1 Department of Anesthesiology, Duke University Medical Center, Box 3094, Durham, NC 27710, USA. 2 Department of Anesthesiology, University of Miami School of Medicine, PO Box 016370, Miami, FL 33101, USA. 3 MEDTAP International, Inc., 7101 Wisconsin Avenue, Suite 600, Bethesda, MD 20814, USA. 4 Department of Anesthesiology, Duke University Medical Center, Box 3094, Durham, NC 27710, USA. 5 RTI Health Solutions, 3040 Cornwallis Road, Research Triangle Park, NC 27709, USA. 6 Pfizer, Inc., 235 East 42nd Street, 235/4/78, New York, NY 10017, USA. 7 Pharmacia Corporation, 5200 Old Orchard Rd., Tower I, 3rd floor, Skokie, IL 60077, USA
*Corresponding author. E-mail: gan00001{at}mc.duke.edu Declaration of interest. Co-authors Connie Chen, PharmD and Tracy Mayne, PhD are employees of Pharmacia Corporation and Pfizer, Inc. respectively, who funded this project.
Accepted for publication: December 13, 2003
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Methods. Fifty patients undergoing major abdominal surgery were enrolled and completed interviews before and after surgery. Measures included an experience with pain questionnaire and an adaptive conjoint analysis (ACA) interview.
Results. Percentage of pain relief obtained post-surgery was between 70 and 80%. Eight-two per cent reported at least one moderate or severe side effect. ACA results demonstrated that pain efficacy and side effect type/severity have almost equal importance scores. Patients varied in their willingness to trade-off pain efficacy for different or milder side effects.
Conclusions. We conclude that people have different relative preferences for different side effects and are willing to trade-off pain relief for less upsetting and/or less severe side effects but to different degrees. Thus, physicians should consider offering pain medications with fewer side effects than narcotics as a first choice. Our study indicates the need to balance analgesia and side effects in order for patients to achieve optimal pain control.
Br J Anaesth 2004; 92: 6818
Keywords: pain, postoperative; pain treatment, patient preference; measurement techniques, conjoint analysis
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Physicians and patients may have differing perspectives regarding the relative importance of these treatment outcomes and thus, which therapy is the best choice. For example, in a study of anaesthesia, physicians considering both frequency and severity of anaesthesia outcomes ranked the top five least desirable outcomes (in order) as incisional pain, nausea, vomiting, preoperative anxiety, and discomfort from i.v. catheter insertion.4 When patients rated anaesthesia outcomes in a companion study, the five most undesirable were (in order) vomiting, gagging on a tracheal tube, incisional pain, nausea, and recall without pain.5 The higher importance placed on postoperative nausea and vomiting in this study is also confirmed by a previous study6 that demonstrated that patients were willing to pay $40100 to avoid postoperative nausea and vomiting.
Conjoint analysis is a technique for indirectly eliciting preferences for products or services features (or attributes) by asking respondents to trade-off combinations of these features.79 The indirect conjoint technique is more effective for identifying true preferences than direct techniques because it forces respondents to make trade-offs similar to the way individuals make decisions in real life. This technique has been used extensively in market research to estimate consumer preferences for products and to predict consumer choice behaviour. More recently, it has been applied to health care to determine patient preferences for treatment, with the ultimate goal of improving medical decision-making and health outcomes. Adaptive Conjoint Analysis (ACA) utilizes a computer-interactive interview that adapts the interview for each respondent, resulting in a less burdensome interview and higher quality data as respondents are more interested and involved in the task.10
Investigating patients preferences in the acute pain setting using conjoint analysis has not been reported previously. We sought to quantify the trade-offs that patients make between pain relief, side effects, and side effect severity. We used adaptive conjoint analysis to quantify patient preferences. Understanding patient preferences allows analgesia to be tailored to an individual. This will maximize the analgesia while also minimizing the side effects.
![]() |
Methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Conjoint analysis
Patients were presented scenarios, which include the following attributes: (1) degree of pain relief, (2) type of side effect, (3) severity of side effect, and (4) setting/route of pain control administration. Each attribute was defined by specific levels (see Appendix for full description). Degree of pain relief ranged from poor pain relief to excellent pain relief; type of side effect included no side effects, vomiting, itching, constipation, etc.; severity of side effect included mild, moderate, and severe; setting/route of pain control administration included inpatient receiving PCA, inpatient receiving pain pills and pain pills at home.
An ACA, which utilizes a computer-interactive interview, was used to administer the trade-off questions. During the conjoint interview, each respondent was asked to trade-off two hypothetical pain control scenarios (defined by the set of attributes and levels) and to provide a rating (i.e. a preference index ranging between 1 and 9) indicating which scenario was preferred and the strength of that preference. The resulting responses were used to estimate a utility function using regression techniques. The respondent utility function quantified the relationship between respondent preferences and the specific pain control attributes or features. Although the utility associated with each pain control scenario is not directly observable, the rating provided by the respondent is observable and is related to the utility value. This rating is used in the regression analysis as the dependent variable in the utility function.11
Data items
At interview 1, subjects provided information on their age, gender, race, marital status, education level, employment status, and co-morbid conditions. Information on pain medications prescribed was obtained from medical records for interviews 2, 3, and 4.
Patients provided written information at interviews 2, 3, and 4 regarding their method of pain control, amount of pain experienced, success and satisfaction with pain control, and side effects they experienced.
Computer-assisted conjoint analysis interview
Subjects completed a computerized interview at each of the four time points using Sawtooth Softwares ACA (Version 5). The four attributes were:
1. Degree of pain control.
2. Type of side effect.
3. Severity of side effect.
4. Setting/route of pain control administration.
Before starting the computer interview, we provided detailed descriptions of each level of each attribute to the respondents (see Appendix). We also told respondents that the time frame for these descriptions was 1 day following surgery. Hence, the patient imagined experiencing the health state defined by each combination of attribute levels on a single day following surgery.
The computer interview first asked respondents to rank their preferences for the various levels of each attribute and to rate the relative importance of each attribute. Examples of these questions are shown in Figure 1. The trade-off questions subsequently completed were different for each respondent depending upon their responses to these initial questions.
|
Data analyses
Data collected were analysed using descriptive statistics. Frequencies, means, and standard deviations were calculated for the socio-demographic characteristics of the sample at interview 1 and the clinical characteristics at interviews 24. ANOVAs were used to compare the experience with pain management at interviews 2, 3, and 4.
Using responses to the conjoint interview at interview 4, we used the ACA program to generate a utility function for each respondent based on OLS regression analysis. Individual utilities for each level of each attribute were generated and normalized so that zero is the lowest utility score for each attribute and higher utilities indicate stronger preference. We also used the ACA program to generate mean utilities and standard errors for the total study population. These indicate the value or utility that respondents have for each level of a particular attribute.
In addition to the utility values, we estimated relative importance scores for each attribute. The importance scores were computed by first summing across all the attributes, and the range of the utility values for each attribute. Then the importance weight for a single attribute was computed as 100 multiplied by the range of the utility values for that attribute divided by the sum of the ranges for all the attributes. The importance scores demonstrate the relative importance of each of the four treatment attributes for determining a patients preferences for treatment. Importance scores were computed for each patient and then mean values computed for all the patients in the study. A combined importance weight of side effect type and side effect severity was also computed and compared with the importance weight for pain control for all patients in the study to demonstrate the relative importance to patients of pain relief vs side effects.
Utility values and importance scores were also estimated using the conjoint interview data from interviews 1, 2, and 3 and compared with those estimated using the data from interview 4. Utility values and importance scores were also estimated for different pre-specified subsets of the population, including gender, pain experience, side effect experience, and type of surgery.
Finally, we conducted simulation analyses to estimate the percentage of patients that prefer one set of treatment attributes over another. These analyses computed the utility (strength of preference) associated with two hypothetical treatments (A and B) with different combinations of attributes for each individual. An individual was assumed to prefer treatment A to treatment B if the utility with treatment A was higher than for treatment B. The proportion of individuals who prefer treatment A to treatment B was estimated based on the results of these calculations in 50 000 hypothetical individuals drawn with replacement from our 50 patient population.12
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
|
|
|
|
|
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The importance scores were stable for the four conjoint interviews given pre-surgery and post-surgery. This result was expected as the conjoint survey asked patients about their preferences for hypothetical scenarios and not about their actual experiences. This stability supports the generalizability of the results from the conjoint estimation.
Results from this study are similar to the findings from a similar study investigating the relative preferences for chronic pain relief.13 Both studies demonstrated that site and delivery mode (PCA in the hospital vs pills at home in this acute pain study; patch vs oral drugs in the chronic pain study) are of less importance to patients than pain relief and side effects. Also, both found that side effect type and severity and pain relief are equally important in determining the utility of the medication outcomes.
This study has several limitations. First, the sample size was small which limited our ability to test for differences in relative preferences between different subpopulations. Secondly, the Sawtooth software assumes that the utility associated with each attribute is independent of the level of the other attributes. For example, this means that the magnitude of the loss in utility from experiencing a side effect is assumed to be the same no matter what level of pain relief is experienced. Thirdly, this study has taken an ex post perspective and investigated hypothetical levels of pain relief and side effect type and severity. This approach has the advantage of generating clear estimates about the utility of different health states, but it does not allow us to estimate the impact of uncertainty in the outcomes. In addition, our study assumes that only one type of side effect is experienced at one time.
In summary, we demonstrated that to optimize patient care, it is necessary for physicians and other healthcare professionals to better understand how patients value these outcomes. Patients have different relative preferences for different types of side effects. They are also willing to trade off pain relief for less upsetting and/or less severe side effects but to different degrees. Such information may guide physicians in selecting pain medications that will best meet the patients needs and expectations and improve his/her compliance, treatment outcomes, and ultimately patient satisfaction.
![]() |
Appendix |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Good
Fair
Poor
Very poor
Side effects
Constipation. You feel bloated and full. It is difficult to have a bowel movement, if you have one at all.
Itching. You feel the need to scratch all or parts of your body.
Mental cloudiness/dizziness. It is difficult to focus. You may be confused, disoriented, or have difficulty concentrating. You or your surroundings may appear to be moving or spinning. If you were to stand up, you might feel unsteady on your feet.
Mood changes/alterations. You may have a rapid increase in your rate of breathing. If you already have symptoms of depression they may become worse. You may have periods of anxiousness, panic, excitability, restlessness, irritability, or mood swings. You may have decreased appetite.
Nausea. You have an upset stomach and you feel like you might throw up.
Nightmares/hallucinations. You may have frightening, strange, or vivid dreams. During waking hours, you may have hallucinations (you think things are happening when they are not).
No side effects. You have no pain medication side effects.
Sleep disorders. You have difficulty falling asleep or staying asleep.
Vomiting. You have an upset stomach and are throwing up.
Route of pain medication
PCA in the hospital. You are receiving i.v. pain medication following surgery. You can control the amount of pain medication you receive by pushing a button. When you push the button, you get more medication through a needle that has already been inserted in your arm.
Pain pills in the hospital. You are taking pain pills by mouth following surgery while in the hospital. You take the pills as often as needed.
Pain pills at home. You have been discharged from the hospital and are taking pain pills by mouth at home. You take the pills as often as needed.
Side effect severity
Mild side effect. Your are not bothered very much by the side effect from your pain medication. You can easily cope with the side effect and do not need any additional medication to treat it. The side effect does not interfere with your daily functioning.
Moderate side effect. You are somewhat bothered by the side effect from your pain medication. It is somewhat difficult to cope with the side effect and you may need additional medication to treat it. The side effect interferes with some of your daily functioning.
Severe side effect. You are very bothered by the side effect from your pain medication. It is difficult to cope with the side effect and you need additional medication to treat it. The side effect interferes with most or all of your daily functioning.
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
2 Strassels SA, Chen C, Carr DB. Postoperative analgesia: economics, resource use, and patient satisfaction in an urban teaching hospital. Anesth Analg 2002; 94: 1307
3 Beauregard L, Pomp A, Choiniere M. Severity and impact of pain after day-surgery. Can J Anaesth 1998; 45: 30411[Abstract]
4 Macario A, Weinger M, Carney S, et al. Which clinical anesthesia outcomes are important to avoid? The perspective of patients. Anesth Analg 1999; 89: 6528
5 Macario A, Weinger M, Truong P, et al. Which clinical anesthesia outcomes are both common and important to avoid? The perspective of a panel of expert anesthesiologists. Anesth Analg 1999; 88: 108591
6 Gan TJ, Sloan F, Dear G, El-Moalem H, Lubarsky DA. How much are patients willing to pay for avoidance of postoperative nausea and vomiting? Anesth Analg 2001; 92: 393400
7 Lancaster K. Consumer Demand. New York: Columbia University Press, 1971
8 Lancaster K. A new approach to consumer theory. J Polit Econ 1996; 74: 13457
9 Hanemann W. Welfare evaluations in contingent valuation experiments with discrete responses: reply. Am J Agric Econ 1984; 69: 33241
10 Huber J. Conjoint analysis: how we got here and where we are? Sequim, WA: Sawtooth Software Research Paper Series, 1987
11 Urban G, Hauser J. Design and Marketing of New Products, 2nd Edn. Upper Saddle River: Prentice Hall, 1993
12 Sawtooth Software ACA User Manual Version 5. Sawtooth Software Inc., Sequim, WA
13 Schmier JK, Palmer CS, Flood EM, Gourlay G. Utility assessments of opioid treatment for chronic pain. Pain Med 2002; 3: 21830[CrossRef][ISI][Medline]