Improving bone mineral density reporting to patients with an illustration of personal fracture risk

Osteoporosis [OP] is a common disease of the skeletal system associated with fragility fractures. OP fractures, particularly of the hip, have been associated with increased morbidity and mortality, and decreased quality of life [1]. Because of these factors and the increased likelihood of chronic pain and dependence, persons with or who are at risk of developing OP will likely want to know their risk of having a fracture. Health care providers can quickly calculate an individual patient's 10-year probability of hip and other major osteoporotic fractures combined (hip, vertebrae, distal forearm and proximal humerus) by using the World Health Organization's FRAX® fracture risk assessment tool (www.​shef.​ac.​uk/​FRAX/​) [2]. This web-based calculator uses 11 factors (age, race, sex, body mass index, prior history of fracture, parental history of fracture, secondary diseases, steroid use, smoking and alcohol intake, and bone mineral density [BMD] as determined by a dual energy X-ray absorptiometry [DXA]) to calculate an individual's personalized 10-year absolute fracture risk. Providers typically use this risk calculation to communicate with patients when considering initiation of medication therapy and lifestyle counseling in the areas of nutrition, exercise, and tobacco and alcohol use. Because OP is asymptomatic and DXA alone is insufficient to predict fracture risk, providing individualized risk is an important strategy for reducing fracture risk and related morbidity. Unfortunately, it is unclear whether patients understand their DXA results, are able to act on this vital information in an appropriate way, or that health care providers know how to convey DXA results and FRAX® in a way that patients understand [3]-[7].

According to the Health Belief Model [HBM], prevention and adherence behaviors rely on patients knowing and understanding that they are at risk for a disease [8]. The HBM theorizes that patients will take necessary actions if they believe they are at risk for poor outcomes and that these poor outcomes can be prevented (e.g., osteoporotic fractures) by taking a recommended action (e.g., pharmacotherapy, calcium, weight-bearing exercises). In other words, patients may choose not to employ fracture risk reduction behaviors if they do not believe they are at risk, if they are not prepared to take action, or if they do not perceive benefits from the action. Accordingly, the OP prevention literature has demonstrated that patients at risk for fractures do not take actions to reduce their risks in part because they do not recognize their at-risk status [4],[7],[9]-[17].

The communication of risk is more complex than simply providing patients with information about their condition, its treatment, and sequelae. Health care providers seeking to communicate risk to patients must take into account their patients' health and numeracy skills. Numeracy is the ability one has understanding risk concepts and basic probability [18],[19]. As is the case for literacy [20], average numeracy in the United States is low [21]. Even highly educated samples perform poorly on fairly simple probability questions [18]. The consequences of poor numeracy for health behavior are high in the case of OP. For example, a 20% chance of major osteoporotic fracture in 10 years represents a high risk of sustaining a fracture, but patients may not understand that 20% is a high risk when receiving this number verbally or in a letter from their doctor [22].

Combining written, numeric, and illustrated representations of risk can ameliorate the effects of lower numeracy on patient understanding [23]-[26]. Studies to maximize patient comprehension of risk have employed a variety of images including bar graphs, pie charts, "thermometers," and icon arrays. However, none of these studies have shown a single approach to be the most preferred or comprehensible to patients. In a study on comprehension of breast cancer risk, patients preferred an icon array to a bar graph [25]. Ghosh et al. found that combining a bar graph with an icon array lead to better understanding for patients who inaccurately perceived themselves to initially be at high risk of breast cancer [26]. Hill et al. found when presented with absolute risk of heart attack in the next five years, most patients preferred a risk thermometer [24]. Despite these promising directions, the published literature clearly lacks consensus on which depiction patients most easily understand, prefer, and are most motivated by [27],[28]. In the field of OP, we were able to identify only a single trial that evaluated the use a pictogram to display fracture risk. The pictogram displayed personal risk of fracture and absolute risk reduction with pharmacotherapy and was used in a decision aid to help patients understand the relative benefits of taking bisphosphonates. Patients who used the decision aid were twice as likely to correctly identify their 10-year fracture risk as patients who did not see the pictogram [29].

We studied the feasibility of developing an instrument combining text and illustration to convey fracture risk to patients in a way that they prefer and understand. We set out to test patients' preferences for, ease of understanding, and perceived fracture susceptibility after presenting subjects with a series of alternative illustrations depicting individualized risk of major fracture derived from a FRAX® score. Our objectives were to identify: which depictions patients prefer and why; whether patients can correctly ascertain fracture risk from the images and not over or underestimate risk; and which depiction is most associated with the desired health behavior outcome of seeking follow-up care from the referring provider. This study informed the instrument development for our current randomized clinical trial, The Patient Activation After DXA Result Notification (PAADRN) study, which tests the efficacy of a direct-to-consumer mailed DXA reporting intervention to activate patients for appropriate follow up with their health care providers based on test results [30].

In this mixed methods study, we found that participants preferred the BAR illustration and they did not prefer the icon array FACES. The use of stoplight colors to reflect fracture risk category was popular. We presented participants with four illustrations of identical risk in random order, asked them to rate their preference, ease of understanding, intent, and worrisome nature, and then asked them to explain their ratings. We found differences in participants' response to the illustrations, demonstrating that the representation of risk has important bearing on patient's consideration of health care information.

There was no strong consensus on the preferred illustration as none was preferred by more than half of participants. The BAR was most preferred of the four with 37% selecting it as their favorite option. Additionally, respondents listed BAR as the easiest to understand illustration. Our findings were similar to the findings of other studies that used a vertical bar graph to display risk [36],[37]. McCaffery et al. found that participants had a strong preference for a bar graph versus an icon array, such as our FACES design. Additionally, they found that participants were more likely to accurately identify risks of >10% using a bar graph when compared to an icon array [36]. Hawley et al. also found that participants performed better on verbatim tasks (the ability to correctly read numbers from graphs) when shown a bar graph versus an icon array [38]. While participants reported that BAR was the easiest to understand of the four illustrations, they did not perform as well when asked about risk severity. Due to the participant feedback, we felt that the color gradation caused this confusion and not the bar-type format.

Risk communication studies are increasingly employing icon arrays as a mechanism for addressing low numeracy among patients. However icons arrays were least likely to lead to a correct identification of risk category by participants. Participants were least likely to prefer the icon array FACES. The FACES option likely suffered from a lack of risk stratification from low to moderate and then to high risks, a design feature in the other types of graphics that was very much preferred by participants. Our data suggest that patients with lower educational attainment may be more likely to prefer icon arrays to other illustrations; however, the difference between educational subgroups was not statistically significant. Additionally, we found that preference for a depiction was not significantly associated with average numeracy levels. Our finding that icon arrays are least preferred by some respondents may be in part due to the arrangement of negative icons within the array. There is evidence to suggest that random arrangement of icons within the array more effectively produces increased susceptibility [39], however open-ended comments by our respondents revealed that they found the depiction of a happy or sad face to be infantile, so it is unclear whether random arrangement would have surmounted that appraisal.

Use of stoplight colors emerged as a clear factor in respondents' preferences and interpretation of information. The universal stoplight coloring system which equates the color green with health and positive movement, yellow with caution and slowed movement, and red with danger, was clearly internalized and applied to the interpretation of risk. Similar uses of the traffic light color system were used in other studies and found to be well perceived by participants in conveying risk [24],[32]. However, the use of graduated colors along these shades in the BAR illustration was perceived to be confusing by respondents as evidenced by 44% of participants stating BAR depicted 21% risk of fracture as a moderate risk rather than as a high risk of fracture. While a color gradation was used in a prior study, there was no mention of their participants' opinions of color gradation [32]. We conclude that, because the color graduated from orange to red slightly above the 20% line, participants still considered 21% to be moderate risk. Our stoplight coloring may limit a color-blind patient's understanding of the depictions. While we used the color system to draw attention and aid in comprehension, we also provided labels to the categories so a color-blind individual could comprehend their risk level. Given the low prevalence of this disorder, specifically in women [40] who represent 80% of patients undergoing DXA, we feel that using this coloring system would be mostly beneficial.

Our study is not without limitations. Instead of running four separate experiments, we opted to test depictions sequentially in random order. Exposing all subjects to all illustrations may have primed them with greater reinforcement of the information when answering the questions for later illustrations; however we moderated this potential effect by randomizing the order in which illustrations were tested. Similarly, we tested three depictions that were somewhat different from one another and one (FACES) that was different from the others, rather than four similar illustrations. This purposeful variation was a methodologic approach to evaluate which depiction is preferred and led subjects to identify clinically correct risk level comprehension. For example, an icon array such as FACES can help patients visualize their true probability of fracture, but it does not aid them in understanding what is clinically considered high risk. This was evidenced by 75% of participants who considered a 21% as a low risk of fracture when viewing FACES. Our study sample was drawn from a general clinic population, which included by chance, some people who had knowledge of OP but others who had no prior knowledge of OP. Because some patients may have more knowledge of OP than others, we do not necessarily know if our results might differ between these two subgroups had samples been recruited specifically along these lines. However, for the purposes of PAADRN, this is not necessarily a limitation as we will be recruiting a similar mix of patients. For the purposes of the PAADRN study we wanted to learn which array was preferred and understood by the majority of participants in our target audience, patients undergoing DXA either for the first time or a repeat DXA and who are 50 years of age and older [30]. We found that non-Whites preferred FACES to other illustrations but this was not statistically significant. A larger sample of respondents with lower educational attainment and numeracy may have different responses. Additionally, we used a convenience sampling method which makes it difficult to ascertain that our quantitative results are reflective of the clinic population. However, we did use a more purposeful sampling technique to ensure that our sample was geographically, racially, and economically diverse. One illustration clearly emerged as the most preferred and most easily understood risk depiction among our target audience; however we concur with other researchers that a single risk illustration might not be unanimously accepted by all people [32].

To our knowledge, this is the first study to examine patient preference for communicating FRAX®. We found that numeracy was not significantly associated with preference, perceived ease of understanding, or correct identification of risk category for any of the illustrations examined in this study. Participants were asked to both rate their preferred format and to explain their reason for doing so, and this type of integration of qualitative and quantitative measures moves our understanding or risk communication forward. While participant preference was important in creating the depiction used for the PAADRN trial, we felt it was most important to use an illustration that patients both liked and which was associated with correct appraisal of fracture risk. Additionally, our goal in creating this depiction was not to recommend treatment but rather to choose a depiction that might motivate patients to communicate with their health care provider about their fracture risk or to make health behavior changes like increasing weight-bearing exercise or dietary calcium. The PAADRN trial will assess patient motivation and behaviors to improve their bone health upon receiving their illustration of personal fracture risk [30]. We found that the majority of participants in our sample significantly preferred and understood a bar-type graph to display 10-year fracture risk. This illustration could easily be added to DXA reports to aid health care providers and patients in making bone health care decisions. This type of work provides valuable and needed information to health care providers who want to improve DXA follow- up care by using illustrations to communicate personalized fracture risk using a depiction that is well perceived by a general audience of 50+ years.

This study describes the methods used to develop a visual depiction of fracture risk that will be mailed with DXA results to patients in the intervention arm of the randomized clinical PAADRN trial. We found that participants had a significant preference for a vertically oriented risk depiction that provided clinically significant risk categories using a stoplight coloring system. As highlighted above, numeracy is an important consideration when communicating risk information to patients. Providing patients with a visual depiction of their personal risk of a disease or disease consequence may assist them in their understanding. However, careful thought should be taken when describing risk to patients, as many adults have low numeracy skills. When determining the best representation of risk obtaining feedback for preference and comprehension from a sample of patients in a target patient population is a critical first step in providing patients with an effective and acceptable risk depiction.

SE conceived of the study, led in its design and coordination, and was one of the lead authors of the manuscript. PC conceived of the study, oversaw and contributed in its design, and assisted in drafting the manuscript. XL assisted in the design of the study, led statistical analysis and assisted in drafting the manuscript. DR assisted in the design of the study, oversaw data collection, and assisted in drafting the manuscript. NW assisted in drafting the manuscript. KS assisted in the design of the study and assisted in drafting the manuscript. SS conceived of the study, led in its design and coordination, and was one of the lead authors of the manuscript. All authors read and approved the final manuscript.