Risk Assessment, Risk Perception and Risk Communication - April 28, 2000

Kenneth Mundt, PhD, President, Applied Epidemiology, Inc.

DR. ERICSSON: Our last speaker is Kenneth Mundt, Ph.D., president and senior epidemiologist for Applied Epidemiology in Amherst, Massachusetts. He's also associate professor for the Department of Family and Community Medicine at the University of Massachusetts Medical Center. He's an abstract reviewer and occupational epidemiology session chair for the Society for Epidemiologic Research at their annual meetings. He's also a peer reviewer in Epidemiology Agency for the Toxic Substances and Disease Registry, also an adviser to the Occupational Medicine Residency Program at the U of Mass., and an epidemiology consultant to the Health Services Department at the World Bank. He has written an extensive number of publications and will today present on risk assessment, risk perception, and risk communication.

Dr. Mundt?

DR. MUNDT: Thank you.

I've been asked to talk about this topic of risk assessment, perception, and communication, and I want to thank the organizers for inviting me to do so because we've heard quite a lot today especially, and a little bit yesterday, about risks to travelers, and both with a focus on psychological risk and stress.

Risk itself is a commodity that we use constantly in epidemiology to identify risk factors, and presumably we identify risk factors so that we can intervene on these risk factors to reduce and eliminate risks. These are very, very basic points, but highly relevant to the conversations that we've heard today, and hopefully we'll convince you that these are not so easy but they're highly interrelated.

A few starting points, probably topics that are intuitive, and one of the nice things about talking about travel risk is that almost everyone is a traveler, almost everyone has experienced travel-related risks, and so we are all to some extent experts on the subject. And it's very easy then to take somewhat subjective impressions and even pseudo-scientific impressions—and I'm very much guilty of doing the same through some of our preliminary research, and that's appropriate at an early stage, but ultimately we need to pin down and quantify risks at an individual level so that they can be utilized for various purposes.

I think you all agree that risk shapes our behaviors, our decision making, and, ultimately, although we may not understand the mechanisms of this, our health.

Perception of risk is highly complex and is not simply based on scientific evidence but, rather, strongly based on our particular individual and cultural values. This has been touched upon by previous speakers.

We aim to manage risk and use it to our advantage because there are some benefits to be had. We also want to minimize risks where they are unnecessary or where they are unacceptable.

Individuals—that is, perhaps, the business travelers—and their organization might face exactly the same quantifiable and qualitative risks, but they'll perceive them and manage those risks from very different perspectives, probably because there are different vested interests from these perspectives.

Communicating risk is subtle. Communicating, as I'll mention again later, is not simply broadcasting what we find numerically or epidemiologically, but it does involve communication and it's a two-way street.

This may be a bit dark—it is a bit dark—but here's a welder sitting on a building being demolished, cutting a beam, and I think he may be sitting on the same beam. And when we look at a worker—and my focus and specialty is in occupational health—it doesn't take a great imagination to understand that this employee is at risk.

Now, what are the risks? Well, there are things that are very obvious: physical hazards. There are perhaps also less obvious risks having to do with exposures to welding or cutting fumes and so on. But I think we have to think about risk in all contexts, and if there is a common basis for this, it's our perception of what is at risk.

What is this employee's perception of risk? Well, it's apparently not great enough to turn down this assignment, or the reward structure is great enough to overcome that employee's perception of risk.

What is the risk and liability to that employee's employer? What are the real risks to this employee's family and friends?

What about a very different work setting? Are the actual physical risks here less severe? Probably so. Maybe there are some risks associated with filing work. Maybe there are some risks with sharing an office with three or four other individuals and lack of control over work flow and so forth. But perhaps the perception of risk in this context is very different. Maybe these employees don't perceive any risks in this environment because they don't perceive risks that aren't physical as important occupational health risks.

So my objective here, given a tall order and not a lot of time, I want to just review—perhaps because most of you know it and see it, but it's easy to forget—the basis for risk assessment in epidemiology. The previous two talks were very good, presented very good material for this clarification because we used the term "risk" quite often. We talk about risk. We talk about incidents. We talk about percentages. We talk about risks in terms of one out of a thousand, one out of a million, and what does that mean to us and to the people we work with. It has its roots here in epidemiology and needs to be understood a little bit so that risk perception can be understood a little bit better from various points of view as well as then the task of communicating risks, both at the individual and group level.

What is risk? Well, it's easy to turn to the dictionary, and I went to my giant Webster International and it listed this definition: the possibility—and that's my emphasis here, the bold—of loss, injury, disadvantage, or destruction. The possibility. A risk is a possibility. It's something that might happen.

Dr. Last has edited a dictionary in epidemiology, and we affectionately refer to it as "the last word" in epidemiology. But the definition for risk in this text is the probability that an event will occur. And it's very similar to the Webster definition, but it takes it a step further. We're now taking a possibility and describing it as a probability which can be quantified and estimated using various objective techniques.

So we, I think, in everyday life, both qualitatively and quantitatively, describe risk. We might ourselves tell our children, oh, that seems to be too risky, why don't you get off of that balcony railing? Where we don't need to quantify it, it's effective enough to say that that exceeds some threshold of tolerance of risk, and qualitatively we think that it needs to be intervened upon.

It is very difficult to quantify that same risk, that is, the child falling from a balcony railing. It can be done. It can be estimated a number of ways. But the specific quantification of the risk for that individual, that child, is nearly impossible. In fact, the true probability of risk for any individual of any outcome is either 0 percent or 100 percent over a certain period of time. Either I will get traveler's diarrhea, however you define it, or I will not get traveler's diarrhea on my next international mission. Either I will have an accident or I will not, and the probability for any individual in any group is going to be 0 or 1 at the individual level.

And so this forces us to look at risk and probability based on groups, and so all of the studies previously have shown that risks have to be estimated based on groups and events that occur within groups, and then we have to decide whether it's fair game to apply the risk we estimate from any group to any individual, whether that individual is part of that group or some other population.

A very basic diagram illustrating five people, which could really be 500 or 5,000 or a million, individuals at a starting point in time, time zero, who are at risk, indicated by the green color, for some disease or outcome. And to estimate risk in this population of five, we follow those individuals to determine which ones over a specified period of time contract the disease or outcome of interest.

So here are very basic steps. Individuals 1 and 2 and 4 go from the beginning of observation time to the end, and this is a 24-month period, a hypothetical example, without contracting that disease. Individuals 3 and 5, however, do contract the disease; 3, it looks like, contracts it after one month and No. 5 after 12 months. And so in this population, we had five individuals at risk at the beginning of a specified time periods. Two individuals contracted the disease during that time period, and so we can very simply calculate risk as what epidemiologists call a cumulative incidence. Simply, the number of people getting the disease, in this case two, the number at risk at the beginning of the risk period, that was five, and over the period of time followed, two years. And so from this we get a very simple derivation that 20 percent, or 0.2, per year got the disease.

This is a proportion, and the proportion can be interpreted, although very cautiously, as a probability and, therefore, we can interpret this then as a risk. Cumulative incidence in epidemiology is synonymous with risk.

Let me go back to that previous slide, if I can. Look at this simple example again. What if this individual in this case who contracted a disease or an event after one month recovered but didn't have the immunity that I've indicated here? That individual would return back to the risk pool and would be at risk again for yet another disease. So you can imagine the extension of this very basic principle to a real dynamic population can be quite tricky, and we have ways of doing that. It's not so bad.

Next, what if we had calculated our estimate of incidence at different points in time here? What if we had looked at this population in a cross-sectional way? Well, toward the end, we actually have one individual in the second year who contracts the disease out of really only four who were at risk because there was one who had come in with some conferred immunity. So on top of this, you can imagine a dynamic population with people coming and going. This is very difficult to equate this very pure concept of probability and risk derived epidemiologically to what we might get in some of the studies that we all have conducted, our studies of insurance claims rates, for instance, our surveys, our cross-sectional. What we are deriving from these is a bit different from risk and is subject to some other determinants.

Nevertheless, I believe that at this stage in the game it's very important to conduct the kinds of research that we are that's totally descriptive and that we can move toward something that's more analytical so that we can quantify better the actual risks so that we can be more effective at intervening. That's not the point of my talk today.

If we're going to derive an estimate of risk from an epidemiological study and then use it for some public health purpose like intervention or prevention, we have to be sure that the estimate we derive is actually applicable to the target population. We can estimate claims rates and risks—risk ratio estimates from claims rates, and a good question yesterday was, well, how generalizable are the results from these studies? Well, they're not at all generalizable unless you can assume or document that there's good comparability between those populations.

Age and gender, huge determinants of risk for various factors, education, cultural factors and so on, depending on what kind of outcome. Clearly, the exposure, health behavior, tolerance, resilience, coping techniques are all going to factor into whether or not a particular population's risk is applicable to any other population's risk.

Also, keep in mind that because we have to study populations in the past in order to determine risk, which is a prediction for the future, there could well be population changes in the risk factor status or in behaviors or in practices over time. And so we have to even assume that there isn't a meaningful change over time in those underlying risk factors. The applicability of a risk derived in some single population may not apply to that very same population several years later, especially if there are changes that occurred related to that risk.

If you can imagine, now take that context of applying the risk estimates from an epidemiological study to a population and extend a step further to any individual in that population. First of all, individual risk is going to be either 0 or 1. Individuals in that population have to determine whether they are a reasonable member of that group and whether that group's assessed or estimated risk is applicable to themselves, and then they have to interpret whether that risk is high enough to modify one's behavior or whether the trade-off between benefit and risk is such that they want to change, modify the behavior.

When we apply a risk estimate from a study or population to an individual, it's going to be a function of what we know about that disease. What are the other risk factors for that disease? Once we understand other risk factors for the same outcome, we can continually factor them in, look at them simultaneously, and then perfect our estimates of risk for that population and, therefore, to those individuals.

If an individual faces a risk, changes a behavior, and thereby doesn't suffer from that outcome or disease, it doesn't mean that that individual is now immune from other diseases. We are constantly facing competing risks and that the modification of one risk doesn't necessarily imply the modification of other risks, and in some cases, the elimination of one risk can actually increase the risk for another outcome or disease.

Finally, the inability to predict the future accurately will always stay with us when we're trying to interpret risk and assess it quantitatively. We don't know what will happen. We can only judge and predict or guess what will happen based on what we've observed in the past already.

If we try to separate the notion of risk into a group level or corporation level to an individual level, we might begin to see that there are different motivations for understanding, assessing, and interpreting risk. First of all, a corporation might want to know, well, this is a reasonable risk of, say, exposure to a certain chemical that we might use, but is it going to increase the numbers of people with a measurable outcome? How great is that risk in terms of number of employees potentially affected?

If it turns out that there's no substitute chemical for this and there are safe handling procedures and the risk is reasonably low and it can be assured that there will not be very many, if any, cases, it might be seen as an acceptable risk. To the individual handling that substance, especially if the safe handling procedures are very complicated, that might not be an acceptable risk. And for that individual, it would be a very different perspective. What is the benefit gained in exchange for that risk?

The organization or corporation might look at other aspects like severity, the impact on the organization. It might look more at the costs per case. Again, not major concerns of the individual, who have to ultimately understand by ranking and weighing—and this is something we do on an everyday basis—our understanding of the multiple risks.

This is really the segue into risk perception. Individual understanding of risk and ability to rank and weigh risks and, therefore, make decisions has to do with an individual's perception of those risks.

Risk perception will be a function of knowledge and education, but not necessarily. Some of those that know the risks best are at greatest risk because they ignore them or they think that they're the 97 out of 100 that won't be afflicted by that disease. It very much varies by gender, by race, by age, and many other cultural and societal factors summarized as values.

We choose risks based on what we value and what we don't value, and we're motivated by those, and those are highly individual.

Here's a result from a survey done about six, seven years ago, a survey of 1,500 U.S. adults, and they were asked 25 questions having to do with: How would you rank such-and-such in terms of risk to society? And for both men and women in this sample, these five risks to society were ranked the highest; in other words, the highest proportion of respondents considered them a high-risk category, and the percentage here suggests that 58 percent of men considered cigarette smoking the highest risk to society out of the list of 25 items. I only had to go down to the top five to be able to include stress.

What's striking in this survey, from top to bottom on this list, men perceived all of the risks on average one grade of severity lower than women. There was no exception. And so we see here large differences in the perception of risk as the societal level for each of these as well.

Determinants of risk are being studied, and it seems that—it has been mentioned earlier—one of the key phrases is: Who controls the risk? Who has control over the decisions they make and, therefore, the risks that they take? Is it the individual? Is it some system that they're part of—their employer, the government? Or is it some unknown entity?

It seems that risks of earthquake and of other natural disasters aren't perceived as such a great risk because the unknown entity is beyond control. It's accepted. It's a fatalistic view that we tend to have. But if it's fear of a plant explosion or a spill of a chemical, those risks, though they might be comparable to the risks of a natural disaster, are perceived much greater because they are unnatural.

Is the source of the risk information trusted? We're getting closer now to the risk communication. What are our sources of risk information? If a physician says that you ought to change this behavior and you trust this physician, you might actually change your behavior instead of a salesperson who's trying to get you to buy the product which helps you change that behavior.

We understand that there are motives behind the advice and recommendations that are given, and I think it's an important point to remember in risk communication. It's very difficult to know if risks are discussed openly, but in any situation where the environment can be made open, there will be some risks in doing so as an organization, but it will improve the ability to communicate risks.

The risk perception, again, if you're working in an environment where you're known by a number and probably listened into—this is a bank of phone operators—might be very different from other populations. This is a child picking cotton. We have many, many occupational risks where the individuals who are at risk have probably very little perception of the risks that they face, and to communicate to these individuals requires different skills and approaches than you might to an international traveler.

Other factors influencing perception of risk have to do with whether the risk is entered voluntarily or not, whether it's natural, whether it's familiar versus exotic, whether it's dread of not. Risk of cancer might be lower than risk of other disease, but are dreaded more in our society for some reason.

The risks can be perceived differently, but an important point is that risks to employees, whether they're real or perceived, are indeed, directly or indirectly, risks to the corporation.

Why communicate risk? Who should communicate what and to whom? What's the intended result? A lot of my work involves research in chronic diseases in persons exposed to hazardous chemicals. And once the studies are done, it's often easy to forget that the people who are handling these substances need to know what are the risks they face. And then the question comes about: When, how, to whom should these be communicated? And these are very good questions to ask, but a main point is that communication implies that both the risk communicator and the person to whom or with whom the communicator is communicating ultimately understands that risk to the degree that their perception is largely based on factual information and is tempered by the proper context of that risk.

Risk communication is done because we believe this is a first and important and essential step toward risk reduction. We generally value health, and so we tend to be motivated to improve our health by reducing risk. We may not know what an effective intervention is to reduce risk, and that may not need to stop us from communicating those risks. Individuals need to know about them, need to make their own decisions about them, whether or not effective interventions are observed.

Here are some principles of risk communication that were derived from a conference a couple years ago of a group of environmental scientists. The stakeholders, as I mentioned before, need to be involved. The quantitative risk measures, the one in a million, the one in a thousand, almost always have to be translated to the persons that are affected by those risks or to whom they apply.

The information on the technology and the health may be inadequate on its own and that the methods for communicating risk must match the message. Broadcasting study results doesn't assume or assure good communication of risks. Many different audiences may need to be targeted in communicating risk, and the impact of that communication has to be evaluated to see if it has been effective at all.

Psychological risks, back to the theme of this conference, are especially difficult to communicate, and I think here there are other experts that are suited to discuss this point. But I think this emphasizes the multidisciplinary nature of our work. We need to derive risks scientifically, epidemiologically, but we really as scientists need assistance in communicating those risks, especially if they have to do with stress.

So managing health risks is important to individuals, corporations, and society. We want to keep in mind that we we deal with those differently. Effective communication of risk will facilitate risk management, both at a corporate level and at an individual level. Fortunately, risk perception can be shaped, and I think that the risk perception, the shaping of risk perception is going to be more effective than beating our audiences over the head with risk numbers. We generate these risk numbers, and so we don't effectively communicate them in most contexts.

But the accurate source of risk information is going to depend on high-quality research in whatever field it is and in the work that I do that's going to be epidemiological research.

Thank you.

[APPLAUSE]

Disclaimer: These Proceedings have been produced from transcripts made from audio tapes. Efforts were made to check the accuracy of information with the various authors, but this accuracy is not guaranteed. If there is information that you believe requires correction, please send a message to our e-mail address.