The notion of risk, broadly defined as the likelihood that a future event will occur, is a fundamental concept in cancer prevention. The scientific design of prevention programs is based on assessments of disease risk and the likelihoods of benefits and harms resulting from possible preventive interventions. Once implemented, individual perceptions of risk influence the effectiveness of prevention efforts. Variations in understanding and perception of risk can result in both higher risk people deciding not to participate in recommended prevention activities and lower risk people seeking out prevention activities that have a greater chance of causing harm than good.

The term risk communication refers to the process of conveying information about the chances of uncertain future events, both good and bad, to a target audience. Although this seems like it should be simple, successful risk communication is a difficult task. The purpose of this talk is to discuss current research and recommended best practices regarding three important factors that influence the effectiveness of risk communication: context, content, and format.

The risk communication context is defined by the goal of the communication. Two common goals are: a) to persuade members of the target audience to change their behavior, and b) to improve their understanding of how likely future events are expected to occur. In this talk, I will assume that the goal of risk communication in cancer prevention is to appropriately align people's perceptions with the most accurate risk assessment possible.

The goal of increased understanding can be further defined in several ways ranging from the simple provision of information about event likelihoods, to providing people with a better understanding of both the likelihoods and consequences of possible future events, to making comparisons about the relative desirability of different courses of action based on the nature and likelihood of their possible consequences. The choice of definition will determine both the appropriate research methods and the implications of research findings. To date, most risk communication research has addressed the question of how to best convey probabilistic frequency information. The extent to which the findings of these studies can be generalized across the entire spectrum of communication goals is as yet unknown and is an important question for future research. Risk information can be characterized using verbal or numerical expressions. Although commonly used, verbal expressions of risk — such as high, moderate, and low — have been shown to be subject to wide variations in interpretation. For this reason, their routine use, especially in isolation, is not recommended. There is evidence, however, that adding verbal information to numerical risk expressions can improve recipients’ understanding of the information presented.

The most common way to describe risk is in numerical terms. Numeric risk expressions- such as 5% or 20 per 1,000 — have been shown to be highly effective ways to communicate risk information, providing they are properly constructed. Factors that affect the effectiveness of numeric risk expressions include how the risk information is expressed and how it is framed.

Common methods used to convey information about the risk of single events include percents, “x in 100” terms, and “1 in y” terms. The x in 100 term reports the number of affected individuals out of an appropriate sample size that is usually a factor of 10 such as 100, 1,000, 10,000, etc. The 1 in y term also includes information about the risk in relation to a total sample but, in this case, using a common numerator as opposed to a common denominator. There is good evidence that it is easier for people to understand quantitative risk information when presented using expressions that use common denominators. For this reason, current guidelines regarding the numeric expression of risk information recommend the use of percents or 1 in x terms.

Framing bias refers to the finding that the interpretation of an expression of risk can be affected by whether the outcome presented in a positive or negative frame, even though they are logically and mathematically equivalent. For example, a 5% risk of harm can be interpreted differently from a 95% chance of no harm. To counteract possible framing effects, the use of both positive and negative frames, called a balanced risk presentation, has been recommended. The extent to which balanced presentations actually improve the accuracy of understanding compared with “unbalanced” presentations is still a question being actively researched.

The final consideration we will discuss is how the risk information is formatted. Formatting is particularly important when the communication objective is to describe the potential outcomes of two or more alternative courses of action. Since not adopting a recommended intervention is always an option, preventive health communications typically involve risk comparisons. Possible formats include text, table, graphs, and combined formats. Theoretical and empirical evidence suggests that comparative risk information should be formatted to minimize the amount of cognitive processing required to interpret the data presented and to take full advantage of human intuitive processing capabilities.

To minimize required cognitive processing, risk information comparing alternative courses of action presented within a line of text should include not only information about the individual risks but also about their differences. Comparative risk information can be presented in terms of relative risk differences, absolute risk differences, and/or number needed to treat. Relative risk information has been repeatedly shown to lead to exaggerated risk perceptions and is not recommended. Absolute risk information, on the other hand, appears to provide most people with a more accurate understanding of the difference in risk expected to result from alternative courses of action. Numbers needed to treat, which are derived from absolute risk differences, were developed to further help people interpret risk differences. Experimental studies to date, however, have found that they have few, if any, added advantages over the simple absolute risk difference. Thus, instead of saying “…risk A is 5% and risk B is 10%” it is better to say “…risk A is 5% and risk B is 10%; the difference between the two is 5%”. Another strategy to minimize cognitive load is to always use whole numbers rather than decimals and fractions when formatting numeric risks.

A fundamental problem with text-based formatting is that it is not well-suited to take advantage of human intuitive processing skills. Tables are better able to take advantage of human processing skills than text-based formats because they provide a way to both include information formatted to minimize cognitive processing and organize it to focus attention on the alternative comparisons. For this reason, many feel that a tabular format is the best way to format numeric risk expressions when comparing alternatives.

Graphic formats provide even greater opportunities to organize risk information such that humans can process it easily and are considered by many to be the preferred risk communication format. Unfortunately, the added capabilities of graphic formats can both serve to improve the clarity of risk information and systematically bias perception of the data.

Graphic communication theory suggests that graphic risk communication formats should present information using an aligned common scale that illustrates both the magnitude of the risk in question and its relationship to the entire population. Examples of formats that meet both of these criteria are bar charts and icon arrays. Current research evidence supports the theoretical advantages of these two graphic formats for many risk communication tasks. Knowledge gaps persist, however, especially with regard to the best way to graphically represent small risks (less than 1-2%) that are hard to depict using these formats and the uncertainty associated with any particular risk estimate.

There has been surprisingly little research into the relative benefits of using combined graphic and numeric risk formats. A major exception has been work showing that frequency diagrams, a series of linked bits of information similar to a treatment algorithm or organizational diagram is an effective way of providing people with probabilistic information in an understandable way.

A final consideration affecting the effectiveness of risk communication is the capabilities of the recipients of the risk message. There is good evidence that education level and numeracy skills affect the effectiveness of risk communication messages. There is also evidence that the ability to interpret graphs may a separate skill, related to, but independent of, general numeracy skills. These findings suggest that separate risk communications may have to be created to better suit these differences in people's abilities to interpret the information properly.

Additional areas of active research include investigation into the merits of including a familiar risk, such as the risk of a car accident or being struck by lightning, to help people interpret the meaning of unfamiliar risks, such as those related to many medical situations, and the effects of risk communication on recipients’ decision-making processes.

Citation Information: Cancer Prev Res 2010;3(12 Suppl):CN15-03.