The measurement and understanding of treatment satisfaction is coming of age; a fact nicely illustrated by this supplement to Value in Health, which presents several approaches to the measurement of treatment satisfaction in multiple diseases. Historically, researchers working within a business and marketing framework were the first to work with the satisfaction construct, and have created an extensive literature on consumer satisfaction. With respect to health, consumer satisfaction was also the first area of study, where such variables as satisfaction with the physician, overall care, or time kept waiting were considered in evaluating the overall quality of care [1]. More recently, treatment satisfaction has become a focus of study, leading to the inevitable development of new disease-specific measures, such as the OsteoArthritis Treatment Satisfaction Questionnaire [2], a measure of treatment satisfaction in erectile dysfunction [3], as well as a broader, generic measure of satisfaction with medication, the Treatment Satisfaction Questionnaire Medication [4]. Theoretical work, which ought to be leading the way in this new field, has not kept pace [5]. I thought bringing together a collection of articles on the subject of treatment satisfaction would both move the field forward and showcase the work of Pfizer scientists and other colleagues in this area. Clearly, much remains to be learned about the concept of treatment satisfaction, both theoretically and empirically, but treatment satisfaction ought to be considered an important outcome, when applied selectively, for the evaluation of medicines.

Treatment satisfaction is a concept that is conceptually distinct from other patient outcomes with which it is sometimes confused, such as quality of life, and different predictions arise from working with it as a result. Satisfaction is inherently among the most subjective of all constructs, but can it be measured reliably? Without reliable measurement, the concept will be useless in making predictions. All of the articles in this supplement provide evidence that treatment satisfaction can be measured reliably.

Satisfaction is fundamentally appetitive, yet also cognitively complex in the case of satisfaction with medication. It doesn’t require much new thought or experience, for instance, to know whether one is satisfied with a meal one has eaten—it's just a gut feeling. On the other hand, satisfaction with medication may be influenced by a variety of factors, such as whether it works, how strong the side effects are, whether it restores or interferes with functioning, is it convenient to take, or even how it tastes, among others. The evaluation of treatment satisfaction becomes more difficult when the patient has no immediate physical sensations they can associate with medication effects, such as reduction in physical symptoms or occurrence of side effects.

Several authors agree that satisfaction is at least partly a function of prior expectation [5]. This assertion raises several questions: What else may drive treatment satisfaction if not expectation, and how do expectations behave? Does the construct of treatment satisfaction behave as other patient-reported outcomes do that may be less related to expectation?

A strong theoretical understanding of the construct of treatment satisfaction is important because it will help to generate new ideas for research, and help regulators and researchers to evaluate empirical results in the area. There are several important social scientific theories that can be mined for insights related to treatment satisfaction. Before my first project in the area of treatment satisfaction, I turned to social learning theory (SLT) as developed by Julian B. Rotter [6]. SLT is of course a broad theory of human behavior, designed to predict behavior across the variety of situations present in human experience. Rotter's theory is perhaps best known for the concept of internal-external locus of control of reinforcement, which was one of the most widely cited and studied concepts in the psychological literature, and which led to the development of measures of health locus of control. It is not this particular application of SLT that was most relevant; rather, it was general predictions about expectancies that seemed most applicable.

Rotter suggests that one of the predictors of behavior in a particular situation is expectancy. There are different types of expectancies and rules for how they behave. One important distinction is that there are generalized expectancies and specific expectancies. In situations in which one has a significant amount of experience, specific expectancies relevant to that situation are more likely to operate. On the other hand, in novel situations, one applies generalized expectancies, which could have been developed from other situations or perhaps from the more general store of one's experience. Satisfaction has been defined as the correspondence between one's actual experience and one's expected experience. Herein lies the importance of the insight from SLT: If one has had a chronic history with a particular illness and tried many different medications, one will form a very different kind of expectation and therefore potentially a different satisfaction level than if one is trying one's first medication for a newly diagnosed condition. Of course, this type of process operates across a vast number of situations that are readily apparent with a little reflection: The example Rotter [7] uses refers to buying spoiled meat in a market. Whether or not one returns to the market to buy meat again might depend on, among other factors, how many times one has previously bought meat at that particular market. The prediction of course is that the fewer the times one has bought meat there, the more one would be influenced not to return by a rotten purchase. Another point is that in tasting spoiled meat, one reaches a threshold of dissatisfaction very quickly!

When considering medications, what does this example tell us? One possibility for testing is that satisfaction with a previously untried medication might be strongly related to one's time with symptoms of the condition being treated and one's previous amount of experience with other treatments. The more a treatment diverges from the patient's past experience, the greater the impact it will have on satisfaction levels. Obviously, the greatest possible divergence might occur when the patient has never tried any treatment before. If the patient suffers bad side effects from a medication they have never tried before for an illness for which they have never been treated, they might be so dissatisfied they would not try treatment again. Another hypothesis to be tested is that valid and stable assessments of treatment satisfaction might be formed rapidly once a certain threshold of experience is reached. In the case of medication, a valid and stable rating of satisfaction can perhaps be obtained once the medication reaches steady state from a pharmacokinetic/pharmacodynamic perspective, or the levels rise into the therapeutic window. Such assessments may not change much with time, once they have been formed. Hence, there might be little value in tracking satisfaction levels over a long period of time. Another implication is that there is no such thing as a premedication baseline level of satisfaction that is relevant for the evaluation of a medication.

Another source of theory that might be applied to make useful predictions about treatment satisfaction is prospect theory [8] and its update, cumulative prospect theory [9]. Prospect theory was developed to address some of the deficiencies of expected utility theory, and concerns itself with how people make decisions under conditions of uncertainty by valuing certain outcomes, including health outcomes not yet experienced [10]. To the extent that satisfaction with medication results from the value achieved through effective treatment, some of the predictions about valuation under conditions of uncertainty from prospect theory may also be relevant to value as experienced. There may be several useful predictions from prospect theory that could be tested in the context of the measurement of treatment satisfaction. First, the value of a change in an outcome is not consistent across the spectrum of possible outcomes, but is much more dramatic at either extreme end of the spectrum. For example, if one is completely asymptomatic and in vibrant health, the presence of a mild symptom would be much more distressing than it would be if the same symptom occurred on top of several other symptoms or closer to the middle of the range of health states. Another prediction from prospect theory is that losses are given much more importance than gains of the same magnitude. A reduction in satisfaction would be more important to the patient than a gain of the same magnitude. Of course, these predictions from prospect theory need to be borne out empirically in the study of treatment satisfaction. They do potentially have an impact, however, in how results from treatment satisfaction studies are interpreted.

Yet another potential source of theory-based predictions about the way satisfaction operates are insights on how well-being functions, as described by Sirgy [11]. He suggests that people seek to maximize well-being by having positive experiences, but only to a certain extent, because frequent positive experiences contribute more to long-term well-being than intense positive experiences do. As a result, there is a threshold of well-being past which most individuals do not wish to go, because that would raise the adaptation level, and the higher the adaptation level, the greater the likelihood that future events will be judged as less satisfying. Might there also be a level of satisfaction with medication treatment that would be enough? In other words, is there a satisfaction threshold? If so, what would be the importance of that in the evaluation of medicines and how satisfaction is measured?

These collected observations from SLT, prospect theory, and theories of well-being have implications for clinical trial design. First, to help control for different levels of experience with medication, a cross-over design is optimal, so that at least every trial participant experiences all of the treatments (this would suggest that the assessment of satisfaction may not be best performed in a dose-ranging study, as a result of the complexity of the crossover). This design also helps to create a baseline of shared experience among trial subjects. Another implication is that satisfaction should not be measured when treatment begins, but should be delayed until the time that the medication reaches therapeutic levels (at minimum) and has had time to take effect. Another implication is that prior experience with medication for the condition under study, and the length of time with the condition, should both be measured and used as a covariate in analyses of satisfaction as an endpoint.

In addition to the importance of expectations, research with the concept of satisfaction has shown that overall treatment satisfaction is affected by a number of other variables, such as medication efficacy, side effects and tolerability, level of functioning, and convenience [4]. There may be other factors as well, such as taste and cost, but the most important determinant of satisfaction appears to be how well the drug works.

Of course treatment satisfaction can’t trump the endpoints of efficacy, safety, mortality, tolerability, and so forth, which we have always used to evaluate new medicines and will continue to do. But as the hurdles to developing new products get higher, and an increasing number of markets have generic competition or at least multiple treatment alternatives, satisfaction can be an important variable to assist in product differentiation [12]. By measuring treatment satisfaction, we can bring another aspect of the patient's experience to the evaluation of new and existing medicines and other treatments. The patient's perspective has become increasingly important over the last several decades and has led to the rapid growth of the field of patient-reported outcomes and quality-of-life assessment, not to mention increasing empowerment of the consumer within the healthcare system. And it will remain important, particularly as the healthcare system will require greater copayments and increased financial contributions from patients in the purchase of their care. Treatment satisfaction can be a very important way in which patients will be able to differentiate between similar products and identify value worth paying for. Nevertheless, without theoretical advances and a rigorous approach to the measurement of satisfaction, the field will not move forward. It is my hope that the work presented by my colleagues in this supplement of Value in Health will help to stimulate the growth of the field of the scientific measurement of treatment satisfaction.


  1. Top of page
  2. References
  • 1
    Cleary PD, McNeil BJ. Patient satisfaction as an indicator of quality care. Inquiry 1988;25: 2536.
  • 2
    Pouchot J, Trudeau E, Hellot SC, et al. Development and psychometric validation of a new patient satisfaction instrument: the osteoARthritis Treatment Satisfaction (ARTS) questionnaire. Qual Life Res 2005;14: 138799.
  • 3
    DiBenedetti DB, Gondek K, Sagnier PP, et al. The treatment satisfaction scale: a multidimensional instrument for the assessment of treatment satisfaction for erectile dysfunction patients and their partners. Eur Urol 2005;48: 50311.
  • 4
    Atkinson MJ, Sinha A, Hass SL, et al. Validation of a general measure of treatment satisfaction, the Treatment Satisfaction Questionnaire for Medication (TSQM), using a national panel study of chronic disease. Health Qual Life Outcomes 2004; 26: 212.
  • 5
    Shikiar R, Rentz AM. Satisfaction with medication: an overview of conceptual, methodologic, and regulatory issues. Value Health 2004;7: 20415.
  • 6
    Rotter JB. Social Learning and Clinical Psychology. New York: Johnson Reprint company, 1973, 1980 (originally published, 1954).
  • 7
    Rotter JB. The Development and Application of Social Learning Theory. New York: Praeger, 1982.
  • 8
    Kahneman D, Tversky A. Prospect theory: an analysis of decision under risk. Econometrica 1979;47: 26391.
  • 9
    Tversky A, Kahneman D. Advances in prospect theory: cumulative representation of uncertainty. J Risk Uncertain 1992;5: 297323.
  • 10
    Wittenberg E. Measuring health-related quality of life: does prospect theory help? In: LenderkingWR, RevickiDA, eds., Advancing Health Outcomes Research and Clinical Applications. Mclean, VA: Degnon Associates, in press.
  • 11
    Sirgy MJ. The psychology of quality of life. In: LenderkingWR, RevickiDA, eds., Advancing Health Outcomes Research and Clinical Applications. Mclean, VA: Degnon Associates, in press.
  • 12
    Rosenstock J, Cappelleri JC, Bolinder B, Gerber RA. Patient satisfaction and glycemic control after 1 year with inhaled insulin (Exubera) in patients with type 1 or type 2 diabetes. Diabetes Care 2004;27: 131823.