Randomized Trial of Presenting Absolute v. Relative Risk Reduction in ...

10 downloads 120 Views 149KB Size Report
Background. The authors performed a randomized con- trolled trial to test the effect of 2 different formats of risk reduction information when using conjoint ...
Randomized Trial of Presenting Absolute v. Relative Risk Reduction in the Elicitation of Patient Values for Heart Disease Prevention With Conjoint Analysis Jennifer M. Griffith, DrPH, MPH, Carmen L. Lewis, MD, MPH, Sarah Hawley, PhD, MPH, Stacey L. Sheridan, MD, MPH, Michael P. Pignone, MD, MPH

Background. The authors performed a randomized controlled trial to test the effect of 2 different formats of risk reduction information when using conjoint analysis to elicit values about heart disease prevention. Methods. Participants ages 30 to 75 were enrolled and presented the same hypothetical scenario: a person with a 13% ten-year risk of heart disease. Participants then worked through a values elicitation exercise using conjoint analysis, making pairwise comparisons of hypothetical treatments that differed on 5 attributes. For the attribute ‘‘ability to reduce heart attacks,’’ participants were randomized to receive either absolute risk reduction (ARR) or relative risk reduction (RRR) information. Participants selected which attribute they felt was most important. Participants’ responses to the pairwise comparisons were then used to generate their most important attribute using ordinary least squares regression. Outcomes included differences

between groups in the proportion choosing and generating ability to reduce heart attacks as the most important attribute. Results. In total, 113 participants completed the study: mean age was 51, 29% were male, 52% were white, and 42% were African American. The proportion who selected the ability to reduce heart attacks as the most important treatment attribute did not differ significantly (64% RRR; 53% ARR, Fisher’s P = 0.26). For the conjoint-generated most important attribute, those receiving the RRR version were significantly more likely to generate ability to reduce heart attacks as the most important attribute (59% RRR; 35% ARR, Fisher’s P = 0.01). Discussion. Risk presentation format appears to affect the perceived value of different treatment attributes generated from conjoint analysis. Key words: conjoint analysis; risk reduction; heart disease prevention. (Med Decis Making 2009;29:167–174)

T

values elicitation in health decision-making research has not been determined to date. One promising technique for values elicitation is conjoint analysis. Conjoint analysis is a technique from marketing research introduced by Green and Rao2 that has been used to identify values for features of different products.3 Conjoint analysis assumes that individuals are not able to reliably express how to weight the various features of products but that they can be estimated indirectly.4 During conjoint analysis, consumers are presented with hypothetical ‘‘products’’ composed of different features

he International Patient Decision Aid Standards (IPDAS) collaborative noted the need to develop tools to assist patients with values elicitation and clarification.1 However, the best means to accomplish

Received 28 March 2008 from the Cecil G. Sheps Center for Health Services Research (JMG, CLL, MPP) and School of Medicine, Division of General Internal Medicine (CLL, SS, MPP), University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, and Center for Behavioral and Decision Sciences in Medicine, University of Michigan, Ann Arbor (SH). Dr. Griffith’s current affiliation is the School of Rural Public Health, Texas A&M Health Science Center, College Station, Texas. Paper presented at the International Shared Decision Making Conference in Freiburg, Germany, May 2007. This study was funded by the Foundation for Informed Medical Decision Making. Dr. Lewis was supported by a K07 Mentored Career Development Award (5K07CA10412 8) from the National Cancer Institute. Dr. Sheridan was supported by the National Heart Lung and Blood Institute Career Development Award (1K23HL074375-01A1). Revision accepted for publication 13 August 2008.

MEDICAL DECISION MAKING/MAR–APR 2009

Address correspondence to Jennifer M. Griffith, DrPH, MPH, School of Rural Public Health, Texas A&M Health Science Center, MS 1266, College Station, TX 77843-1266; e-mail: [email protected]. DOI: 10.1177/0272989X08327492

167

GRIFFITH AND OTHERS

or ‘‘attributes.’’ The characteristics, or ‘‘levels,’’ used to describe the attributes are varied in multiple pairwise comparisons or rating and ranking exercises. Responses are then used to determine the relative importance of the attributes.5 Conjoint analysis has been applied to eliciting values for attributes of health care treatments and services, including HIV testing,6,7 miscarriage management,8 and treatment of osteoporosis.9 Because conjoint analysis asks individuals to make multiple comparisons among hypothetical treatments or services with different attribute levels, responses may be affected by how information about the attributes and their levels is presented. One important consideration in a conjoint analysis study is whether the format of quantitative information about treatment risks and benefits affects choices made by individuals. As such, a limitation of conjoint analysis is that the interpretation of the values of both the attributes and levels depends on the ranges presented.10 Studies have shown that how risk reduction information is presented affects decision making by influencing an individual’s understanding and intuitive perception of risk.11–30 Two formats are predominantly used to convey risk reduction associated with health-related treatments or decisions: 1) relative risk reduction (RRR) and 2) absolute risk reduction (ARR). Studies suggest that RRR presentations may be more persuasive in health-related decision making, resulting in higher perception of risk, increased intent for risk reduction, and greater acceptance of risk-reducing treatments.12–15,17,20,22–28,30 If this is the case, it is possible that values elicited through conjoint analysis could be particularly susceptible to magnifying this effect. No studies to our knowledge have evaluated whether the presentation of risk reduction information in different formats affects the value placed on treatment attributes during values elicitation using conjoint analysis or other similar techniques. The goal of this study was to evaluate whether presentation of risk reduction information in RRR or ARR formats affects values for heart disease prevention treatments elicited during conjoint analysis. METHODS We performed a randomized trial of the effect of 2 different formats for presentation of heart disease risk reduction information when using conjoint analysis. 168 • MEDICAL DECISION MAKING/MAR–APR 2009

Study Site and Participants Our study was approved by the Biomedical Institutional Review Board at the University of North Carolina at Chapel Hill (UNC) and was performed at the Decision Support Lab (DSL) in the UNC Sheps Center for Health Services Research. All participants were recruited from our DSL participant registry, which currently has more than 700 individuals enrolled. The registry recruits participants using various media (e-mail lists, TV, print advertising, bus signage, and fliers) and word of mouth from previous study participants and registry members. For this study, we used the registry to recruit participants via their preferred method of contact, either email or letter, who were between the ages of 30 and 75, who had not had a heart attack or stroke, and who did not have congestive heart failure. Those who agreed to participate were asked to come to the DSL for the study session. Participants’ study sessions ranged from 30 minutes to 1 hour in duration, with most approximately 45 minutes. All participants who completed the study received $25. Values Elicitation Using Conjoint Analysis We first generated a set of treatment attributes and levels related to heart disease prevention from literature reviews, focus groups, and our previous research.31,32 The attributes included ability of the treatment to reduce the risk of heart attacks, other health benefits of the treatment, how easy or difficult it is to do the treatment, chance of bad reactions from the treatment, and out-of-pocket costs. Levels associated with each of the attributes are provided in Table 1. We determined that the meaningful levels for one attribute—the ability to reduce the risk of heart attacks—could be described using either RRR or ARR formats for the different treatment options. To make the information understandable, we developed different wording for each version (RRR = ability to reduce heart attacks [XX%]; ARR = chance of a heart attack drops from X% to Y%; see Figure 1A, B). The conjoint analysis process requires participants to view a series of pairwise comparisons of hypothetical treatment choices and choose between them. To generate the hypothetical scenarios for our pairwise comparisons, we entered the 5 key attributes and their levels into the Sawtooth Software traditional conjoint analysis software package.33 The software produced a set of 15 pairwise scenarios required for adequate review of all attributes and their levels.3 In addition, we added 2 other scenarios, both with an

RISK REDUCTION FORMAT AND CONJOINT ANALYSIS

Figure 1 (A) Sample screen from relative risk reduction version of values elicitation exercise and (B) sample screen from absolute risk reduction version of values elicitation exercise.

RISK COMMUNICATION

169

GRIFFITH AND OTHERS

Table 1 Attributes and Levels Attribute

Ability to reduce heart attacks Absolute risk reduction Relative risk reduction

Other health benefits How easy or difficult to do

Chance of having a bad reaction

Out-of-pocket costs

Levels

13% to 11% 13% to 9% 13% to 6% 10% 30% 50% None Medium Many Easy Neither easy nor difficult Difficult Very small Small Moderate $100

expected dominant choice, to check whether participants attended to the values elicitation exercise. The software required responses to all 17 pairwise comparisons to generate the overall importance of attributes. An example of a hypothetical scenario is shown in Figure 1. The scenarios are identical, except that Figure 1A shows the RRR version and Figure 1B the ARR version for the attribute ‘‘ability to reduce heart attacks.’’ Study Process Each participant was first presented with the same hypothetical patient scenario for heart disease risk. Participants were asked to imagine that they were 50 years old with hypertension, high cholesterol, and no diabetes, smoking, or family history of coronary heart disease. For a family of 4, they had a yearly family income of $60,000 and were not currently taking any steps to prevent heart disease. They had an untreated 13% baseline chance of having a heart attack over the next 10 years. Using this information and considering their own values, participants were asked to respond to the values elicitation exercise. Each participant completed the same 17 pairwise comparisons. For an individual participant, attributes were presented in the same order for all 17 pairwise comparisons. For each participant session, 170 • MEDICAL DECISION MAKING/MAR–APR 2009

however, the order of attribute presentation was randomized to reduce ordering effects between participants. For example, a participant who saw Figure 1A as his or her first pairwise screen would always see attributes in that order (adoption of treatment, chance of side effects, ability to reduce heart attacks, etc.), but the next participant would have another ordering (e.g., ability to reduce heart attack, other health benefits, out-of-pocket costs, etc.). To address our study question about the effect of risk reduction format, we randomized participants to view either the RRR or ARR version of the attribute ‘‘ability to reduce heart attacks.’’ Randomization was achieved using a random-number generator used to create sequential, sealed, opaque assignment envelopes. Each envelope was opened at the beginning of a participant’s session by the research assistant, and the research assistant started the appropriate version of the program to be completed by the participant. Measures For this analysis, we chose to use the ‘‘most important attribute’’ as a proxy for participant values. We evaluated 2 primary outcome measures: 1) participant-selected most important attribute and 2) conjoint-generated most important attribute. The participant-selected most important attribute was chosen directly by the participant upon completion of the values elicitation exercise but prior to receiving the results of the conjoint analysis. We used Sawtooth traditional conjoint analysis software with ordinary least squares (OLS) regression to calculate attribute importance or ‘‘utilities’’ and identify the most important attribute for each participant. OLS regression uses a recoding strategy that determines the presence or absence of an attribute level in a pairwise comparison and provides a range of ‘‘utilities’’ based on the levels associated with each attribute. Utility, in the context of conjoint analysis, is a numerical representation of the difference between the best and worst levels for that attribute based on participants’ responses to the pairwise comparisons. The most important attribute was determined by dividing the range for each attribute by the sum of the ranges for all attributes and multiplying by 100.33,34 The attribute with the highest importance (expressed as a percentage) was coded as the most important attribute. Analyses Fisher’s exact tests were used to compare the effect of risk reduction format for the participant-selected

RISK REDUCTION FORMAT AND CONJOINT ANALYSIS

Table 2 Demographics by Group (n = 113)

Age, mean+SD Male, % White, % African American, % College graduate or more, %

Absolute Risk Reduction Group (n = 55)

Relative Risk Reduction Group (n = 58)

50+11 24 57 38 51

52+11 35 50 45 57

and conjoint-generated most important attribute. We also assessed whether race, age, gender, and education level influenced the effect of risk format. We used logistic regression and evaluated the Wald statistic’s P values to test for significant interaction effects.

Table 3 Proportion of Participants Selecting Each Attribute as Most Important, by Risk Reduction Format (in Percentages)

Ability to reduce heart attacks Other health benefits How easy or difficult to do Chance of having a bad reaction Out-of-pocket costs

Absolute Risk Reduction Group (n = 55)

Relative Risk Reduction Group (n = 58)

53 11 15 15 7

64a 0 16 7 14

a. Fisher’s exact P = 0.26; proportion of those participants in each group who selected ability to reduce heart attacks as most important by risk reduction format.

Interactions With Risk Presentation Format RESULTS We identified and contacted 390 potentially eligible subjects from our participant registry between September 2006 and March 2007. Of the 148 who responded to the study announcement, 142 were eligible and scheduled to participate; nonrespondents were slightly younger and included more women, fewer whites, and more African Americans. Of the 142, 113 (80%) attended their study session and were randomized and completed the study: 55 in the ARR group and 58 in the RRR group. Demographics are shown in Table 2. Effect of Risk Presentation Format on Participant-Selected Most Important Attribute The attribute ‘‘ability to reduce heart attacks’’ was selected most often as the most important treatment attribute in both ARR and RRR groups. The proportion of participants who selected this attribute was slightly higher in the RRR group compared with the ARR group, but this difference was not statistically significant (64% RRR v. 53% ARR, Fisher’s exact test P = 0.26; Table 3). Effect of Risk Presentation Format on Conjoint-Generated Most Important Attribute Those in the RRR group were significantly more likely to have a conjoint-generated most important attribute of ability to reduce heart attacks than those in the ARR group (59% RRR v. 33% ARR, Fisher’s exact test P < 0.01; Table 4).

RISK COMMUNICATION

In stratified analyses, we assessed whether 4 demographic variables—race, gender, education level, and age—influenced the effect of risk reduction format on the odds that the participant-selected and conjointgenerated most important attribute was the ability to reduce heart attacks. For the participant-selected most important attribute, race, education level, and age showed no interaction with risk reduction format, with the most important attribute being the ability to reduce heart attacks. However, we did find an interaction between gender and the risk reduction format’s effect. Men were more likely to be influenced by risk reduction format: 85% selected ability to reduce heart attacks as most important with RRR v. 39% with ARR, but there was no difference for women (53% RRR v. 57% ARR; P = 0.01 for interaction). For the conjoint-generated most important attribute, we did not find any significant interactions. Both men and women were more likely to generate ability to reduce heart attacks as the most important attribute when provided with RRR as compared with ARR (men: 70% RRR v. 46% ARR; women: 53% RRR v. 31% ARR). DISCUSSION The presentation of treatment benefit information in absolute or relative terms affected participants’ assessment of the importance of this attribute for decision making about heart disease prevention. Participants were more likely to generate, through the conjoint process, the ability to reduce heart

171

GRIFFITH AND OTHERS

Table 4 Proportion of Participants Generating Each Attribute as Most Important, by Risk Reduction Format (in Percentages)

Ability to reduce heart attacks Other health benefits How easy or difficult to do Chance of having a bad reaction Out-of-pocket costs

Absolute Risk Reduction Group (n = 55)

Relative Risk Reduction Group (n = 58)

33 7 26 16 18

59a 3 12 14 12

a. Fisher’s exact P < 0.01; proportion of those participants in each group who generated ability to reduce heart attacks as most important by risk reduction format.

attacks as the most important attribute when they received RRR compared with ARR. Our finding that those in the RRR group were more likely to generate the ability to reduce heart attacks as the most important attribute is consistent with previous work that found the RRR format to be more persuasive than ARR.12,20,22,25,27 This has important implications for values elicitation using conjoint analysis. Because participants in our study were provided with multiple pairwise comparisons using all 5 attributes, the effect of repeatedly receiving the RRR format compared with the ARR format may have been magnified. Our findings suggest that the persuasiveness of the RRR format could be the reason behind a higher number of participants selecting and generating ability to reduce heart attacks as the most important attribute and the higher level of consistency in the RRR group. How best to present this information is unclear. One option would be to present the RRR format in conjunction with ARR when using conjoint analysis with pairwise comparisons to elicit values. Using both formats of information when eliciting values in conjoint analysis may reduce the potential of a single format having a large impact on assessed values, which could lead to suboptimal decisions. However, presenting both formats could be confusing and cause information overload; this bears further study. Our findings also suggest that men and women may be affected differently by risk reduction presentations for heart disease prevention. Men were influenced by the risk reduction format when selecting their most important attribute, but women were not. We did not observe a difference by gender for the 172 • MEDICAL DECISION MAKING/MAR–APR 2009

conjoint-generated most important attribute; men and women both were affected in a similar manner. The explanation for these findings is not clear from our data. When selecting their most important attribute, men may have been more likely than women to focus on the large numbers used in the RRR format and thus be more likely to see it as most important. Research conducted in marketing suggests that men are more likely to be ‘‘selective processors’’ and thus may not engage in detailed elaboration of message content but rely on heuristics (such as defaulting to the large number) instead.35–38 However, this theory would not be consistent with the finding for the conjoint-generated most important attribute, as both men and women were affected similarly. Further research is required to examine this effect in different decision tasks, but it poses an important question about how information is provided and processed by men and women when making health decisions. When interpreting our study results, it is important to highlight several limitations. First, we did not assess the participant-selected most important attribute prior to the values elicitation exercise using conjoint analysis. We made this choice to prevent participants from focusing on their selected attribute during the exercise, but doing so negated our ability to measure a change from before to after the exercise. Second, all 5 attributes were presented simultaneously for each of the 17 pairwise comparisons. This may have overwhelmed participants, resulting in a simplification of information processing from focusing on only one or a few of the attributes rather than all 5, possibly exaggerating the effect of risk reduction format. Another method of conjoint analysis, adaptive conjoint, presents only a few of the attributes at a time rather than all attributes simultaneously and could reduce the chance that choices of participants are driven by focusing on one attribute.3 This should be tested in future work. Third, we used a hypothetical scenario for heart disease prevention with lab study participants, potentially making the decision less salient by reducing engagement. It is possible that our results would differ when patients are making decisions using their own risk information. Finally, the wording used in the 2 versions of the program for the risk reduction attributes was different. This difference is implicit in the risk reduction format and is needed to make the information understandable. However, the wording difference may be an important part of the observed effect as opposed to the numbers themselves. We did not measure how well participants

RISK REDUCTION FORMAT AND CONJOINT ANALYSIS

comprehended the information, and future work should do so. On the basis of our findings and the limitations, we can only make preliminary conclusions regarding risk reduction format and use of conjoint analysis for values elicitation related to heart disease prevention. Our conclusions, however, are consistent with other literature suggesting that the RRR format is more persuasive in health-related decision making12,20,22,25,27 and that men and women may process the information differently,35–38 giving validity to our findings. When presented with RRR, which feels large, participants may have downplayed other attributes, such as ease of adoption, costs, and chance of bad reactions, which may be just as or more important to them. Whether using RRR and ARR formats together or ARR alone is a better approach is unclear and requires further study to elicit patients’ underlying values by more than one means, allowing comparison between methods.

9. Fraenkel L, Gulanski B, Wittink D. Patient treatment preferences for osteoporosis. Arthritis Rheum. 2006;55:729–35.

ACKNOWLEDGMENTS

17. Forrow L, Taylor WC, Arnold RM. Absolutely relative: how research results are summarized can affect treatment decisions. Am J Med. 1992;92:121–4. 18. Gigerenzer G. Why does framing influence judgment? J Gen Intern Med. 2003;18:960–1.

We thank Shrikant I. Bangdiwala, PhD, for his assistance with the statistical analysis in this article. In addition, we thank our research assistants Destiney S. Nettles, Alison R. Tytell Brenner, and Chris DeLeon for their assistance with data collection.

REFERENCES 1. Elwyn G, O’Connor A, Stacey D, et al. Developing a quality criteria framework for patient decision aids: online international Delphi consensus process. BMJ. 2006;333:417. 2. Green P, Rao V. Conjoint measurement for quantifying judgmental data. J Marketing Res. 1971;8:355–63. 3. Orme B. Which Conjoint Method Should I Use? Sawtooth Software Research Paper Series. Sequin, WA: Sawtooth Software; 2003. 4. Orme B. Getting Started with Conjoint Analysis: Strategies for Product Design and Pricing Research. Madison, WI: Research Publishers LLC; 2006. 5. Sawtooth Software. Sawtooth Software: Technical Paper Series Conjoint Value Analysis (CVA), Version 3.0. Sequim, WA: Sawtooth Software; 2002. 6. Phillips KA, Johnson FR, Maddala T. Measuring what people value: a comparison of ‘‘attitude’’ and ‘‘preference’’ surveys. Health Serv Res. 2002;37:1659–79. 7. Phillips KA, Maddala T, Johnson FR. Measuring preferences for health care interventions using conjoint analysis: an application to HIV testing. Health Serv Res. 2002;37:1681–705. 8. Ryan M, Hughes J. Using conjoint analysis to assess women’s preferences for miscarriage management. Health Econ. 1997;6:261–73.

RISK COMMUNICATION

10. Amaya-Amaya M, Gerard K, Ryan M. Discrete choice experiments in a nutshell. In: Ryan M, Gerard K, Amaya-Amaya M, eds. Discrete Choice Experiments to Value Health and Health Care. Dordecht, The Netherlands: Springer; 2008. 11. Bergus GR, Levin IP, Elstein AS. Presenting risks and benefits to patients. J Gen Intern Med. 2002;17:612–7. 12. Berry DC, Knapp P, Raynor T. Expressing medicine side effects: assessing the effectiveness of absolute risk, relative risk, and number needed to harm, and the provision of baseline risk information. Patient Educ Couns. 2006;63:89–96. 13. Bobbio M, Demichelis B, Giustetto G. Completeness of reporting trial results: effect on physicians’ willingness to prescribe. Lancet. 1994;343:1209–11. 14. Bucher HC, Weinbacher M, Gyr K. Influence of method of reporting study results on decision of physicians to prescribe drugs to lower cholesterol concentration. BMJ. 1994;309:761–4. 15. Cranney M, Walley T. Same information, different decisions: the influence of evidence on the management of hypertension in the elderly. Br J Gen Pract. 1996;46:661–3. 16. Edwards A, Evans R, Elwyn G. Manufactured but not imported: new directions for research in shared decision making support and skills. Patient Educ Couns. 2003;50:33–8.

19. Gigerenzer G, Edwards A. Simple tools for understanding risks: from innumeracy to insight. BMJ. 2003;327:741–4. 20. Gyrd-Hansen D, Kristiansen IS, Nexoe J, Nielsen JB. How do individuals apply risk information when choosing among health care interventions? Risk Anal. 2003;23:697–704. 21. Heller RF, Sandars JE, Patterson L, McElduff P. GPs’ and physicians’ interpretation of risks, benefits and diagnostic test results. Fam Pract. 2004;21:155–9. 22. Hembroff LA, Holmes-Rovner M, Wills CE. Treatment decisionmaking and the form of risk communication: results of a factorial survey. BMC Med Inform Decis Making. 2004;4. Available from: http:// www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd = Retrieve& db = PubMed&dopt = Citation&list_uids = 15546488 23. Hux JE, Naylor CD. Communicating the benefits of chronic preventive therapy: does the format of efficacy data determine patients’ acceptance of treatment? Med Decis Making. 1995;15:152–7. 24. Malenka DJ, Baron JA, Johansen S, Wahrenberger JW, Ross JM. The framing effect of relative and absolute risk. J Gen Intern Med. 1993;8:543–8. 25. Misselbrook D, Armstrong D. Patients’ responses to risk information about the benefits of treating hypertension. Br J Gen Pract. 2001;51:276–9. 26. Naylor CD, Chen E, Strauss B. Measured enthusiasm: does the method of reporting trial results alter perceptions of therapeutic effectiveness? Ann Intern Med. 1992;117:916–21. 27. Nexoe J, Gyrd-Hansen D, Kragstrup J, Kristiansen IS, Nielsen JB. Danish GPs’ perception of disease risk and benefit of prevention. Fam Pract. 2002;19:3–6.

173

GRIFFITH AND OTHERS

28. Nikolajevic-Sarunac J, Henry DA, O’Connell DL, Robertson J. Effects of information framing on the intentions of family physicians to prescribe long-term hormone replacement therapy. J Gen Intern Med. 1999;14:591–8.

Available from: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi? cmd = Retrieve&db = PubMed&dopt = Citation&list_uids = 17005051

29. Replogle WH, Johnson WD. Interpretation of absolute measures of disease risk in comparative research. Fam Med. 2007;39: 432–5.

33. Sawtooth Software. CVA System 3.0.4. Sequim, WA: Sawtooth Software; 2006. 34. Fraenkel L, Gulanski B, Wittink DR. Preference for hip protectors among older adults at high risk for osteoporotic fractures. J Rheumatol. 2006;33:2064–8.

30. Sheridan SL, Pignone MP, Lewis CL. A randomized comparison of patients’ understanding of number needed to treat and other common risk reduction formats. J Gen Intern Med. 2003;18: 884–92.

35. Meyers-Levy J. Gender difference in information processing: a selectivity interpretation. In: Cafferata P, Tybout A, eds. Cognitive and Affective Responses to Advertising. Lexington, MA: Lexington; 1989. p 219–60.

31. Pignone M, Sheridan SL, Lee YZ, et al. Heart to Heart: a computerized decision aid for assessment of coronary heart disease risk and the impact of risk-reduction interventions for primary prevention. Prev Cardiol. 2004;7:26–33. 32. Sheridan SL, Shadle J, Simpson RJ Jr, Pignone MP. The impact of a decision aid about heart disease prevention on patients’ discussions with their doctor and their plans for prevention: a pilot randomized trial. BMC Health Serv Res. 2006;6.

36. Meyers-Levy J, Maheswaran D. Exploring differences in males’ and females’ processing strategies. J Cons Res. 1991;18:63–70.

174 • MEDICAL DECISION MAKING/MAR–APR 2009

37. Meyers-Levy J, Sternthal B. Gender differences in the use of message cues and judgments. J Marketing Res. 1991;28:84–96. 38. Putrevu S. Exploring the origins and information processing difference between men and women: implications for advertisers. AMS Rev. 2001;2001. Available from: http://www.amsreview.org/ articles/putrevu10-2001.pdf