Much has been made of the American public's general lack of health literacy and the effect it has on patients’ abilities to make informed health decisions. A key component of health literacy is the ability to understand statistics and concepts like “probability” and “risk.” Thus, it is important that physicians be able to explain these concepts in a way that enables patients to incorporate the information into their decision-making process.

 Life is uncertain, and, as stated by Kathryn Montgomery in her treatise 

, “few experiences feel as uncertain as illness.” Furthermore, what is true for patients is equally true for their physicians. Uncertainty has been rightly 

 as “a fundamental feature of medicine… the physician’s constant companion.” Much of medicine is, in fact, about making effective decisions in the face of uncertainty.

	Sound medical decisions depend on an informed physician and an informed patient. Physicians must know or be able to find relevant data to support assessments of the risks and benefits of studies and treatments. They must understand the limitations of such data along with basic statistical concepts of uncertainty. Furthermore, physicians must be able to communicate this information in such a way that a patient can comprehend it within his or her unique cultural and educational background and values. Clearly, guiding patients to sound, informed medical decisions in the face of inherent uncertainty is no small feat.

	Much of the medical decision-making process involves evaluating risk. Risk is 

 as uncertainty about the state of the world that can be quantified as a probability. Assessment of risk therefore depends on the availability of reliable empirical data. Unfortunately, in spite of the advances made by evidence-based medicine, such data often remains illusory or incomplete, introducing uncertainty even into the things we “know.”

	Risk is classically expressed using several methods. The absolute risk reduction (ARR) (or, in the case of harm, absolute risk increase) is the most straightforward measure of risk. The ARR is defined as the absolute difference in the proportion of treated and untreated patients who experience an outcome. Another commonly used measure is the relative risk ratio (RRR). RRR is the difference in the proportions of untreated and treated patients who experience a particular outcome relative to the proportion of untreated patients who would experience the outcome. The major problem with RRR is that relative risk measures tend to exaggerate small differences in effect (it is for this reason that relative risk measures might be preferentially reported when the purpose is to support sales of a product or to create sensationalism).

	Another risk descriptor preferred by many is number needed to treat. The number needed to treat is the reciprocal of the absolute risk reduction and represents the total number of patients who must be treated to produce the target outcome in one patient. Number needed to treat has the advantage of facilitating comparisons of risks and benefits between several options. The smaller the number needed to treat, the better the treatment. Conversely, the larger the number needed to harm, the less the risk.

	The relative merits and disadvantages of using each of these measures to communicate risk can be illustrated by the following example. Suppose 100,000 subjects are given a drug thought to prevent death from heart disease and 100,000 subjects are given placebo. If two subjects in the control group die while only one in the drug group dies, the absolute risk reduction is 1/100,000 (2 out of 100,000 minus 1 out of 100,000). If a similarly designed study of 100,000 subjects using another drug found that 20,000 control subjects and 10,000 drug subjects died, the absolute risk reduction would be 10,000/100,000 (20,000/100,000 minus 10,000/100,000). The number needed to treat in the first study would be 100,000, whereas it would be just 10 for the second. In essence, the second drug is 1000 times more effective. This benefit is completely obscured, however, if expressed as relative risk reduction (a 50% reduction in each study). Clearly, RRR is not the best way to convey risk information.

Do doctors and patients understand risk? 

	It turns out that a large proportion of people, including otherwise well-educated adults, simply do not know how to think and understand using numbers. This problem has been termed innumeracy by the mathematician 

. Kurz-Milcke and colleagues describe one 

 that reported findings in terms of a relative risk:

	“In the mid-1990s, the British press reported the results of a study that women who took [a particular] contraceptive pill increased their risk of thromboembolism by 100%... Hearing the bad news, thousands of British women panicked and stopped taking the pill, which led to a wave of unwanted pregnancies. But what did the study in fact show? Out of every 14,000 who took it, this number increased from one to two. That is, the absolute risk increase is 1 in 14,000, but the relative increase is 100%. Most of these women, like the majority of people, had never learned the difference between absolute and relative risks and thus could easily be frightened by the 100% figure.”

 that found that 88% of subjects indicated they would take a hypothetical lipid-lowering drug that would result in a 34% reduction in heart attacks (relative risk reduction), but only 42% would take the drug when told it would result in 1.4% fewer patients having heart attacks (the equivalent absolute rate reduction). When informed that 71 subjects would have to be treated for five years to prevent one heart attack (the equivalent NNT), only 31% would do so. Yet, 

 that death rates expressed as “1,286 out of 10,000” were rated by subjects as more risky than rates of “24.14 out of 100.”

	Unfortunately, statistical innumeracy is not limited to the lay public. Hoffrage and Gigerenzer 

 48 physicians with an average of 14 years of professional experience. Each physician was asked to estimate the chances of breast cancer in a woman age 40-50 years given a positive routine screening. Physicians who were given the relevant information in probabilities (eg, 0.8%) estimated the likelihood to be between 1% and 90% with surprisingly little consensus (the correct answer was 9%). Physicians did better when they were given the same information in natural frequencies (for example, “8 out of every 1,000”) instead of probabilities. Presented this way, the majority were able to respond with the correct answer or close to it. Other researchers 

 similar results. Since patients also find percentages difficult to comprehend, it is suggested that risks always be considered and described in terms of natural frequencies.

	Some patients prefer to avoid any reference to numerical descriptions. For example, Woloshin and colleagues 

 that nearly 30% of subjects preferred that risk information be conveyed using words only. However, avoiding numerical risk information did not remove the potential for misunderstanding. Only 80.3% of their subjects agreed that “certain” meant 100 of 100 people, and only 67.8% agreed that “never” meant zero of 100 people. As would be expected, words describing less polarized alternatives were even more ambiguously interpreted.

	Newer approaches to communicating risk involve visual representations designed to convey information more intuitively. One simple approach uses a diagram containing 100 icons designed to represent patients and possible outcomes. For example, an absolute risk reduction of 7% might be visually represented by shading seven of the 100 icons differently to indicate that this number of patients would be expected to benefit from a therapy relative to the total population of 100. Such a diagram incorporates the advantages of absolute risk reductions represented as a natural frequency. Rifkin and Bouwer 

,” a simple graphic illustration of a typical theater seating chart with 1000 seats. In this easily understandable example of a population sample, seats are colored differently to indicate the number of individuals in a hypothetical theater audience who would be expected to experience a given outcome.

	Dr. Chris Cates has further extended visual risk representation in an innovative and entertaining method that represents numbers needed to treat and/or numbers needed to harm in a single diagram (Figure 1). In my opinion, this clever plot provides a significant and practical advance in risk communication. Furthermore, Dr. Cates has created the program 

, which allows anyone to generate such a plot by entering data available from the medical literature for the condition and treatment of interest.

	Given the difficulty that both patients and physicians have in innately understanding numerical concepts of risk and the lack of consensus about the meaning of even commonly used words to describe uncertainty, how can we communicate risk to patients? Are there available guidelines?

	First, clinicians must be sure that they understand the merits and drawbacks of the commonly used measures of risk, as reviewed above. Risks should be communicated using absolute measures, such as number needed to treat (or harm) and absolute risk reduction (or increase). Given their inherent potential for misinterpretation, relative risk measures should generally be avoided in discussions with patients. Probabilities should be described as natural frequencies instead of percentages, and commonplace words such as “likely” and “uncommon,” if used, should be defined as explicitly as possible. Visual aids, such as the Risk Characterization Theater and Cates Plots, may be particularly helpful in conveying quantitative concepts more intuitively.

	Apart from the content, attention should also be given to the context in which risk information is given. For example, patients 

 to be sensitive to the ways in which risk information might be framed. Framing the benefit of a screening test in terms of what the patient might lose by not having it has been shown to be more persuasive than framing involving potential gains. Likewise, framing risk information in positive terms (for example, chance of survival) is more likely to persuade patients to accept risky options than information presented in negative terms (chance of death).

	Risk communication is multidimensional and must be primarily patient-centered. To remind physicians of this, 

, suggest approaching risk communication using the mnemonic RISK: R — Relate relevant evidence; I – Individualize the message, using a flexible approach; S – Seek the patient’s perspectives and share the decision-making; and K-checK the patient’s understanding, monitor, and review decisions. In regard to the last point, I have found that asking a patient to teach back to me what I have just explained is one of the simplest ways to find out whether or not they truly comprehend the message.

	We cannot attain certainty in life (with the exception of death and taxes, according to Benjamin Franklin). At best, we can try to offer clear information to our patients by attempting to quantify and communicate anticipated risks. Ultimately, the success of that communication will not be gauged by how we convey risk information, but by whether or not it is understood by patients who make important decisions based on the information we provide. Certainly, we owe this to them. 

 editorial board member, and assistant professor of clinical medicine and associate program director in the Department of Medicine at the University of Illinois College of Medicine, Peoria, IL.

		The Cates Plot shown above is an example of the output from a Cochrane Review of the use of antibiotics in Acute Otitis Media (middle ear infection) in children. The data is taken from the Cochrane Library 1999 issue 4.

		The picture represents 100 children who are all given antibiotics for ear infections. The 86 green faces are children who would have been free from pain at 2 to 7 days even if they had not received an antibiotic. The 9 red faces are children who are still in pain even with antibiotics. The 5 yellow faces are the only children who show a benefit; they would have been in pain without the antibiotic but are not when they receive one.

		Since it is not possible to identify which children will benefit all 100 need to be given the antibiotic for 5 to benefit. This represents the number needed to treat (NNT) of 20 for a single child to benefit.

. Reproduced with permission of Dr. Chris Cates.

 Communicating Risk: The Main Work of Doctors

Demystifying Risk: Understanding and Communicating Medical Risks 

Fuzzy Math: Some Medical Stats Confuse Patients 

Helping Doctors and Patients Make Sense of Health Statistics

How Can Doctors Communicate Information about Risk More Effectively? 

Making Sense of Risk Information on the Web 

Risk Communication in Practice: The Contribution of Decision Aids 

Risk Communication: Problems of Presentation and Understanding 

The Role of Risk Communication in Shared Decision Making 

Simple Tools for Understanding Risks: From Innumeracy to Insight 

Strategies to Help patients Understand Risks

Teaching Medical Students and Doctors How to Communicate Risk 

Articles

A key component of health literacy is the ability to understand statistics and concepts like "probability" and "risk."

The Risk of Risk: Explaining Difficult Concepts to Patients

	"Nobody knows the form of the information tool that may transform medicine, but we can begin to discern some of its characteristics. The tool must be able to answer highly complex questions and so will have to be connected to a large valid database. Inevitably, it will be electronic, but it must be portable, fast, and easy to use."

Healthcare computing is unique. First, healthcare is mobile. It is practiced wherever the patient resides at the moment. Second, time is scarce for clinicians. Our interactions with our tools must be effi cient and seamless. Finally, healthcare information is sensitive, demanding high-level security to ensure our patients’ privacy. As evidenced from the opening quote, physicians have long anticipated a device that will satisfy our distinctive information management needs. Could it be the tablet PC?

	What is a tablet PC? According to Microsoft, tablet PCs are computers powered by Windows operating systems that support functionality for input using digitizing screens designed to interact with a complementary pen. Tablet PCs have the full functionality of laptop PCs, but are primarily distinguished by pen input capability. One can use the pen directly on the screen just as you would a mouse to do things such as select, drag, and open fi les, or in place of a keyboard to handwrite notes and communication. By interacting directly with the screen, the PC becomes both more comfortable and easier to use (especially for poor typists).  

	Handwriting recognition, which translates handwritten screen input into editable text, distinguished the original tablet PC versions of Windows XP. Windows Vista now incorporates integrated support for tablet PCs, which, according to Debbie Martin, Microsoft’s OEM account manager for mobility, refl ects Microsoft’s solid commitment to the tablet PC platform. “We’ve always felt that the tablet is an important technology,” Martin said. “Some people at Microsoft believe that tablets are already mainstream,” which is why support for tablet PCs was embedded within Vista, and why further improvements in pen input and inking were included in Windows XP’s successor.

	Physicians should know that contemporary tablets also have the hardware horsepower to run voice recognition applications. Thus, one device can off er a variety of input options (keyboard, handwriting recognition, digital ink, point and click, speech), depending on the need and the user.

	In addition to pen input, the tablet PC’s form factor off ers additional advantages to clinicians. Most tablet PCs are designed as “slates” without an attached keyboard (although keyboards are available as optional accessories), or as “hybrids” with attached keyboards that can be optionally folded out of the way beneath the screen. Therefore, there is no need to find a fl at space on which to use the device, nor does a vertical screen impose a barrier between clinician and patient. The form factor lends itself well to use during rounds and other patient settings, much like the ubiquitous clipboard. Screen size for most tablet PCs is about 12” (measured diagonally), resulting in a device that is similar in size, or in some cases, smaller than a clipboard.

	Some tablet PCs, such as Motion’s new C5 model, are engineered specifi cally for healthcare environments and include features such as shock-resistance, integrated RFID readers, optional bar code scanners, and cameras for documentation of clinical findings. The C5’s case is even designed to allow for disinfection.

	Tablet PCs are generally priced between $1,000 and $2,500, and are off ered by a variety of manufacturers. Tablet PCs should not be confused with ultra-mobile PCs (UMPC). Although UMPCs might be considered smaller siblings of tablet PCs, the two devices are not functionally interchangeable. A UMPC is about the size of a paperback book and weighs less than two pounds. UMPCs feature displays of 4-7 inches and touch capabilities (using pressure-sensitive touch screens as opposed to the digitized screen input used in tablet PCs). Th ese devices also use the Windows Vista or Windows XP operating systems. However, their small screen size and hardware limitations currently make them inadequate replacements for tablet PCs for clinical computing tasks.

	In 2001, Bill Gates proclaimed that the tablet PC “is a PC that is virtually without limits, and within fi ve years I predict it will be the most popular form of PC sold in America.” Based on the device’s usability, such a prediction might indeed have seemed justifi ed (even considering the source). Yet even in healthcare, an industry that would seem ideally suited to adoption of the device, tablet PCs are not yet commonplace. Why not?

	Many experts, such as Joel French, vice president of Motion Computing and general manager of its Healthcare Business Group, appropriately emphasize that the hardware interface is only one component of the total computing experience. “It’s a mistake to think of [a tablet PC] as a discrete device,” he says. Successful design “has to encompass infrastructure, software, user needs, and hardware.”

	As an example, consider wireless networking. Mobile clinicians must be able to depend upon seamless and reliable network access. Without it, even the best hardware interface is unlikely to be used. Yet a survey by Spyglass Consulting found that 64% of nurses believe that currently available wireless infrastructure is not reliable enough to support point-of-care computing solutions.

	Similarly, software vendors must specifically tailor their interfaces to best exploit pen input. Motion Computing, in designing their C5 model specifi cally for use in healthcare, began working with software suppliers two years before the device was marketed. Recognizing the importance of the platform, clinical software vendors are increasingly developing applications that are optimized for use on tablets.

	In the age of HIPAA, security is a critical component of healthcare computing— especially for a device designed for mobility. According to Microsoft’s Martin, “Vista was a huge leap forward in making the tablet more secure.” Some tablets, such as Motion’s C5, also off er biometric user authentication.

	As for the device itself, battery life remains one of the most frequently cited limitations. Battery life for most current tablet PCs is at best about 3.5 hours. French rhetorically asks, “Should a clinician ever have to change a battery during a shift? No. Clinicians shouldn’t have to do things that detract from patient care.”

	In studying clinician workfl ow patterns, Motion found that clinicians often place their tablets on a desktop surface while doing other tasks, suggesting that the devices might instead be dropped into a docking station for a quick charge. When docks are provided in patient care areas in ratios of 2-3 docks per device, battery life becomes “a non-issue,” according to French. Another approach, of course, is to build a better battery. TabletKiosk, another manufacturer of tablet PCs used in healthcare, is now off ering an optional extended battery capable of 10 hours of continuous use in a wireless setting.

	Given the tablet PC’s evolution, it seems surprising that the devices are not yet more ubiquitous. The reason undoubtedly has more to do with the healthcare environment than it does with the device. As French puts it, “There’s a digital threshold below which it’s not worth migrating to a mobile device. As the threshold of digital information increases, [we] expect to see a multiplier effect.”

	Once healthcare exceeds that threshold (and it won’t take long), I suspect the tablet PC may yet prove to be “the information tool that may transform medicine.”

	As part of my research for this article, I decided to make hospital rounds with a tablet PC. The only device available to me was an aging fi rstgeneration device with an anemic battery. Given those limitations, here are my thoughts.

	I found the device to be unobtrusive. (I’m not even sure that all of the patients recognized it as a computer instead of a clipboard.) The pen interface worked pretty well for selecting patients and retrieving clinical results. (We have not yet implemented computerized provider order entry, so I can’t address data input.) It was fairly easy to transport, and it was convenient to have immediate wireless access to our enterprise electronic medical record. I did have to restrain my tendency to multitask between reviewing data and interacting with the patient so as not to distract my attention from the latter.

	My frustrations paralleled those of other clinicians. I intermittently lost my connection with the wireless network as I moved around the hospital. Each dropped connection resulted in having to go through several login screens to re-establish access. (Th is was obviously not a fault of the device.) Also, the tablet I used was powered by an old battery that didn’t last through rounds.

	My conclusion? Given a more contemporary machine with faster performance and better battery life, coupled with a more reliable wireless network, I suspect that I might make the tablet part of my routine. (Perhaps readers using tablets in other settings would share their experiences?)

editorial board member and an assistant professor of clinical medicine and associate program director, department of medicine, University of Illinois College of Medicine at Peoria.

Tablet PCs are computers powered by Windows operating systems that support functionality for input using digitizing screens designed to interact with a complementary pen.

G. Stephen Nace, MD

Articles

The Risk of Risk: Explaining Difficult Concepts to Patients

Where's the Love for Tablet PCs?