Importance of risk comparison for individual and societal decision-making after the Fukushima disaster.

Risk comparison is essential for effective societal and individual decision-making. After the Fukushima disaster, studies compared radiation and other disaster-related risks to determine the effective prioritizing of measures for response. Evaluating the value of risk comparison information can enable effective risk communication. In this review, the value of risk comparison after the Fukushima disaster for societal and individual decision-making is discussed while clarifying the concept of radiation risk assessment at low doses. The objectives of radiation risk assessment are explained within a regulatory science framework, including the historical adoption of the linear non-threshold theory. An example of risk comparison (i.e. radiation risk versus evacuation-related risk in nursing homes) is used to discuss the prioritization of pre-disaster measures. The effective communication of risk information by authorities is discussed with respect to group-based and face-to-face approaches. Furthermore, future perspectives regarding radiation risk comparisons are discussed.

INTRODUCTION

The methodology of risk comparison was conceived in the 1660s by Graunt, and has been beneficial for both individual and societal decision-making [1]. Since then, several risk indicators, including cancer risk, loss of life expectancy (LLE), quality-adjusted life year (QALY) and disability-adjusted life year (DALY), have been developed and used to compare multiple risks [2-4]. Starr analyzed the mortality risks and benefits of multiple technologies, and demonstrated public acceptance, including acceptance of a nuclear power plant [5]. Gamo et al. ranked the risks of 12 major environmental pollutants, including radon, using the same scale indicator, LLE [6]. Murray et al. evaluated global DALYs for 291 diseases from 1990–2010. These risk comparisons are useful for prioritizing measures as part of societal decision-making. Specifically, a risk trade-off analysis is required in order to minimize risks in society [7].

Risk comparison is also used to deliver information to the public as part of risk communication, because the concept of risk is regarded as the most difficult skill among all numeracy elements [8]. Lay people and even expert interpretations of risk information include various heuristics and depend on the communication method [9-12]. Specifically, low-probability risk is generally overestimated [13], and comparison information is necessary in order to accurately judge the differences between low-probability risks [14]. It is known that a lack of risk comparison information leads people to overestimate one specific risk [15]; therefore, risk comparison is also important for individual decision-making.

The Fukushima disaster in 2011 posed multiple risks: including radiation [16-18]; indirect health effects such as evacuation from nursing homes during the acute phase (i.e. days to months after the crisis) [19, 20]; lifestyle-related diseases [21]; and psychological distress [22] during the chronic phase. The number of earthquake-related deaths (deaths indirectly associated with the earthquake) in Fukushima increased to more than 2100 by 31 March 2017 [23]. Comparison of these risks at the same scale indicator is necessary in order to understand the levels of multiple risks and to prioritize measures; however, to the best of my knowledge, this approach has been limited [24] by a lack of interdisciplinary scientific research. In addition, for assessment of radiation risk at low doses, there are arguments regarding the validity of cancer incidence or mortality risk assessment [25], suggesting that it is necessary to clarify the concept of ‘radiation risk’.

A radiation risk comparison approach using units of dose (mSv) has been used to communicate risk information by several experts and institutions (e.g. through a comparison with standards, natural and artificial radiation doses, other chemicals, and smoking risks [26-28]). Risk comparison approaches, which aim to support individual decision-making such as risk acceptance or risk avoidance, including food consumption behavior and returning to one’s hometown, has often been criticized, especially in terms of the Covello’s guidelines [29]. However, these guidelines were originally based only on personal experience, and application of the guidelines to different cultural situations is poorly understood. It is necessary to demonstrate the effects of risk comparison information on residents’ understanding of risk or decision-making.

The objectives of this review are 3-fold. First, the concept of radiation risk assessment at low doses is clarified under a regulatory science framework, taking into account its historical development [30]. Second, after introducing examples of non-radiation health risks in the aftermath of the Fukushima disaster [19, 31], a radiation risk assessment is used to compare the radiation risk with another risk, namely the evacuation risk from nursing homes, as an example of tools for societal decision-making [24]. Third, to evaluate the value of risk comparison for individual decision-making, the effects of risk comparison information on multiple outcomes are assessed, including the level of subjective and objective understanding and confidence in the information [32]. The importance of risk communication for risk trade-off problems is also discussed [33]. Finally, future perspectives regarding radiation risk comparisons are discussed.

LOW-DOSE RADIATION RISK ASSESSMENT

Radiation exposure attributable to the Fukushima accident is limited to a lifetime effective dose of ~30 mSv in evacuation order areas and 2 to <20 mSv in other areas in the Fukushima Prefecture [16, 34, 35]. Epidemiologically, cancer incidence or mortality under such low doses has been well studied. Epidemiological follow-up studies showed no indication of genetic effects including congenital malformations, stillbirths, early mortality, or cancer increases in the offspring of atomic bomb survivors, who received several hundred mGy [36, 37]. Epidemiological studies on atomic bomb survivors showed no significant increase of cancer incidence for doses below 100 mGy, whereas a significant increase was observed for doses >200 mGy [38]. An increase in the radiation-related risk of non-cancer outcomes (e.g. cardiovascular disease) was recently reported in atomic bomb survivors at higher doses [39]. Thyroid cancer incidences in epidemiological studies after the Chernobyl accident were also consistent with model estimates based on the dose–response results of atomic bomb survivors [40]. Other epidemiological studies, including medical and occupational exposure, have shown contradictory findings regarding cancer incidence and mortality rates below 100 mSv of additional doses [41]. Furthermore, epidemiological studies on natural radiation exposure have also shown inconsistent findings in cancer incidence rates [42, 43]. The results of epidemiological studies in atomic bomb survivors were regarded as the most informative by the United Nations Scientific Committee on the Effects of Atomic Radiation [44], because of the large number of data, the accuracy of the dose assessment, and the length of the follow-up period. It should be noted that ‘no statistically significant increase in risk’ does not prove an absence of risk.

The methodology of radiation risk assessment depends on the presence of a threshold. The International Commission on Radiological Protection (ICRP) defined the threshold dose as 1% of incidence [45]. Even if there is a threshold and the dose is below that threshold, it does not prove there is zero risk to the public. Historically, the cancer risk at low doses was described in ICRP Publication 1 in 1958 [46] as a de facto linear non-threshold (LNT) theory. The concept whereby the incidence of leukemia is proportional to the accumulated dose was adopted: ‘The most conservative approach would be to assume that there is no threshold and no recovery … and the incidence might be proportional to the accumulated dose’. It is noted that risk estimation based on the LNT theory or the linear–quadratic model emphasizes ‘how to protect or manage radiological risk’ rather than ‘how to predict the actual probability of cancer incidence or mortality’. For example, this theory was then used to determine the dose limit for occupational and public exposure, estimated from an unacceptable risk, which in turn was estimated from occupational accident risks and other risks [47-50]. In other words, radiation risk assessment had a regulatory science role in societal decision-making [30]. Risk assessment under these assumptions has two advantages: (i) it provides information for making risk management decisions, and (ii) it enables us to compare risks among various factors or to evaluate risk trade-offs.

VALUE OF RISK COMPARISON IN SOCIETAL DECISION-MAKING

After the Fukushima disaster, a large increase in mortality was observed for nursing home residents who underwent evacuation, typically within 3 months [19, 31], whereas no significant increase was observed in nursing home residents who did not evacuate [51]. Similarly, indirect health effects were observed, especially for elderly women, within the first 3 months after the disaster [20]. It should be noted that these changes might be mitigated with a greater societal understanding and preparedness for such risks before and immediately after a disaster. However, in hindsight, evacuation itself led to these changes, triggering such risks.

To understand risk levels and demonstrate the importance of policy and measures, a risk comparison study is introduced below: radiation risks versus evacuation-related risks in nursing homes [24]. All the above risks relate to post-disaster increases (i.e. additional risks after the disaster). Use of the same scale indicator is required for risk comparison; therefore, LLE was used as the risk indicator. LLE has two advantages over other risk indicators such as DALYs: (i) LLE can be quantitatively estimated using measurable data, and (ii) differences in the timing of mortality between radiation exposure and other risks (e.g. evacuation-related) can be expressed by the age reached at the time of death. LLE is based on the concept that a length of time is of equal worth to everybody. LLE can be ethically supported by health-maximization ageism related to efficiency [52] and fair-innings ageism related to equity [53]. LLE has been used in various fields, including radiological protection [50].

One main objective of risk comparison in societal decision-making is prioritizing countermeasures, which includes minimization of health risks based on ordering or comparisons. Decision depends on the civil authority, targeted population, and scope of the countermeasures, and therefore the target population involved is an important factor for assessment. In this comparison, nursing home staff and residents were targeted as the subjects of decision/management in nursing home facilities. Four scenarios were considered to discuss risk-trade off: ‘rapid evacuation’ (in conjunction with the actual situation, i.e. evacuation on 22 March 2011); ‘deliberate evacuation’ (instead of immediate evacuation, slow evacuation after completion of preparations for hospitalization, i.e. evacuation on 20 June 2011); ‘20-mSv exposure’; and ‘100-mSv exposure’. The ‘rapid evacuation’ scenario consisted of evacuation-related risk, whereas other scenarios included radiation risk. Radiation exposures of 20 mSv or 100 mSv in the first year were used as benchmarks because they are the lowest and highest reference levels, respectively, in the ICRP effective dose bands for emergency exposure situations [45].

For the comparison [24], to assess the radiation risk at low doses, cancer mortalities were estimated from linear–quadratic dose–response models for all solid cancers and leukemia. LLE was then calculated using the Japanese survival probability and excess relative risk models developed in the Life Span Study for atomic bomb survivors [54-56]. Moreover, LLE for the evacuation-related risk in three nursing homes was estimated using the pre- and post-disaster survival probability for nursing home residents in Minamisoma City. Although the areas were outside the mandatory evacuation order areas, all nursing home residents and staff chose to be evacuated voluntarily by 22 March 2011. LLEs in two nursing homes in Soma City, where there was no evacuation, were also estimated as a control. Mortality increases only in the 90 days after the evacuation were considered. As described in the section ‘Low-dose radiation risk assessment’, this approach was not used for accurate prediction but was used for determining the implications for risk management. With this objective, the radiation risk was conservatively estimated (i.e. overestimated), and another risk (i.e. evacuation-related risk in nursing homes) was underestimated. Detailed description and justification of methods, including estimation of mortality timing, can be found in Murakami et al. [24].

The risk comparison is illustrated in Fig. 1. The total LLE due to evacuation-related risk for residents was 11 000 person-days. This evacuation-related risk may have included non-evacuation related effects (e.g. disaster shock); in this case the LLE was calculated to be 880 person-days by using the values for two nursing homes where there was no evacuation. The total LLE due to evacuation risk for residents was much higher than the total LLE for residents and staff due to radiation in the other scenarios (27, 1100 and 5800 person-days for avoidable radiation doses over 90 days, at 20 mSv exposure, and at 100 mSv exposure, respectively). The high level of evacuation-related risk could be attributed to the combined effects of physical stress due to movement, limitations on medical resources, changes in medical staff, and poor nutritional conditions for elderly people [19, 31, 57]. Since the residents were evacuated in late winter/early spring, the evacuation-related risk might also be affected by the weather conditions.

Fig. 1.

Risk comparisons after the Fukushima disaster using the same indicator, LLE. Evacuation-related risk in nursing homes versus cancer risk due to radiation [24]. The avoidable radiation risk shows a difference in the LLE of the cancer risk due to radiation between actual evacuation and deliberate evacuation (instead of immediate evacuation, slow evacuation after completion of preparations for hospitalization, i.e. evacuation after 90 days post disaster).

From the result, the intention is not to discuss whether the decision regarding evacuation after the disaster was inappropriate. Rather, this result suggests that the risk comparison highlights the importance of countermeasures related to evacuation. It should be noted that assessments of evacuation-related risk from nursing homes have uncertainties due to limitations in the survey (e.g. number of participants, observation periods, emergency conditions, etc.), in contrast to long-term radiation risk studies consisting of a large number of participants, with measurable dose information, and control data. However, a recent systematic review of studies demonstrated that evacuation-related risk was consistently found for people in nursing homes in various man-made or natural disasters [58]. Evacuation-related risk from nursing homes likely depends on preparedness status as well as natural conditions (e.g. weather). The results suggest that evacuation regulation and planning are important for reducing overall risk as emergency preparedness. Cost effectiveness for evacuation regulation and planning is not clear at this stage. Cost effective analysis as well as risk comparison would provide useful information for improved societal decision-making.

VALUE OF RISK COMPARISON IN INDIVIDUAL DECISION-MAKING

After the Fukushima disaster, several risk comparison approaches were used to communicate the radiation risk [26-28]. Following their application, the value of risk comparison information was assessed through questionnaires to residents in the Fukushima, Tokyo and Osaka prefectures [32]. Residents received 1 of 15 risk comparison information packs (comprising combinations of 3 items of numerical risk data and 10 items of risk comparison information).

Numerical risk data were as follows: (A) current dietary radiocesium doses (mSv/year); (B) current dietary radiocesium doses (mSv/year) and corresponding lifetime cancer mortality rates based on LNT theory (%); and (C) current dietary radiocesium doses (mSv/year) and corresponding LLE (s). Current dietary radiocesium doses used in the study were as follows: 0.0016 mSv/year for Fukushima; 0.0010 mSv/year for Tokyo; and 0.0007 mSv/year for Osaka. The 10 items of risk-comparison information were as follows: (i) the radiation dose only (no comparison information); (ii) the food standard dose: ‘Current standards for restrictions on the distribution of foods have been established from 1 mSv/year’; (iii) the results for 100 mSv: ‘Clear health effects below 100 mSv have not been observed through epidemiology so far’; (iv) the 1960s dose: ‘The average dose of dietary radiocesium in 1964 in Japan derived from nuclear bomb tests was 0.019 mSv/year’; (v) the doses in other prefectures (current doses in two other prefectures were provided); (vi) the natural radiation dose: ‘The natural radiation dose in Japan, excluding radiation from the 2011 accident, is 2.1 mSv/year (1 mSv/year from the diet; 1.1 mSv/year from other sources)’; (vii) the total cancer mortality rate: ‘Approximately 20% of Japanese die from cancer’; (viii) the airplane dose: ‘The dose from a round-trip between Tokyo and New York by airplane is ~0.2 mSv’; (ix) the arsenic risk: ‘The cancer risk from inorganic arsenic in rice and hijiki seaweed corresponds to ~0.2 mSv/year, if converted to radiation dose units’; (x) the smoking risk: ‘The cancer risk from smoking corresponds to approximately 1000 to 2000 mSv, if converted to radiation dose units.’ Risk comparison information points (ii)–(iv) correspond to Rank 1 (most acceptable as risk-information sources) in Covello’s guidelines [29], point (v) is Rank 2, points (vi) and (vii) are Rank 3, points (viii) and (ix) are Rank 4, and point (x) is Rank 5 (rarely acceptable). Information regarding the risk of cancer from smoking came from the National Cancer Center Japan [28]; here, the hazard ratio of all cancer incidences was used to demonstrate the comparison [59]. Subjective understanding was assessed using a 5-point Likert scale with the question: ‘From the above information, do you intuitively understand the level of risk currently posed by dietary radiocesium in your prefecture?’ Answers of ‘comprehensible’ or ‘highly comprehensible’ are regarded as a presence of ‘subjective understanding’. Similarly, backlash against information was assessed using a 5-point Likert scale with the question: ‘Do you think that the information provided on the level of risk currently posed by dietary radiocesium in your prefecture is accurate?’ An answer of ‘highly inaccurate’ was regarded as a presence of ‘backlash against information’. Objective understanding was assessed through the following question: ‘Compare the level of risk currently posed by dietary radiocesium in your prefecture with the risk of dying in a traffic accident. How do you think the possibility of dying from cancer as a result of 1 year’s current intake of dietary radiocesium in your prefecture compares with the annual possibility of dying in a traffic accident?’ Five choices were provided: ‘higher than for a traffic accident,’ ‘comparable to that of a traffic accident,’ ‘about 1/10 of that of a traffic accident,’ ‘about 1/100 of that of a traffic accident,’ and ‘about 1/1000 of that of a traffic accident.’ According to statistics on annual traffic death rates, the mortality rate from one year of intake of dietary radiocesium, as calculated on the basis of LNT theory, was about 1/1000 of the annual traffic death rate. An answer of about 1/1000 of that of a traffic accident’ was regarded as a presence of ‘objective understanding’. Because LNT theory is used as a radiological protection concept and the estimated risk was conservative (i.e. overestimated) as mentioned in ‘Low-dose radiation risk assessment’, answers other than ‘about 1/1000 of that of a traffic accident’ represent overestimation of the risk compared with the value calculated from LNT theory.

Figure 2 shows the effects of the risk comparison information. All the risk comparison information significantly enhanced the subjective (intuitive) understanding of risk to residents in comparison with dose information alone. On the other hand, residents’ objective understanding of risk (i.e. understanding of radiation risk in comparison with traffic accident risk) was significantly increased only by a combination of information related to the cancer risk from radiation and the smoking risk (B10). It should be noted that a low odds ratio for objective understanding indicates that residents overestimated the risk compared with the value calculated from LNT theory. In other words, providing dose information only caused residents to overestimate the radiation risk in comparison with the cancer risk from radiation and the smoking risk. There were no significant differences among risk comparison information regarding the backlash against information; a proxy for the distrust of risk information. These results suggest that providing a combination of the cancer risk from radiation and the smoking risk as risk comparison information is the most effective for enhancing residents’ subjective and objective understanding. Contrary to Covello’s guidelines [29], the use of this combination did not cause respondents to distrust the risk information. The reason for the advantages of this information is not clear; however, it may be attributed to the fact that this risk comparison information provides both an actual risk (i.e. cancer risk from radiation) and a relative comparison indicator known in daily life (i.e. smoking).

Fig. 2.

Adjusted odds ratios for risk-comparison information provided for respondents’ attitudes to risk [32]. (a) Subjective understanding, (b) objective understanding, (c) backlash against information. A1 (radiation dose only) was used as a reference. Other risk comparison information was as follows: A2 (food standard dose); A3 (results for 100 mSv); A4 (1960s dose); A5 (doses in other prefectures); A6 (natural radiation dose); A8 (airplane dose); A9 (arsenic risk); A10 (smoking risk); B1 (cancer risk from radiation); B7 (cancer risk from radiation and total cancer mortality rate); B9 (cancer risk from radiation and arsenic); B10 (cancer risk from radiation and smoking risk); C1 (LLE from radiation); C10 [LLE from radiation and smoking risk (see text for full details)]. Odds ratios were adjusted by location (including evacuation experiences), gender, age, employment status, absence/presence of spouse, children, and grandchildren, educational background, completion of a humanities or science course, smoking habits, and perception of trustworthy information sources. Error bars represent 95% confidence intervals.

This risk comparison approach comprised the initial stage of group-based risk communication. However, after the Fukushima disaster, risk communication then shifted to small focus groups or face-to-face dialogues [60]. Regarding risk communication, authorities should carefully consider the objectives: namely, whether dialogues are implemented for risk acceptance or risk trade-off problems. The science of risk perception and risk communication historically began with the question ‘How safe is safe enough?’ [61]. Since then, studies on risk perception and even risk communication have often cited risk acceptance as the main outcome [12, 62]. In the case of the Fukushima disaster, controversial ‘expert’ opinions on radiation risks were provided to the residents [63]. In response to this, some residents think ‘we should consider radiation dangerous (overestimate the risk) to protect ourselves, if there is a variety of ‘expert’ opinions on the safety or danger of radiation’ [33]. Experts or authorities should consider that risk communication targeting the acceptance of only one specific risk (e.g. radiation risk alone) can blind residents to risk trade-off problems. Rather, providing risk comparison information under the framework of risk trade-off (e.g. a lack of outside exercise that may elevate the lifestyle-related disease risk versus avoidance of radiation) can support resident decisions concerning daily activities. The purpose of this paper is not to insist that risk comparison or risk trade-off information is useful for resident risk acceptance, but to highlight that authorities or experts should communicate risk trade-off problems to residents and contribute both to promoting overall resident health and supporting resident decision-making.

PERSPECTIVES

In this review, comparisons of risks after the Fukushima disaster are presented and their usefulness for societal and individual decision-making are discussed, while clarifying the concept of ‘risk’ and the objective of a ‘risk comparison’. Future study is required from the perspective of societal and individual decision-making.

First, this should include other post-Fukushima risks than those shown in this review. During the chronic phase, an increase in lifestyle-related diseases such as diabetes and hyperlipidemia was observed, especially in evacuees [21, 64, 65]. Table 1 shows the odds ratio, relative risk, or hazard ratio of the prevalence of three lifestyle-related diseases (i.e. diabetes, hyperlipidemia/dyslipidemia, and hypertension) for evacuees after the Fukushima disaster. Comparisons in pre- and post-disaster prevalence showed a significant increase of diabetes and hyperlipidemia, while no significant increase was generally observed for hypertension. Comparisons between evacuees and non-evacuees also highlighted the significant impacts of evacuation on diabetes and hyperlipidemia, although it is still not well understood why evacuation caused lifestyle-related diseases for evacuees during the chronic phase. Specifically, lifestyle diseases or psychological distress should be discussed as part of a risk comparison with radiation risks using the same scale indicator. For that, a risk indicator that considers the quality of life may be required. In these assessments, it should be noted what the objectives of risk comparisons are. In particular, as part of effective prioritizing of countermeasures, the targeted population involved should be carefully considered in order to rank various risks. Second, the effects of communicating risk comparison or risk trade-off information in a small focus group should be investigated to enable effective risk communication. Third, from a preparedness perspective, the pre-assessment of multiple risks from a hypothetical nuclear accident should be discussed, and not only risk, but also cost effectiveness should be demonstrated to suggest effective countermeasures. Overall, a comprehensive understanding of the multiple pre- and post-disaster risk assessment is still lacking, and improved comparisons of radiation and other disaster-related risks is required for effective risk management and communication.

Table 1.

Odds ratio, relative risk, and hazard ratio of lifestyle-related disease prevalence for evacuees after the Fukushima disaster [21, 64, 65]

Location	Data source	Subjects	Indicator	Year	Diabetes	Hyperlipidemia/dyslipidemia	Hypertension	Ref
Soma City and Minamisoma City	Public health check-ups administrated city offices	Evacuees	Relative risk (vs 2008–2010)	2011	1.12 (0.70–1.79)	1.10 (0.94–1.27)	1.05 (0.91–1.21)	[64]
				2012	1.21 (0.88–1.67)	1.16 (1.05–1.29)	1.04 (0.94–1.14)
				2013	1.55 (1.15–2.09)	1.30 (1.18–1.43)	1.10 (1.00–1.21)
				2014	1.60 (1.18–2.16)	1.20 (1.08–1.32)	0.94 (0.85–1.05)
		Non-evacuees/temporary-evacuees	Relative risk (vs 2008–2010)	2011	0.94 (0.81–1.10)	1.00 (0.95–1.05)	1.05 (1.01–1.10)
				2012	1.11 (0.97–1.27)	1.03 (0.98–1.08)	1.03 (0.99–1.07)
				2013	1.33 (1.17–1.52)	1.12 (1.07–1.17)	1.01 (0.97–1.05)
				2014	1.27 (1.11–1.45)	1.14 (1.09–1.20)	0.95 (0.91–0.99)
		Evacuees	Odds ratio (vs non-evacuees/temporary-evacuees)	2012–2014	1.14 (0.96–1.35)	1.18 (1.06–1.32)	0.97 (0.86–1.09)
Kawauchi Village	Japanese National Healthcare System medical examination data	Evacuees (including returners)	Odds ratio (vs 2008–2010)	2012	1.35 (1.13–1.61)	1.54 (1.33–1.78)	0.85 (0.75–0.96)	[65]
				2013	1.60 (1.32–1.95)	1.72 (1.47–2.02)	0.92 (0.80–1.06)
Evacuation order areas	Public health checkups by National Health Insurance and Fukushima Health Management Survey	Evacuees	Hazard ratio [odds ratio for hyperlipidemia] (vs non-evacuees)	2011–2012	1.40 (1.20–1.63)	Men: 1.41 (1.20–1.67) Women: 1.35 (1.08–1.69)	Men: 1.24 (1.10–1.39) Women: 1.06 (0.94–1.18)	[21]

Values in parentheses represent 95% confidence intervals. Ages were adjusted in all studies (other covariates depend on the studies).