is also provided in the Fishbein model. Researchers can assess consumers' normative belief and motivation to comply with important referents. An overall attitudinal and normative index is computed with the Fishbein model. If the researcher is interested in additional diagnostic information and in the relative importance of attitudinal and normative factors, the Fishbein model is more appropriate. Both the Subjective Probability and Fishbein models are potentially useful to consumer researchers.

DECISION ROLE SEGMENTATION AMONG PHYSICIANS

The paper by Rosenberg and Webster (1983) is characterized by a large sample, replication, and sophisticated analytical techniques. The authors interviewed two hundred and thirty physicians across two studies. Data were analyzed using conjoint analysis, cluster analysis, and discriminant function analysis.

The authors stated "it was hypothesized that physicians would be more likely to use compensatory decision rules." However, the authors failed to statistically test the aforementioned research hypothesis. The two studies reported suggest that physicians do not use compensatory decision rules, but statistical testing is not undertaken.

The authors conducted two research studies among the physician population; however, they failed to show a relationship between the two studies. Conjoint analysis is used in the first study, while conjoint analysis, cluster analysis, and confirmatory discriminant function analysis is used in the second study. Since the two studies are similar in design, why are different analytical techniques used?

The current study and a number of other consumer research studies have attempted to use "fitting" methodologies to detect and assess decision making strategies. Can an approach which fits a statistical model to consumer preference data determine the underlying decision processes used by consumers? While individual choice models can be used to predict future purchasing behavior, it is unlikely that statistical "fitting" approaches can discover the processes used by consumers to make decisions. Conjoint analysis is useful for testing new product concepts and for determining the various levels of product attributes which may be desired by consumers. However, conjoint analysis requires the researcher to fix various levels of specified attributes. Therefore, the technique is affected by the attributes and levels established by the investigator. Therefore, when only one attribute has emerged as salient from a conjoint analysis study, the consumer [nay be employing a noncompensatory strategy by focusing on only a single attribute and failing to average across attributes. However, an equally plausible explanation is that only one of the attributes supplied by the investigator is important to the consumer and/or the levels fixed by the investigator are not sufficiently sensitive to differences in consumer preference. Conjoint analysis is extremely useful for testing preferences for new products. However, the technique is inappropriate as a method for determining decision strategies used by consumers. Procedures based on free response protocols are more useful in determining decision making strategies.

REFERENCES

Brinberg, David and Cummings, Vicki (1983), "Purchasing Generic Prescription Drugs: An Analysis Using Two Behavioral Intention Models," Advances in Consumer Behavior, 10, in press

Fishbein, Martin and Ajzen, Icek (1975), Beliefs, Attitudes, Intention and Behavior: An Introduction to Theory and Research, Reading, Mass.: Addison-Wesley.

Gardner, David U. (1975), "Deception in Advertising: A Conceptual Approach," Journal of Marketing, 39 (January), 40-46.

Jacoby, Jacob and Small, Constance (1975), "The FDA Approach to Defining Misleading Advertising," Journal of Marketing, 39 (October), 65-73.

Oliver, Richard L., Walbridge, R. Hoyt, and Rheinstein, Peter .H. (1983), "A Study of Physicians' Perception of Advertising Judged Deceptive by the FDA," Advances in Consumer Behavior, 10, in press.

Rosenberg, Phillis A. and Webster, Sandra K. (1983), "Drug Therapy Decision Rules Among Physicians: Decision Rule Segmentation," Advances in Consumer Research, 10, in press.

Russo, J. Edward, Metcalf, Barbara L., and Stephens, Debra (1981), "Identifying Misleading Advertising," Journal of Consumer Research, 8 (September), 119-131.

Wyer, Robert S. and Goldberg, Lee (1970), "A Probabilistic Analysis of Relations Among Beliefs and Attitudes," PsYchological Review, 77,100-120.

----------------------------------------