Reducing exposure to electromagnetic fields: the effects of low- and high-threat risk messages on behavior change.

Introduction

Before the industrial revolution, the principal hazards faced by human beings were infectious diseases, floods, famines, and other natural events. Increasingly, however, society has been faced with hazards of its own creation. Dangers arise from the failures of technology and other innovations (as in the events at Three Mile Island and Chernobyl) (1). Indeed, the possibility of negative effects from modernization and technology has become a significant public health concern. Accompanying this change has been an increase in attention to the manner in which health risks related to technological hazards are communicated to workers and the public.

The ultimate goal of a health-related risk communication message is usually behavior change. For example, motivating people to evacuate after a nuclear-power-plant release might involve the following:

* making them aware of the dangers of ionizing radiation,

* changing their attitudes about their own susceptibility to risk, and

* convincing them that they can successfully reduce their own exposure.

The process by which scientists communicate to the public about risk continues to be an important and complex issue. In a review of risk communication research, Covello et al. (1986) discuss several categories of problems that arise in the communication of risk information (2). Problems can arise from

1. the source of the risk message (scientists, government officials, company presidents);

2. the way information is channeled (television, newspapers, radio);

3. the receiver of the risk message (other risk experts, laypersons); and

4. the content of the message (scientific information, emergency information).

Message problems have been described as a major bottleneck in the process of communicating information about technological risks, possibly because of the levels of complexity and uncertainty inherent in scientific information (3).

There has been much research on how individuals perceive risk, how people make decisions about risks, and how risks are quantified by experts. Despite this impressive volume of research, there has been little exploration of how to communicate risks in a manner that creates an adaptive behavioral response (4).

To effectively evoke an adaptive behavioral response, a risk message must contain both a threat component (to create a perception of personal susceptibility and severity) and an efficacy component (to provide information about measures an individual can adopt to reduce the threat). Unfortunately, the construction of effective risk messages is seldom as straightforward as ensuring the presence of these two important elements. Several factors work together to make the design of such messages difficult, including

* deficiencies in scientific knowledge,

* lack of expert consensus, and

* the public's concern with qualitative risk factors (such as "voluntariness" and "dread") when it makes risk judgements.

Electromagnetic fields (EMFs) have been the subject of increasing concern and controversy. Electric and magnetic fields are produced by the generation, transmission, and use of electricity. In fact, individuals are exposed to EMFs whenever they turn on light switches or use computers, toasters, or other electrical appliances. Even in the absence of scientific certainty, Morgan (1989) contends that risk experts have an ethical obligation to inform the public that exposure to EMFs may have adverse effects on their health and that they should avoid such exposure whenever doing so is economically, logistically, and behaviorally practicable (5). Witte (1994) notes that risk managers may prevent panic responses in cases in which risks are uncertain by

1. acknowledging this uncertainty and

2. offering specific steps that individuals can take to effectively reduce the occurrence of the threat or minimize harm from the threat should it occur (6).

Two characteristics of EMFs make effective risk reduction strategies difficult to plan. First, some research has found that EMFs may not follow a typical dose-response relationship, meaning that reducing exposure in some situations could actually lead to increased risk of harm. Second, it is currently unknown which specific aspect of exposure to EMFs is most strongly linked with health effects - average exposure time or peak field strength. Studies have found that both factors may influence biological effects.

The purpose of this study was to test low- and high-threat risk messages about EMFs and to determine whether one type is more successful in motivating individuals to make adaptive behavioral response changes.

Methods

Subjects were asked to complete attitude questionnaires immediately after reading a risk message concerning EMF exposure. Each subject received either a low- or high-threat risk message. Both messages included an explanation of the properties of EMFs, a brief summary of the findings of epidemiological studies regarding EMF exposure and health effects, and an acknowledgement of scientific uncertainty regarding exposure assessment. Both messages also discussed the source of EMFs and the nature of the hazard, and suggested that exposure levels can be reduced simply by increasing one's distance from electrical devices.

Although the messages were similar in many ways, the high-threat message was designed to produce an intense perception of severity and susceptibility, while the low-threat message was designed to produce less acute perceptions of severity and susceptibility. The low-threat message informed the reader that some EMFs occur naturally (caused by the earth's magnetic fields, for example) while the high-threat message did not discuss this natural source of EMFs and instead emphasized that EMFs are invisible and cannot be detected by humans. Unlike the low-threat message, the high-threat message used the term "radiation" and presented exposure estimates for a typical home and for individuals in close proximity to electrical household appliances. The high-threat message also mentioned that some studies have linked childhood cancer to EMFs, whereas the low-threat message stressed that the literature on EMF health effects was inconclusive.

Subjects

The subjects were 251 college undergraduates who had been offered extra credit for participating in the study. They were told that educational materials about EMFs, targeted for college students, were being developed and that their honest opinions and responses were essential so that the materials could be further refined. The subjects were assured anonymity and chose their own secret codes, which were used to match the preliminary and follow-up questionnaires. Each subject was randomly assigned either the low-threat or high-threat message, and both groups were also provided with a take-home copy of a list of recommendations for reducing EMF exposure (e.g., increase distance between self and electrical appliances, move electrical appliances away from bedside, reduce use of electric blankets).

TABLE 1 Respondent Demographics

Gender                    n                 %

Male                     81               32%
Female                  165               66%
Missing                   5                2%

Ethnicity

Caucasian                53               22%
Hispanic                 35               14%
Asian                   136               54%
Other                    22                8%
Missing                   5                2%

Status

Freshman                121               48%
Sophomore                36               14%
Junior                   56               22%
Senior                   33               13%
Missing                   5                2%

Marital Status

Single                  237               94%
Married                   5                2%
Other                     3                1%
Missing                   6                3%

Number of Children

None                    237               94%
One                       4                2%
Two or More               2                1%
Missing                   8                3%

Age

Average: 22 years
Range = 18-39 years

Two weeks later, subjects completed a follow-up questionnaire that had been developed to measure attitudes and behavioral change. The time constraints of the 10-week school term did not allow for a greater lapse, and a two-week period was judged to be adequate for the subjects to modify their behavior to reduce their EMF exposure.

The low-threat message was read by 115 subjects; 136 subjects read the high-threat message. The sample was predominantly female (66 percent) and single (94 percent). Just over half (54 percent) were Asian, and approximately half (48 percent) were in their first year of college. The subjects ranged in age from 19 to 39 years; the mean age was 22 years (Table 1). The generalizability of the Study should be assessed with caution. Although studies have shown that college students perceive risks like other laypersons, there has been little evaluation of ethnicity and the perception of risk from EMFs (7).

[TABULAR DATA FOR TABLE 2 OMITTED]

[TABULAR DATA FOR TABLE 3 OMITTED]

[TABULAR DATA FOR TABLE 4 OMITTED]

Measures

In addition to the demographics discussed above, the questionnaire also asked for proximity of home to power lines and time per day spent using electrical appliances. The average distance from a subject's home to the nearest power line was 348 feet, and the average subject spent roughly 2.5 hours per day using electrical appliances. There were no significant differences between those who read the low-threat and those who read the high-threat messages either for proximity to power lines [t = 0.49, not significant (NS)] or time spent using appliances (t = 1.43, NS). The questionnaire measured eight important variables:

1. Susceptibility refers to the level of threat subjects perceived from EMFs - that is, whether participants thought EMFs would be harmful to their health (e.g., "It is likely that exposure to electromagnetic fields will be harmful to my health"; [Alpha] = 0:75). (Cronbach's alphas are reported throughout this article to provide a scale of reliability for each index). The index items were modeled after those utilized by Witte, (1991) (8,9).

2. Severity, a second measure of EMF threat level, refers to whether subjects felt health effects related to exposure to EMFs were "severe," "significant," and "serious" ([Alpha] = 0.91).

3. Response efficacy refers to beliefs regarding utility of exposure reduction techniques in overall health schema (e.g., "Reducing exposure ... is effective in preventing unwanted health effects"; [Alpha] = 0.82).

4. Self-efficacy refers to the perceived difficulty of implementing adaptive behaviors (e.g., "Measures to control exposure ... are easy to use"; [Alpha] = 0.72).

5. Fear refers to how "scared," "frightened," and "anxious" subjects were about health effects from exposure to electromagnetic fields ([Alpha] = 0.83).

6. Attitudes refers to whether subjects thought it would be desirable to implement the recommended adaptive behavioral responses ([Alpha] = 0.86).

7. Intent refers to subjects' actual plans to implement the recommended control measures ([Alpha] = 0.94).

8. Current use of recommended control measures to reduce exposure to EMFs was the eighth variable ([Alpha] = 0.86).

Perceptions about the accuracy, objectivity, comprehensibility, and interest of the presentation were also measured, with one item each (e.g., "This message was clearly written").

Results

Bivariate Results

Threat and Efficacy Components

How serious one perceives a threat to be is an important factor in the decision to change behavior. The subjects' perceptions about the threat from EMFs were separated into three components: fear, susceptibility, and severity. As expected, readers of the high-threat message scored higher than those who read the low-threat message, as follows:

* fear of EMFs (t = -3.01, p [less than] .01);

* feelings that health effects related to EMF exposure are severe (t = -7.95, p [less than] .0001); and

* belief that EMF exposure was likely to increase susceptibility to harmful health effects (t = -3.23, p [less than] .001).

No significant differences existed between the groups on response efficacy (t = -1.92, NS) or self-efficacy (t = -1.01, NS) (Table 2). Also, no significant differences were found between the groups with respect to clarity of the message. These findings indicate that the impact of the messages varied by threat and fear levels, but not by efficacy or clarity.

Attitudes, Intentions and Behavior

Readers of the high-threat message were significantly more likely to agree that reducing exposure to EMFs is a sound idea (t = -1.98, p [less than] .05). Interestingly, there were no significant differences between low- and high-threat message groups in intent to implement the control measures recommended (t =-1.58, NS). Similarly, there were no significant differences between the groups in current use of the control measures (t = 0.04, NS) (Table 3).

Follow-Up Results

Subjects' perception of the threat presented by EMFs declined significantly over time; subjects' fear of EMFs (t = 2.61, p [less than] .01), feelings about the severity of health effects related to EMF exposure (t = 1.94, p = .05), and feelings that EMF exposure increase susceptibility to harmful health effects (t = 2.08, p [less than] .001) decreased from Time 1 to Time 2. Subjects were less likely at Time 2 to agree that they intended to implement control measures to reduce their exposure (t = 2.70, p [less than] .01), but their existing use of recommended control measures did increase significantly (t = -4.42, p [less than] .0001) (Table 4). These findings show that the effects of risk communication messages "wear off" over time, including the intent to make adaptive behavioral changes (although subjects were, in fact, more likely to actually implement the recommended behavioral actions at Time 2 when compared to Time 1).

As expected, the threat level of the message was an important factor in change. Nearly half (46 percent) of those who read the high-threat message implemented controls compared to only nine percent of the low-threat readers ([Phi] = 0.36, p [less than] .0001).

Multivariate Results

Most (65 percent, or 104) of the subjects reported making at least one change; the remaining subjects said they made from one to 11 changes. Because this distribution of the data was skewed, the data were dichotomized and a logistic regression analysis was performed. The effects that several independent factors had on whether the readers made any of the recommended behavioral changes were regressed on the dependent variable, whether the subjects implemented at least one of the suggested techniques to reduce their exposure to EMFs. The independent variables included

* type of risk message (low versus high threat),

* fear (those who are more afraid should be more likely to desire to protect themselves from harm),

* attitude about reducing exposure to EMFs (the stronger one's belief in the desirability of reducing EMFs, the more likely one should be to make recommended changes),

* strength of belief that health effects related to exposure to EMFs are severe (another motivational factor), and

* level of belief that the message was intentionally manipulative (a control factor - those who believe the risk message to be manipulative should be less likely to implement any changes).

[TABULAR DATA FOR TABLE 5 OMITTED]

Additionally, response efficacy (those who feel that control measures would be effective in reducing adverse health effects should be more likely to implement them) and self-efficacy (subjects who feel that control measures are inconvenient should be less likely to implement them) were entered into the equation. One additional factor - whether participants thought EMFs would be harmful to their health, could not be included in the multivariate equation because of multi-colinearity with three other variables (fear of EMFs, severity of ill health effects, and response efficacy).

It appears that behavioral change to reduce exposure to EMFs was related to threat level and to the belief that using control measures would reduce adverse health effects (Table 5). Those who read the high-threat message were nearly seven times more likely to make changes, compared with those who read the low-threat message [odds ratio (OR) = 6.76, p [less than] .01], even when other factors were controlled.

Similarly, response efficacy played a role in adaptive behavioral change. Those subjects who believed that reducing their exposure would prevent adverse health effects were much more likely than their less certain counterparts to implement the recommended changes (OR = 1.9, p [less than] .05).

No other variables significantly affected the subjects' use of adaptive behavioral changes in this model. In other words, the subjects' individual levels of fear about EMFs, their beliefs regarding the severity of the harms attributable to EMFs, and their feelings about whether the risk message was intentionally manipulative had no significant effects in this model. Ironically, their attitudes toward change and their beliefs about self-efficacy also had no significant impacts on their choice to implement the recommended control measures.

A logistic regression analysis controlling for demographic factors (e.g., age, race [white vs. nonwhite], gender); level of concern about EMFs before reading the risk message; and use of control measures before participating in the study did not yield any significant differences in results.

[TABULAR DATA FOR TABLE 6 OMITTED]

The factors that had significant effects in this model were those that could easily be manipulated by designers of risk messages. Ensuring a high level of threat and simultaneously demonstrating the ease with which individuals can avoid negative health effects appears to be the most important way to create adaptive behavioral change.

Effect of Implementing Change on Subjects' Fear Levels

One interesting question concerns whether engaging in adaptive behavioral change serves to reduce subjects' levels of fear regarding a given threat. Presumably, individuals could lower their level of fear of adverse health effects by implementing the suggested controls. This idea was not supported in this study, however. Among those subjects who implemented controls, level of fear did not change significantly over time (t = 0.18, NS), while it decreased among those who failed to make changes (t = 2.77, p [less than] .01) (Table 6). In addition, the level of fear was higher at both time points among those who ultimately made 'changes (t = -3.84, p [less than] .0001 for Time 1; t= -4.69, p [less than] .0001 for Time 2) (Table 6). In other words, fear began at higher levels and remained at high levels for those who made changes, while it decreased only for those who ultimately failed to make any of the recommended changes.

Discussion

Because experts and laypersons process risk messages differently, it is critical to find the elements of risk communication messages that promote self-protective and adaptive behavior. The results of this study suggest that effective risk messages should promote high levels of perceived threat and fear and at the same time promote high levels of response efficacy.

Response efficacy was more important than self-efficacy in evoking adaptive behavioral change. This may be due to the fact that EMF exposure represents an unknown risk whose effects cannot be easily understood by the public (unlike the established link between seat belt use and level of injury in auto accidents). Additionally, self-efficacy includes measures of the convenience of implementing control actions. Since most behaviors that reduce EMF exposure are fairly easy to implement, self-efficacy may have played a much smaller role in influencing change in the current study. It is possible, that self-efficacy is important in situations in which risks are known or in which adaptive behavioral changes are more demanding.

Readers of the high-threat message were much more likely than those who read the low-threat message to make changes, even when a host of additional factors were controlled. There were no significant differences between the two groups resulting from demographic or other factors. Nevertheless, the homogeneous nature of the participants may have affected the findings of this study. Future studies should sample a more heterogeneous population to rule out any potential biases. The findings of this study, however, indicate that effective messages need to be clear about threat. Risk messages may have failed in the past because the perception of threat was not present or was not strong enough to induce change. The cases for home radon testing and reduced sunbathing, for example, have failed because people do not believe these hazards are threatening (10).

Risk messages also must clearly state what actions can be taken and exactly how these changes will benefit the recipient of the message. It is critical to note that messages that increase perceived threat and fear but do not address response efficacy run the risk of producing maladaptive outcomes.

Corresponding Author: Shari McMahan, Assistant Professor, School of Social Ecology, University of California Irvine, Irvine, CA 92717.

REFERENCES

1. Baker, F. (1990), "Rick Communication about Environmental Hazards," Journal of Public Health Policy, 3(3):351-360.

2. Covello, V.T., P. Slovic, and D. Von Winterfeldt (1986), "Risk Communication: A Review of the Literature," Risk Abstract, 3(1):171-182.

3. Weterings, R.A.P.M. and J.C.M. Van Eijndhoven (1989), "Informing the Public About Uncertain Risks," Risk Analysis, 9(4):473-482.

4. Witte, K. (1994), "Generating Effective Risk Messages: How Scary Should Your Risk Communication Be?" Communication Yearbook, 18(1):229-246.

5. Morgan, M.G. (1989), Electric and Magnetic Fields from 60 Hertz Electric Power: What Do We Know About Possible Health Risks? Pittsburgh, Penn.: Carnegie Mellon University, Department of Engineering and Public Policy.

6. Witte, K. (1994), "Fear Control and Danger Control: A Test of the Extended Parallel Process Model (EPP)." Communication Monographs, 61(2):113-134.

7. Slovic, P., B. Fischoff, and S. Lichtenstein (1979), "Rating the Risks," Environment, 21(3):14-39.

8. Witte, K. (1991), "Preventing AIDS Through Persuasive Communications: Fear Appeals and Preventive Action Efficacy," Irvine, Calif.: University of California Irvine, Unpublished Doctoral Dissertation.

9. Blalock, H. (1960), Social Statistics, New York:McGraw-Hill, pp.8-9.

10. Sandman, P.M. (1987), "Risk Communication: Facing Public Outrage," EPA Journal, 13/14(13):21-22.