You are here

Article Text

Article menu

Download PDFPDF

XML
Qualitative research
Research
The value of personalised risk information: a qualitative study of the perceptions of patients with prostate cancer
  1. Paul K J Han1,2,
  2. Norbert Hootsmans1,
  3. Michael Neilson2,
  4. Bethany Roy2,
  5. Terence Kungel3,
  6. Caitlin Gutheil1,
  7. Michael Diefenbach4,
  8. Moritz Hansen2,5
  1. 1Center for Outcomes Research and Evaluation, Maine Medical Center, Portland, Maine, USA
  2. 2Tufts University School of Medicine, Boston, Massachusetts, USA
  3. 3Maine Coalition to Fight Prostate Cancer, Augusta, Maine, USA
  4. 4Division of Urology, Mount Sinai School of Medicine, New York, New York, USA
  5. 5Genitourinary Cancer Program, Maine Medical Center, Portland, Maine, USA
  1. Correspondence to Dr Paul K J Han; hanp{at}mmc.org

Abstract

Objective To explore the experiences of patients with prostate cancer with risk information and their perceptions of the value of personalised risk information in treatment decisions.

Design A qualitative study was conducted using focus groups. Semistructured interviews explored participants’ experiences with using risk information, and their perceptions of the potential value of personalised risk information produced by clinical prediction models.

Participants English-speaking patients, ages 54–82, diagnosed with prostate cancer within the past 3 years, residing in rural and non-rural geographic locations in Maine (USA), and attending prostate cancer patient support groups.

Setting 6 focus groups were conducted with 27 patients; separate groups were held for patients with low-risk, medium-risk and high-risk disease defined by National Comprehensive Cancer Network guidelines.

Results Several participants reported receiving risk information that was imprecise rather than precise, qualitative rather than quantitative, indirect rather than direct and focused on biomarker values rather than clinical outcomes. Some participants felt that personalised risk information could be useful in helping them make better informed decisions, but expressed scepticism about its value. Many participants favoured decision-making strategies that were heuristic-based and intuitive rather than risk-based and deliberative, and perceived other forms of evidence—emotions, recommendations of trusted physicians, personal narratives—as more reliable and valuable in treatment decisions.

Conclusions Patients with prostate cancer appear to have little experience using personalised risk information, may favour heuristic-based over risk-based decision-making strategies and may perceive personalised risk information as less valuable than other types of evidence. These decision-making approaches and perceptions represent potential barriers to the clinical use of personalised risk information. Overcoming these barriers will require providing patients with greater exposure to risk information, education about the nature and value of personalised risk information and training in deliberative decision-making strategies. More research is needed to confirm these findings and address these needs.

  • QUALITATIVE RESEARCH
  • HEALTH SERVICES ADMINISTRATION & MANAGEMENT
  • ONCOLOGY

This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 3.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/3.0/

https://doi.org/10.1136/bmjopen-2013-003226

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Article summary

    Strengths and limitations of this study

    • The study provides empirical evidence on the value of personalised risk information from the patient perspective, and identifies previously unexamined barriers to effective use of this information.

    • The study sample was relatively small, geographically limited and racially and ethnically homogeneous.

    • Qualitative methods cannot definitively establish the prevalence, causes or effects of patients’ experiences and perceptions. Further research using quantitative methods is needed to confirm the study's findings.

    Introduction

    Personalised, or individualised, risk information—information about the probability of future health outcomes for individual patients1—is an increasingly common and important form of clinical evidence. In recent years, burgeoning efforts to develop clinical prediction models (CPMs)—statistical algorithms that utilise characteristics of the patient, disease and treatment to estimate individualised probabilities of health outcomes2—have increased the supply of personalised risk information. Meanwhile, a growing emphasis on the ideals of personalised healthcare, patient-centred outcomes research, and informed and shared decision making have heightened clinical demand for this information.3–8 Personalised risk information advances each of these important healthcare ideals, enabling decision-making based on the expected outcomes of individuals rather than groups, on prognostic estimates rather than diagnostic categories,9 and on patient values and preferences.2 Emerging evidence supports this vision; a recent Cochrane review concluded that personalised risk information promotes informed patient decision-making in cancer screening.10

    Nevertheless, the clinical value of personalised risk information is limited by several barriers, including the conceptually abstract nature of risk information,11–14 psychological biases and well-documented deficits in numeracy that impede its comprehension by both patients and health professionals.15–22 Correspondingly, several studies have shown that patients’ understanding of personalised risk estimates produced by CPMs is poor.21–29 A more serious problem, however, is that precise, quantitative risk information may not be what patients really want or need. Zikmund-Fisher30 has argued that such information is not always informative, and ‘simpler, less precise representations’ of risk are often more useful to patients. In a similar vein, Reyna31 has contended that people prefer to construe risk in qualitative terms representing its ‘gist’ meaning, rather than in precise, quantitative terms representing its ‘verbatim’ details. Gigerenzer has gone further to challenge the conception of rationality underlying efforts to apply risk information to decision making. Extending an argument put forward by the economist Simon in the 1950s,32 he emphasises that rationality is ‘bounded’ by limitations in the cognitive and environmental resources available to decision makers. Consequently, in real-world decisions people do not ‘optimize’—that is, they do not calculate and weigh probabilities and values in an exhaustive computational search for the best option. Instead, they use heuristics—‘fast and frugal’ rules of thumb that facilitate adaptive decisions.

    These insights raise fundamental questions about the usefulness of precise, quantitative risk information in healthcare. Is such information really needed and desired by patients? Do patients—as opposed to health professionals—perceive personalised risk information as valuable, and how might their perceptions influence the success of efforts to use CPMs in clinical practice?

    The aim of the current research was to explore these questions, focusing on personalised risk information in the treatment of prostate cancer—the most common and second most common male cancer in the USA and worldwide, respectively, and the second and sixth leading cause of male cancer deaths.33–35 Approximately 80% of newly diagnosed patients with prostate cancer have clinically localised disease for which there are multiple treatment options—surgery, radiation therapy, conservative treatment (active surveillance)—each with differing potential benefits and harms.34 The same is true for the treatment of more advanced, higher-risk disease. Prostate cancer has thus become a primary focus of predictive modelling activities, resulting in the development of more than 100 CPMs36–42 and growing efforts to disseminate and apply these models in clinical practice.41–46

    Little is known, however, about patients' perspectives on the value of CPMs and the information they produce. To this end, the current study explored experiences of patients with prostate cancer regarding the use of risk information in treatment decisions, and their perceptions of the value of precise, personalised risk information produced by CPMs. The ultimate goal was to generate insights that might inform efforts to use CPMs to promote personalised, informed decision-making in the treatment of prostate cancer and other conditions.

    Methods

    Study design, participants and data collection

    This qualitative study employed focus groups, a useful methodology for exploring people's perceptions, beliefs and attitudes.47 ,48 From September 2010 to February 2011, seven focus groups were conducted with 27 patients with prostate cancer in the state of Maine (3–6 patients/group). Participants were members of the Maine Coalition to Fight Prostate Cancer (MCFPC), a patient advocacy organisation that administers nine statewide support groups. Eligible participants were recruited by MCFPC and consisted of English-speaking prostate cancer survivors within 3 years of initial diagnosis. Purposive recruitment soliciting participation of men at various disease stages was conducted to obtain a study sample with diverse treatment experiences and prognoses; no volunteers were excluded. To achieve sufficient within-group homogeneity to encourage open discussion,49 participants were stratified into three groups (box 1) according to their risk for recurrent disease (low, medium, high), as defined by US National Comprehensive Cancer Network guidelines.50 A total of two low risk, three medium risk and two high risk focus groups were conducted at MCFPC support group sites in three rural and non-rural towns and cities (Brunswick, Lewiston, Portland). Participants received $50 compensation. Sessions lasted approximately 2 h, were audio-recorded with prior consent of participants, and transcribed verbatim by a professional transcription service.

    Box 1

    Focus group categories

    • Low-risk group: Early-stage, clinically localised disease (Gleason Score ≤6, stage T1C-T2A, PSA <10)

    • Medium-risk group: Intermediate-stage, clinically localised disease (Gleason Score 7, stage T2B-T2C, PSA 10–20)

    • High-risk group: patients with advanced, treatment-refractory or recurrent disease (Gleason Score 8–10, stage T3-T4, PSA >20)

    Interview content

    Groups were led by PH and either a professional moderator or another study investigator (TK). Interviews were semistructured and followed a moderator guide consisting of open-ended questions and close-ended probes designed to elicit patients’ past experiences with risk information, preferences for personalised risk information and attitudes towards CPMs. To illustrate how CPMs work and the types of information they provide, participants were briefly shown visual aids consisting of screenshots of web-based CPMs (appendix A) for different prostate cancer outcomes (eg, risk of cancer recurrence, cancer-free survival, death) produced by the Cleveland Clinic (http://www.lerner.ccf.org/qhs/risk_calculator/index.php)42 and Memorial Sloan-Kettering Cancer Center (http://www.mskcc.org/mskcc/html/10088.cfm).51 During the course of the study, minor revisions were made in the interview guide to clarify emergent themes.

    Data analysis

    In-depth analysis and line-by-line software-assisted coding of anonymised interview transcripts were conducted using the programme NVivo (V.8; QSR International). First, three investigators (PH, MN and BR) developed a preliminary conceptual schema and codebook by independently reading three transcripts, categorising participants’ verbatim statements according to thematic content, and organising emergent themes—consistent with a ‘grounded theory’ approach to the data.52 ,53 Preliminary codebooks were reviewed by the team and areas of disagreement were resolved through further discussion. A single working codebook was then produced, which two investigators (MN, BR) used to code the remaining transcripts, employing a ‘constant comparative’ method to compare new data, concepts and themes with ones previously identified.52 ,54 The research team held periodic meetings to discuss coding decisions, identify new themes and resolve areas of disagreement. Finally, two investigators (PH and NH) conducted a secondary review of all coded text to organise dominant themes. An outline of identified themes is provided in appendix B. The current analysis focused on participants’ perceptions of the value of personalised risk information; analyses of other themes will be reported separately.

    Results

    Characteristics of participants are summarised in table 1. The primary interview domains consisted of participants’ (1) past experiences with risk communication, and (2) perceptions of the value of personalised risk information produced by CPMs.

    View this table:
    Table 1

    Characteristics of focus group participants

    Past experiences with risk communication

    Within these two broad domains several dominant themes emerged; the first was the broad spectrum of risk communication experiences and practices reported by participants.

    Absence of explicit risk communication

    At one end of the spectrum were a small number of participants who reported having never received risk information of any kind during the decision-making process: I wasn't actually told about the numbers. I just went by what they were saying. I guess—they knew what the numbers were, what the odds were for me, and I went with that and yeah, I would've wanted to know what the numbers were. [Medium-Risk Participant]

    For these few participants treatment risks were implicitly understood rather than explicitly communicated: I think the conversation that I had with my urologist was we each knew all of those possibilities. It was kind of like a given, and our conversation was based on the fact that we each knew those facts, so they weren't really mentioned specifically. It was just, you've got it, here are your options to deal with it because you want to deal with it because it could be fatal, it could spread. [Low-Risk Participant]

    Quantitative risk communication

    At the other end of the spectrum were just two participants, among all groups, who reported receiving precise, detailed, quantitative risk information. One low-risk and another medium-risk participant each recalled being shown a ‘histogram’ and other visual aids detailing different risk levels for different treatment outcomes, broken down by demographic characteristics. Neither participant, however, recalled their risk estimates.

    More commonly, quantitative risk information was communicated in imprecise terms. For example, several participants recalled being provided with ‘average’ risk estimates applicable to patient subpopulations—for example, stratified by cancer stage—while several others reported receiving a range rather than point estimate of risk.

    Categorical risk communication

    The more commonly reported mode of risk communication was categorical and non-quantitative. A low-risk participated noted: ‘he didn't say anything about the likelihood of spreading, but, you know, just, what he told me on the phone it was a low, that it was low.’ Other participants recalled the use of similarly broad categories—for example, ‘good,’ ‘most likely,’ ‘very remote,’ ‘high,’ ‘very likely.’

    Indirect risk communication: the biomarker heuristic

    The most commonly reported mode of risk communication by far, however, was indirect, through the use of biomarker information—primarily prostate specific antigen (PSA) values or Gleason scores (box 2). Participants in all groups described using biomarker values in lieu of probability estimates to understand the magnitude of their risk: I really, you know, I read the books … And, I'm sure they had all sorts of graphs and stuff like that. But I've got to admit when I looked at my Gleason scores, this is gonna sound, you know it's all irrelevant. [High-Risk Participant]

    Box 2

    Biomarker heuristic: illustrative quotes

    I was told that with the Gleason score of 9 that that's frightening … That's extremely aggressive is what this one doctor said, and at the next place that doctor said it's kind of scary, he says, but it can be taken care of. [High-Risk Participant]

    INTERVIEWER: And what was your understanding of what the Gleason score meant?

    PATIENT: A high number is, you know—the top number is, you know, if that's higher than your bottom number, you've got problems. [High-Risk Participant]

    … he gave me the lab report itself and he showed me these different things and he says this here's the Gleason score. The Gleason is this three over four. He said if it had been four over three, you know, we would have been more concerned. He said this shows it's an active cancer but it's not an aggressive cancer and he said so we've got some time to wait. We don't have a lot of time to wait. He said we don't have to worry about going into surgery in the next three weeks or something. [Medium-Risk Participant]

    The way they explained to me was that if it's under a four Gleason it's probably not gonna do much. If it's around a six it's kind of getting up there. If it gets up around and eight then it's real aggressive, you gotta do something fast. [Medium-Risk Participant]

    This ‘biomarker heuristic’—which has not been previously described as such, to our knowledge—functioned as a shorthand rubric or guide to treatment decisions. As one medium-risk participant articulated, referring to his PSA values: “when you don't have any guidelines to go by, you gotta have some idea where that is, whether it's a four or six or ten or whatever.” The use of biomarkers was a heuristic strategy that obviated the need to deal directly with outcome probabilities per se: INTERVIEWER: So they gave you a sense of what those numbers meant … a four means you have an X percent chance of— PATIENT 3: No, they didn't. INTERVIEWER: But they just said—it was good—versus bad, right? PATIENT 3: Yeah. PATIENT 2: That's what I got out of it when I was told I had—the PSA was 9 I think and my Gleason was 7—and like everybody else, you know, you're just completely naïve about it, but, by explaining then, telling me that, well, the 9, you better do something about it, you know—I mean, it's not a necessity, but they didn't recommend not doing anything about it all … So those numbers I think … kind of helps you towards whichever goal you wanna go to. [High-Risk Participant]

    The usefulness of biomarker values appeared to stem from their apparently ‘hard,’ tangible nature, and their straightforward connection to a defined course of action: PATIENT: My decision was based upon what we had for biopsies and stuff like that. INTERVIEWER: Okay. Alright. PATIENT: Not look, you know, a lot of people get hit crossing the street. And, but that doesn't’ really—that's irrelevant to me … but, based upon the information that I had, you know, hard information, I felt very comfortable.

    The influence of biomarker information contrasted with participants’ lack of understanding of its precise meaning: I really wasn't sure, and to this day, I don't know what the heck a Gleason score is. I'm not certain exactly what the PSA score is and what that is supposed to indicate, but I do know I had several PSA tests, and they started at 17, and there were several in 13, 14 and then a couple with 9 and 10, and then I had one that was 7. [High-Risk Participant]

    Perceptions of the value of personalised risk information

    Patients’ perceptions of the potential value of personalised risk information produced by CPMs, spanned a broad range.

    Enabling informed decision making

    On the one hand, participants acknowledged several potential benefits related to the ideal of informed decision-making. Some participants noted that personalised risk information could help clarify trade-offs involved with alternative treatment options: You take a look at that and say the statistics say, wow, you know, I'm not anywhere near as good if I do that over there. I think it would help you with your decision. [High-Risk Participant]

    Patient empowerment vis-a-vis their physicians was another perceived value of personalised risk information: So this is like a whole new concept where, you know, we're empowered, each of us is empowered with making the decision, and that's, like, you think about it and that's pretty unique and wonderful rather than to have a doctor…you know, the doctor can make a mistake, too … So, like, the doctor is not the god. [Low-Risk Participant]

    Participants believed that such empowerment was necessary because of inherent biases of treating physicians: I can see this being used to help people make a decision as to what course of treatment they should seek versus what side effect they can live with. Now, if you go to X and ask them what their percentage of incontinence is … they're gonna give you a skewed number because they don't want to make that million-dollar piece of equipment look bad. [High-Risk Participant]

    Some participants—particularly those who had chosen active treatment (surgery, radiation therapy) over active surveillance or watchful waiting—felt that personalised risk information produced by CPMs might have changed their treatment decisions: PATIENT 2: I mean, if you could plug in these numbers to see what the prediction would be based on what my diagnosis numbers were doing the various procedures, I, I probably would have looked at all those—and that may have strongly influenced me to do something other than what I did … in my case I think if I had a predictor like this which showed me the same result with watchful waiting or robot or radiation, I would, I would think real strongly about watchful waiting. [Medium-Risk Participant]

    Even if personalised risk information did not influence patients’ actual decisions, furthermore, many valued such information as a means of simply being informed: INTERVIEWER: Imagining that these kinds of tools were available when you were going through decision making, do you think this kind of information would have changed the decisions that you actually made about your own treatment? MULTIPLE PATIENTS (in unison): No. INTERVIEWER: Then what good is it? PATIENT 4: Like D. said, it's just a tool for you to go by, you know. PATIENT 5: Any information you can get adds to your knowledge of where you're gonna go with your treatment. [Medium-Risk Participants]

    Scepticism about the relative value of personalised risk information

    Most participants, however, expressed scepticism about the value of personalised risk information. A primary source was the fundamental uncertainty inherent to all risk estimates. Several participants specifically mentioned the uncertainty involved in applying probability estimates to the realm of single events experienced by individuals:14 You know, I don't care how much information you have, or how much the odds are, it can be thrown a monkey wrench. The Patriots are a good example. [Laughter] I mean, you looked at all the stats and it said they were gonna win. That ain't how it turned out! [Medium-Risk Participant] You don't know which percentage you're going to be in. I mean … the five-year number is 95%. Five percent of the people, their cancer's going to spread. So—I mean, it's a very high chance you won't spread, but you could be in the five percent. [Medium-Risk Participant]

    Another source of scepticism was related to the reliability of statistical models and risk evidence itself. Several participants raised questions about the authenticity of CPMs, the qualifications and experience of CPM producers, and the novel nature of CPMs. They also felt that different researchers and models should be ‘close to giving you the same type of prediction,’ and that conflicting estimates would lower their trust in CPMs. These concerns reflected ‘epistemic uncertainty’ or the consciousness of what has been termed ‘ambiguity’—limitations in the reliability, credibility or adequacy of risk information.14

    Participants’ uncertainties were manifest in scepticism about the value and influence of personalised risk information in decision-making. Participants reported that at the most, they would use CPMs as a ‘second opinion’ or ‘adjunct’ that would augment—but not take precedence over—other types of evidence. As a medium-risk participant stated, “I would consider the numbers very important, but it wouldn't drive the whole decision process.”

    Risk information versus other types of evidence

    More fundamentally, many participants did not accord privileged status to personalised risk information, but instead viewed it as merely one of several types of evidence—with at least equal legitimacy and weight in the decision-making process: Oh, no. I don't—I don't think it's bullshit, but I do think it is just one of the factors that you would use to make a decision. I agree with the man. But I think that, you know that there's more than just emotion too. I think there's, you know, statistical data and information, there is emotional consideration, there is life—your spouse's circumstance too. [Medium-Risk Participant] I think anything that you get anywhere is a guide … Doesn't matter whether it comes from your doctor, the Internet—support group, or anything. Even prayer is a guide. And you got to take that information that you can get, as much as you can get, and make the best decision for you and your family. [Medium-Risk Participant]

    Emotions were one prominent form of non-statistical evidence that participants utilised in decision-making, manifesting use of the affect heuristic:55 ,56 I wouldn't choose from that. Well certainly if it were me, I would review what are the side effects of each, and the one thing that hasn't been touched on, and you may not be able to, is the emotional part of the decision. The statistics for most people probably wouldn't be more than 50% of the decision-making process. [Medium-Risk Participant]

    For other participants the recommendations of trusted physicians constituted the primary form of evidence, manifesting reliance on what Wegwarth and Gigerenzer57 have termed the trust-the-doctor heuristic: I'm a person who, who really respects and honors education … doctors work damn hard to get a degree and become a medical professional. I have to respect them and their decisions. [High-Risk Participant]

    Personal narrative was another form of evidence that participants prioritised over statistical evidence.58–60 As one participant articulated: “To me, you know, we can listen to the statistics from the physicians but it's also nice to hear from the patients and maybe even the wives.” For some participants, the use of narrative evidence appeared to function as a heuristic—a mental shortcut that obviated the need to rely on statistical evidence. Reliance on this ‘narrative heuristic’—well described in the literature although not typically characterised as a heuristic—manifested a greater trust in the reported experiences of identifiable individuals than in statistical information: Trust is experience-based. The experience happens here with people who are vouching and people who have had the experience … It doesn't happen in a marketplace. [Medium-Risk Participant] So I think there's two ways to look at it. I'm not sure, you know, perhaps how does this 67% get generated? Is it based on just numbers or is it also based on doctors and nurses who have actually face-to-faced and worked, you know, and stuff. [Low-Risk Participant]

    Need for simplicity

    Several participants expressed an overarching perception that risk information was not only potentially unimportant but detrimental in introducing complexity—another major source of uncertainty14 that could obscure the pertinent issues in decision-making: It can cloud the issue for an untrained person. Confuse the situation. Maybe more information than you really need to make a good decision. And you get so boggled with all the information that you lose sight of the forest. [Medium-Risk Participant] But the die was cast and I made my choice. I was going to go with it, and I had that much faith. And the robotic surgery and how really, really, really good it is as opposed to the other types of surgery … I had that much faith in my choice. And sometimes I don't want to say, gee, I don't want the issues to get clouded by the numbers, but I felt comfortable in what I was doing. [Medium-Risk Participant]

    Yet, this perception of personalised risk information as obscuring rather than aiding decisions, along with patients’ reported reliance on various heuristics (biomarker, affect, white-coat, narrative), reflected a more fundamental need for simplicity in information and the decision-making process.

    For many patients, the need for simplicity was ultimately manifest in the choice of active prostate cancer treatment (ie, surgery, radiation therapy): “what's simpler than if you have cancer here, what's simpler than taking it out?” This cognitive strategy exemplified the use of what could be characterised as a simplifying ‘disease heuristic’: a conceptual equation of the mere diagnosis of disease with the necessity of active treatment: PATIENT 2: There was no watchful waiting for me … Whether he told me that or not, I just— INTERVIEWER: There was no option. PATIENT 2: —no. PATIENT 1: I, I had the same. I didn't—that wasn't a decision. [Medium-Risk Participants] And, and you know, when I got those numbers of, you know, the, the biopsies and what not … I wanted to be done with it … I wanted it gone … Let's just rip it out, throw it in the trash can over there in the corner and be done with it. [Medium-Risk Participant]

    Although the conceptual equation of disease diagnosis with treatment necessity has been previously recognised,9 it has not been characterised as a decision-making heuristic. However, our data suggest that it has a heuristic function—obviating patients’ need and desire to attend to outcome probabilities and leading many patients to perceive no real choice options: I, I didn't know these types of models existed. I don't know if I would have used them or not—I might have. Again, I, it was so obvious to take the prostate out, you know—get rid of the problem, but I didn't really care what the percentages were. [Medium-Risk Participant] PATIENT 1: I had to obviously make a decision, but it was very, for me it was such an easy decision. PATIENT 3: Yeah. PATIENT 2: A no-brainer. PATIENT 1: That may have been naïve, but, I guess, I don't know, it just seemed so logical. So I didn't even think twice about—I didn't even think once about weighing options or risks or … I never wanted to get into a discussion with my doctor about all that stuff, because it just wasn't, wasn't necessary. [Medium-Risk Participants]

    Discussion

    Our study provides preliminary insights on an important but understudied issue in the use of CPMs to advance personalised healthcare: the value of personalised risk information from the patient perspective. The study first demonstrates that such information has significant potential value, since the risk information at least some patients currently receive is typically imprecise rather than precise, qualitative rather than quantitative, indirect (biomarker-based) rather than direct (outcomes-based), diagnostic rather than prognostic. To our knowledge, the current study provides the first empirical documentation of these phenomena in prostate cancer care; however, they are not unique to this disease. Indirect risk communication through biomarker values, for example, occurs in the treatment of common conditions such as hypercholesterolaemia and hypertension,9 and heuristics are employed in numerous types of decisions.61 ,62 What the current study newly brings to light, however, is the critical role of patient values in reinforcing these processes. Although many prostate cancer patients in our study clearly perceived personalised risk information as valuable, they viewed other types of evidence as relatively more influential and expressed preferences for intuitive, heuristic-based rather than deliberative, risk-based approaches to decision-making.

    These findings should be interpreted cautiously, however, given several study limitations. The sample was relatively small, geographically limited and homogeneous in race and ethnicity. Many participants had already made treatment decisions, and their negative attitudes towards personalised risk information may thus have been biased by a motivation to avoid regret or dissonance over not having used such information. Participants’ negative attitudes could also have been influenced by the particular ways in which the information was represented by the websites shown in the interviews. It is conceivable that alternative representational methods—for example, using visualisations or other patient-centred risk communication strategies aimed at improving comprehension—may have encouraged more favourable attitudes. This possibility remains to be explored and is an important focus for future research. Finally, the study assessed patient perceptions rather than behaviours and used qualitative methods, a powerful approach for ascertaining the nature and breadth—but not the prevalence, causes and effects—of people's beliefs and attitudes. Further quantitative studies using larger, racially and ethnically diverse samples are thus needed to confirm our preliminary findings. Nevertheless, the validity of these findings is supported by their consistency with mounting evidence that precise, quantitative risk information is not always what patients need or want.30 ,31 Our findings have important implications for potential users of CPMs because they identify patient perceptions of the value of personalised risk information as a barrier to the use of CPMs, and suggest potential modifiable reasons for these perceptions.

    The most straightforward reason is lack of exposure to personalised risk information; most study participants reported having never received such information, and its value was thus simply beyond their imagination. This problem might be addressed by exposing patients to what they have been missing; as Steve Jobs famously quipped, “People don’t know what they want until you show it to them.”63 Yet, our study suggests that even if patients were provided with personalised risk information, they may not always want it. Study participants expressed an overarching desire for simplicity in information and decision-making approach, manifested in preferences for heuristic—rather than risk-based decision-making. This desire for simplicity is common,30 ,64–66 and reflects fundamental limitations in human cognitive capacities (eg, memory, literacy and numeracy) and available environmental resources (eg, time and decision support)32 ,67 ,68 that constrain people's ability to engage in effortful deliberation in decision-making, and instead promote intuitive, heuristic-based decision-making based on factors other than ‘estimations of probabilities, gains, costs and the like.’61 Intuitive and deliberative decision-making approaches have advantages and disadvantages;69 however, if the goal is to increase the perceived value and clinical use of personalised risk information, then the limitations that reduce people's capacity for deliberation must somehow be overcome.

    Another factor limiting the perceived value and use of personalised risk information among study participants was a perception that such information is less reliable than non-quantitative evidence. This perception raises a need that has not been addressed in efforts to apply risk information to patient care: to increase patients’ epistemological understanding—that is, their comprehension of the nature of risk knowledge and the strengths and weaknesses of the evidence at hand. Our study illustrates the twofold challenge patients face in using personalised risk information: they must weigh not only the magnitude of competing probabilities but the value of competing types of evidence, each with their own strengths and weaknesses. Personalised risk information represents the strongest form of evidence from the expert perspective; nevertheless, its clinical value is diminished by uncertainties arising from methodological problems in risk modelling70 ,71 and the limited applicability of risk estimates to single events experienced by individuals.72 On the other hand non-quantitative forms of evidence (eg, ‘gut feelings,’73 physician recommendations, personal anecdotes) provide a means of mitigating irreducible uncertainties of risk estimates and are thus valuable from the lay perspective; however, such evidence is susceptible to numerous biases. Exactly how patients should weight these different types of evidence is a critical question for future research, but at the very least this task requires an understanding of their strengths and weaknesses. The lack of such understanding among study participants suggests that epistemological education should be a primary focus of efforts to apply personalised risk information to patient care.

    Our study thus identifies several potentially important tasks for enhancing the value and use of personalised risk information: (1) increasing patients’ exposure to personalised risk information, (2) providing resources to support deliberative decision-making and (3) providing epistemological education on the nature of medical knowledge and evidence. The first task has been a primary focus of CPM proponents; however, more work is needed to effectively disseminate and implement CPMs in clinical practice. The second task has only begun to be addressed, but promising approaches include the delivery of personalised risk information through patient-centred decision support interventions (DeSIs) such as decision aids,74 ,75 and the use of risk communication strategies such as visual representations to improve the evaluability of risk estimates.30 ,31 ,76–83 A more challenging task is to provide the environmental resources—for example, clinical time, processes and incentives—needed for deliberative and collaborative decision-making.74 The final task, the provision of epistemological education, is a new frontier that has yet to be explored but calls for efforts to expand the content of risk information communicated to patients—whether through DeSIs or larger-scale educational efforts delivered through other channels. It remains for further research to address these challenges, and to determine how best to help patients translate personalised risk information into better informed healthcare decisions.

    Acknowledgments

     The authors would like to thank the patients with prostate cancer who gave generously of their time and insights to participate in this study.

    Appendix A Clinical prediction model screenshots used as interview aids

    Embedded Image

    Embedded Image

    Appendix B Theme outline—focus group study

    1. Past experiences with risk communication

      1. Extent of communication

        1. No risk communication

        2. Detailed communication of individualised, quantitative risk information

        3. Non-individualised quantitative risk communication

        4. Imprecise quantitative risk communication

        5. Categorical risk communication: qualitative categories

        6. Indirect risk communication: biomarker heuristic

      2. Sources of risk information

        1. Doctor: urologist, radiation oncologist, PCP, other

        2. Other health professionals: nurse navigators, etc

        3. Friends, family

        4. Internet, books

    2. Perceived value of personalised risk information

      1. Promote SDM

      2. Provide insight into physician thinking

      3. Provide second opinion/counter physician bias

      4. Confirmatory

      5. Facilitate advance care planning

      6. Justify decisions

    3. Perceived challenges in use of personalised risk information

      1. Lack of self-efficacy in understanding risk information

      2. Logistic barriers: lack of time

      3. Physician bias/self-interest

      4. Preference for simplicity

      5. Low perceived need for information

        1. Low preference for information and participation in decision-making

        2. Trust the doctor heuristic

        3. Cancer heuristic (removal heuristic: ‘cut it out’): perceived lack of decision

        4. Reliance on intuitive versus rational decision-making approaches

      6. Distrust of statistical information: ‘anti-statistics’ viewpoint

        1. Preference for anecdotal/narrative versus statistical information: ‘unreality’ of numbers versus narrative information

        2. Questionable relevance to individual: reference class problem

        3. Uncertainty: epistemic (ambiguity)

          1. Imprecision

          2. Conflicting data

          3. Missing data, multiple/unknown/unaccounted risks

          4. Source credibility

        4. Uncertainty: aleatory (randomness)

      7. Motivated reasoning: desire for good news/fear of bad news

    4. Value of indirect risk communication (Biomarker Heuristic)

      1. Provide sense of control

      2. Communicate the bottom-line/gist

      3. Represent ‘hard information’

      4. Facilitate decision resolution

      5. Simplify decision-making

    5. Interpretation of personalised risk estimates

      1. Understanding of risk estimate: frequency or confidence statement

      2. Perceived difficulty

    6. Communication of uncertainty (ambiguity)

      1. Preferences for/against

      2. Reactions to use of range/CI

    References

    1. ↵
      1. Edwards A,
      2. Hood K,
      3. Matthews E,
      4. et al
      . The effectiveness of one-to-one risk communication interventions in health care: a systematic review. Med Decis Mak 2000;20:290–7.
      OpenUrl
    2. ↵
      1. Steyerberg EW
      . Clinical prediction models: a practical approach to development, validation, and updating. New York: Springer, 2010.
    3. ↵
      Personalized Health Care: Pioneers, Partnerships, Progress. 2008. http://www.hhs.gov/myhealthcare/ (accessed 12 Mar 2013).
    4. ↵
      1. Briss P,
      2. Rimer B,
      3. Reilley B,
      4. et al
      . Promoting informed decisions about cancer screening in communities and healthcare systems. Am J Prev Med 2004;26:67–80.
      OpenUrlCrossRefPubMedWeb of Science
    5. ↵
      1. Makoul G,
      2. Clayman ML
      . An integrative model of shared decision making in medical encounters. Patient Educ Couns 2006;60:301–12.
      OpenUrlCrossRefPubMedWeb of Science
    6. ↵
      1. Edwards A,
      2. Unigwe S,
      3. Elwyn G,
      4. et al
      . Effects of communicating individual risks in screening programmes: cochrane systematic review. BMJ 2003;327:703–9.
      OpenUrlAbstract/FREE Full Text
    7. ↵
      1. Barry MJ,
      2. Edgman-Levitan S
      . Shared decision making—pinnacle of patient-centered care. N Engl J Med 2012;366:780–1.
      OpenUrlCrossRefPubMedWeb of Science
    8. ↵
      1. Elwyn G,
      2. Frosch D,
      3. Thomson R,
      4. et al
      . Shared decision making: a model for clinical practice. J Gen Intern Med 2012;27:1361–7.
      OpenUrlCrossRefPubMed
    9. ↵
      1. Vickers AJ,
      2. Basch E,
      3. Kattan MW
      . Against diagnosis. Ann Intern Med 2008;149:200–3.
      OpenUrlPubMedWeb of Science
    10. ↵
      1. Edwards AG,
      2. Naik G,
      3. Ahmed H,
      4. et al
      . Personalised risk communication for informed decision making about taking screening tests. Cochrane Database Syst Rev 2013;(2):CD001865.
    11. ↵
      1. Gillies D
      . Philosophical theories of probability. London: Routledge, 2000.
    12. ↵
      1. Hacking I
      . The taming of chance. Cambridge: Cambridge University Press, 1990.
    13. ↵
      1. Han PK
      . Conceptual, methodological, and ethical problems in communicating uncertainty in clinical evidence. Med Care Res Rev 2013;70:14S–36S.
      OpenUrlAbstract/FREE Full Text
    14. ↵
      1. Han PK,
      2. Klein WM,
      3. Arora NK
      . Varieties of uncertainty in health care: a conceptual taxonomy. Med Decis Mak 2011;31:828–38.
      OpenUrl
    15. ↵
      1. Brewer NT,
      2. Richman AR,
      3. Defrank JT,
      4. et al
      . Improving communication of breast cancer recurrence risk. Breast Cancer Res Treat 2012;133:553–61.
      OpenUrlCrossRefPubMed
    16. ↵
      1. Lipkus IM,
      2. Samsa G,
      3. Rimer BK
      . General performance on a numeracy scale among highly educated samples. Med Decis Mak 2001;21:37–44.
      OpenUrl
    17. ↵
      1. Sheridan SL,
      2. Pignone M
      . Numeracy and the medical student's ability to interpret data. Eff Clin Pract 2002;5:35–40.
      OpenUrlPubMed
    18. ↵
      1. Schwartz LM,
      2. Woloshin S,
      3. Black WC,
      4. et al
      . The role of numeracy in understanding the benefit of screening mammography. Ann Intern Med 1997;127:966–72.
      OpenUrlCrossRefPubMedWeb of Science
    19. ↵
      1. Wegwarth O,
      2. Schwartz LM,
      3. Woloshin S,
      4. et al
      . Do physicians understand cancer screening statistics? A national survey of primary care physicians in the United States. Ann Intern Med 2012;156:340–9.
      OpenUrlCrossRefPubMedWeb of Science
    20. ↵
      1. Reyna VF,
      2. Nelson WL,
      3. Han PK,
      4. et al
      . How numeracy influences risk comprehension and medical decision making. Psychol Bull 2009;135:943–73.
      OpenUrlCrossRefPubMedWeb of Science
    21. ↵
      1. Han PK,
      2. Klein WM,
      3. Lehman TC,
      4. et al
      . Laypersons’ responses to the communication of uncertainty regarding cancer risk estimates. Med Decis Mak 2009;29:391–403.
      OpenUrl
    22. ↵
      1. Han PK,
      2. Lehman TC,
      3. Massett H,
      4. et al
      . Conceptual problems in laypersons’ understanding of individualized cancer risk: a qualitative study. Health Expect 2009;12:4–17.
      OpenUrlCrossRefPubMedWeb of Science
    23. ↵
      1. Ravdin PM,
      2. Siminoff LA,
      3. Davis GJ,
      4. et al
      . Computer program to assist in making decisions about adjuvant therapy for women with early breast cancer. J Clin Oncol 2001;19:980–91.
      OpenUrlAbstract/FREE Full Text
    24. ↵
      1. Peele PB,
      2. Siminoff LA,
      3. Xu Y,
      4. et al
      . Decreased use of adjuvant breast cancer therapy in a randomized controlled trial of a decision aid with individualized risk information. Med Decis Mak 2005;25:301–7.
      OpenUrl
    25. ↵
      1. Siminoff LA,
      2. Gordon NH,
      3. Silverman P,
      4. et al
      . A decision aid to assist in adjuvant therapy choices for breast cancer. Psychooncology 2006;15:1001–13.
      OpenUrlCrossRefPubMed
    26. ↵
      1. Belkora JK,
      2. Rugo HS,
      3. Moore DH,
      4. et al
      . Oncologist use of the adjuvant! model for risk communication: a pilot study examining patient knowledge of 10-year prognosis. BMC Cancer 2009;9:127.
      OpenUrlCrossRefPubMed
    27. ↵
      1. Hutton DW,
      2. Belkora JK,
      3. Shachter RD,
      4. et al
      . Are patients getting the ‘gist’ in risk communication? Patient understanding of prognosis in breast cancer treatment. J Cancer Educ 2009;24:194–9.
      OpenUrlCrossRefPubMedWeb of Science
    28. ↵
      1. Lipkus IM,
      2. Peters E,
      3. Kimmick G,
      4. et al
      . Breast cancer patients’ treatment expectations after exposure to the decision aid program adjuvant online: the influence of numeracy. Med Decis Mak 2010;30:464–73.
      OpenUrl
    29. ↵
      1. Zikmund-Fisher BJ,
      2. Fagerlin A,
      3. Ubel PA
      . Improving understanding of adjuvant therapy options by using simpler risk graphics. Cancer 2008;113:3382–90.
      OpenUrlCrossRefPubMedWeb of Science
    30. ↵
      1. Zikmund-Fisher BJ
      . The right tool is what they need, not what we have: a taxonomy of appropriate levels of precision in patient risk communication. Med Care Res Rev 2013;70:37S–49S.
      OpenUrlAbstract/FREE Full Text
    31. ↵
      1. Reyna VF
      . A theory of medical decision making and health: fuzzy trace theory. Med Decis Mak 2008;28:850–65.
      OpenUrl
    32. ↵
      1. Simon H
      . Models of man, social and rational: mathematical essays on rational human behavior in a social setting. New York: Wiley, 1957.
    33. ↵
      Cancer Research UK. Prostate Cancer Statistics. 2013. http://www.cancerresearchuk.org/cancer-info/cancerstats/types/prostate/ (accessed 4 Apr 2013).
    34. ↵
      1. Rothenberg BM,
      2. Marbella A,
      3. Belinson SE,
      4. et al
      . Future Research Needs for Comparative Effectiveness of Treatments of Localized Prostate Cancer: Identification of Future Research Needs from Comparative Effectiveness Review No. 13. Rockville (MD) 2010.
    35. ↵
      National Cancer Institute. Surveillance, Epidemiology, and End Results (SEER) Stat Facts Sheets: Prostate. 2013; http://seer.cancer.gov/statfacts/html/prost.html (accessed 4 Apr 2013).
    36. ↵
      1. Cooperberg MR
      . Prostate cancer risk assessment: choosing the sharpest tool in the shed. Cancer 2008;113:3062–6.
      OpenUrlPubMed
    37. ↵
      1. Shariat SF,
      2. Karakiewicz PI,
      3. Roehrborn CG,
      4. et al
      . An updated catalog of prostate cancer predictive tools. Cancer 2008;113:3075–99.
      OpenUrlCrossRefPubMedWeb of Science
    38. ↵
      1. Lughezzani G,
      2. Briganti A,
      3. Karakiewicz PI,
      4. et al
      . Predictive and prognostic models in radical prostatectomy candidates: a critical analysis of the literature. Eur Urol 2010;58:687–700.
      OpenUrlCrossRefPubMed
    39. ↵
      1. Cowen ME,
      2. Halasyamani LK,
      3. Kattan MW
      . Predicting life expectancy in men with clinically localized prostate cancer. J Urol 2006;175:99–103.
      OpenUrlCrossRefPubMedWeb of Science
    40. ↵
      1. Eastham JA,
      2. Scardino PT,
      3. Kattan MW
      . Predicting an optimal outcome after radical prostatectomy: the trifecta nomogram. JUrol 2008;179:2207–10; discussion 2210–2201.
      OpenUrlCrossRefPubMed
    41. ↵
      1. Ross RW,
      2. Kantoff PW
      . Predicting outcomes in prostate cancer: how many more nomograms do we need? J Clin Oncol 2007;25:3563–4.
      OpenUrlFREE Full Text
    42. ↵
      Lerner Research Institute, Cleveland Clinic. Risk calculators. 2010. http://www.lerner.ccf.org/qhs/risk_calculator/index.php (accessed 12 Nov 2010).
    43. ↵
      1. Touijer K,
      2. Scardino PT
      . Nomograms for staging, prognosis, and predicting treatment outcomes. Cancer 2009;115(13 Suppl):3107–11.
      OpenUrlCrossRefPubMed
    44. ↵
      1. Cegala DJ,
      2. Post DM,
      3. McClure L
      . The effects of patient communication skills training on the discourse of older patients during a primary care interview. J Am Geriatr Soc 2001;49:1505–11.
      OpenUrlCrossRefPubMedWeb of Science
    45. ↵
      nomogram.org. Prostate Cancer. 2010. http://nomogram.org/Prostate/pros_calc.php (accessed 12 Nov 2010).
    46. ↵
      1. Hoffman A,
      2. Montgomery R,
      3. Aubry W,
      4. et al
      . How best to engage patients, doctors, and other stakeholders in designing comparative effectiveness studies. Health Aff (Millwood) 2010;29:1834–41.
      OpenUrlAbstract/FREE Full Text
    47. ↵
      1. Sofaer S
      . Qualitative methods: what are they and why use them? Health Serv Res 1999;34(5 Pt 2):1101–18.
      OpenUrlPubMedWeb of Science
    48. ↵
      1. Kitzinger J
      . Qualitative research. Introducing focus groups. BMJ 1995;311:299–302.
      OpenUrlFREE Full Text
    49. ↵
      1. Morgan D
      . Focus groups in qualitative research. 2nd edn. Thousand Oaks, CA: Sage Publications, 1998.
    50. ↵
      National Comprehensive Cancer Network. NCCN Practice Guidelines for Prostate Cancer. 2010; Version 1: 2010. http://www.nccn.org (accessed 16 Dec 2012).
    51. ↵
      Memorial-Sloan Kettering Center . Prostate Cancer Nomograms: a Tool for Doctors & Patients. 2010. http://www.mskcc.org/mskcc/html/10088.cfm (accessed 12 Nov 2010).
    52. ↵
      1. Strauss AL,
      2. Corbin J
      . Basics of qualitative research: techniques and procedures for developing grounded theory. 2nd edn. Thousand Oaks, CA: Sage, 1998.
    53. ↵
      1. Ryan GW,
      2. Bernard HR
      . Data management and analysis methods. In: Denzin NK, Lincoln YS. eds. Collecting and interpreting qualitative materials. Thousand Oaks, CA: Sage, 2003: 259–309.
    54. ↵
      1. Glaser BG
      . The constant comparative method of qualitative analysis. Soc Probl 1965;12:436–45.
      OpenUrlFREE Full Text
    55. ↵
      1. Slovic P,
      2. Finucane M,
      3. Peters E,
      4. et al
      . The affect heuristic. In: Gilovich T, Griffin D, Kahneman D. eds. Heuristics and biases: the psychology of intuitive judgment. Cambridge: Cambridge University Press, 2002: 397–420.
    56. ↵
      1. Slovic P,
      2. Peters E,
      3. Finucane ML,
      4. et al
      . Affect, risk, and decision making. Health Psychol 2005;24(4 Suppl):S35–40.
      OpenUrlCrossRefPubMedWeb of Science
    57. ↵
      1. Wegwarth O,
      2. Gigerenzer G
      . Trust-your-doctor: a simple heuristic in need of a proper social environment. In: Hertwig R, Hoffrage U. eds. Simple heuristics in a social world. New York: Oxford University Press, 2013: 67–102.
    58. ↵
      1. Winterbottom A,
      2. Bekker HL,
      3. Conner M,
      4. et al
      . Does narrative information bias individual's decision making? A systematic review. Soc Sci Med 2008;67:2079–88.
      OpenUrlCrossRefPubMedWeb of Science
    59. ↵
      1. deWit JBF,
      2. Das E,
      3. Vet R
      . What works best: objective statistics or a personal testimonial? An assessment of the persuasive effects of different types of message evidence on risk perception. Health Psychol 2008;27:110–15.
      OpenUrlCrossRefPubMedWeb of Science
    60. ↵
      1. Shaffer VA,
      2. Zikmund-Fisher BJ
      . All stories are not alike: a purpose-, content-, and valence-based taxonomy of patient narratives in decision AIDS. Med Decis Mak 2013;33:4–13.
      OpenUrl
    61. ↵
      1. Gigerenzer G,
      2. Selten R
      . Bounded rationality. Cambridge: MIT Press, 2002.
    62. ↵
      1. Tversky A,
      2. Kahneman D
      . Judgment under uncertainty: heuristics and biases. Science 1974;185:1124–31.
      OpenUrlAbstract/FREE Full Text
    63. ↵
      1. Isaacson W
      . Steve jobs. New York: Simon and Schuster, 2011.
    64. ↵
      1. Peters E,
      2. Dieckmann N,
      3. Dixon A,
      4. et al
      . Less is more in presenting quality information to consumers. Med Care Res Rev 2007;64:169–90.
      OpenUrlAbstract/FREE Full Text
    65. ↵
      1. Iyengar SS,
      2. Kamenica E
      . Choice proliferation, simplicity seeking, and asset allocation. J Public Econ 2010;94:530–9.
      OpenUrlCrossRefWeb of Science
    66. ↵
      1. Iyengar SS,
      2. Lepper MR
      . When choice is demotivating: can one desire too much of a good thing? J Pers Soc Psychol 2000;79:995–1006.
      OpenUrlCrossRefPubMedWeb of Science
    67. ↵
      1. Legare F,
      2. Ratte S,
      3. Gravel K,
      4. et al
      . Barriers and facilitators to implementing shared decision-making in clinical practice: update of a systematic review of health professionals’ perceptions. Patient Educ Couns 2008;73:526–35.
      OpenUrlCrossRefPubMedWeb of Science
    68. ↵
      1. Lin GA,
      2. Aaronson DS,
      3. Knight SJ,
      4. et al
      . Patient decision aids for prostate cancer treatment: a systematic review of the literature. CA Cancer J Clin 2009;59:379–90.
      OpenUrlCrossRefPubMedWeb of Science
    69. ↵
      1. de Vries M,
      2. Fagerlin A,
      3. Witteman HO,
      4. et al
      . Combining deliberation and intuition in patient decision support. Patient Educ Couns 2013;91:154–60.
      OpenUrl
    70. ↵
      1. Spiegelhalter DJ,
      2. Riesch H
      . Don't know, can't know: embracing deeper uncertainties when analysing risks. Philos Transact A Math Phys Eng Sci 2011;369:4730–50.
      OpenUrl
    71. ↵
      1. Bilcke J,
      2. Beutels P,
      3. Brisson M,
      4. et al
      . Accounting for methodological, structural, and parameter uncertainty in decision-analytic models: a practical guide. Med Decis Mak 2011;31:675–92.
      OpenUrl
    72. ↵
      1. Han PK
      . Conceptual, methodological, and ethical problems in communicating uncertainty in clinical evidence. Med Care Res Rev 2013;70(1 Suppl):14S–36S.
      OpenUrlAbstract/FREE Full Text
    73. ↵
      1. Gigerenzer G
      . Gut feelings: the intelligence of the unconscious. New York: Penguin Books, 2007.
    74. ↵
      1. Elwyn G,
      2. Frosch D,
      3. Volandes AE,
      4. et al
      . Investing in deliberation: a definition and classification of decision support interventions for people facing difficult health decisions. Med Decis Mak 2010;30:701–11.
      OpenUrl
    75. ↵
      1. Stacey D,
      2. Bennett CL,
      3. Barry MJ,
      4. et al
      . Decision aids for people facing health treatment or screening decisions. Cochrane Database Syst Rev 2011;(10):CD001431.
    76. ↵
      1. Lipkus IM
      . Numeric, verbal, and visual formats of conveying health risks: suggested best practices and future recommendations. Med Decis Mak 2007;27:696–713.
      OpenUrlCrossRef
    77. ↵
      1. Visschers VH,
      2. Meertens RM,
      3. Passchier WW,
      4. et al
      . Probability information in risk communication: a review of the research literature. Risk Anal 2009;29:267–87.
      OpenUrlCrossRefPubMedWeb of Science
    78. ↵
      1. Galesic M,
      2. Garcia-Retamero R,
      3. Gigerenzer G
      . Using icon arrays to communicate medical risks: overcoming low numeracy. Health Psychol 2009;28:210–16.
      OpenUrlCrossRefPubMedWeb of Science
    79. ↵
      1. Garcia-Retamero R,
      2. Galesic M
      . Who profits from visual aids: overcoming challenges in people's understanding of risks [corrected]. Soc Sci Med 2010;70:1019–25.
      OpenUrlCrossRefPubMedWeb of Science
    80. ↵
      1. Dolan JG,
      2. Qian F,
      3. Veazie PJ
      . How well do commonly used data presentation formats support comparative effectiveness evaluations? Med Decis Mak 2012;32:840–50.
      OpenUrl
    81. ↵
      1. Akl EA,
      2. Oxman AD,
      3. Herrin J,
      4. et al
      . Using alternative statistical formats for presenting risks and risk reductions. Cochrane Database Syst Rev 2011;(3):CD006776.
    82. ↵
      1. Gigerenzer G,
      2. Gaissmaier W,
      3. Kurz-Milcke E,
      4. et al
      . Helping doctors and patients make sense of health statistics. Psychol SciPublic Interest 2007;8:53–96.
      OpenUrl
    83. ↵
      1. Spiegelhalter D,
      2. Pearson M,
      3. Short I
      . Visualizing uncertainty about the future. Science 2011;333:1393–400.
      OpenUrlAbstract/FREE Full Text

    Footnotes

    • Contributors PKJH designed the study, obtained funding, conducted interviews and data analysis, drafted the manuscript and edited the manuscript for important content. MN, BR and NH conducted data analysis and edited the manuscript for important content. TK assisted in the design of the study, conducted interviews and edited the manuscript for important content. CG edited the manuscript for important content. MD edited the manuscript for important content. MH assisted in the design of the study and edited the manuscript for important content.

    • Funding This study was supported by a Patient Education and Support Grant from the Maine Cancer Foundation.

    • Competing interests None.

    • Ethics approval The study was approved by the Maine Medical Center Institutional Review Board (IRB#3805X), and granted a waiver of informed consent.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Data sharing statement No additional data are available.