Measures of severity tend to be associated with a particular theoretical framework. For example, there are measures of severity developed in the context of the Health Belief Model (Maiman et al., 1977; Champion, 1984), the Extended Parallel Process Model (Witte et al., 1996), and Leventhal's Self-regulation Theory operationalized via the Illness Perception Questionnaire (Weinman et al., 1996; Moss-Morris et al., 2002) (see Appendix for severity measures). Although the latter are validated measures, questions remain about whether they are appropriately worded, suitably specific and fully assess the concept of severity.
Scope and the role of emotional arousal. There appears to be general consensus among health behaviour theories that the scope of perceived severity is extremely broad. For example, Janz and Becker (1984) state: "This dimension includes evaluations of both medical/ clinical consequences (e.g., death, disability, and pain) and possible social consequences (e.g., effects of the conditions on work, family life, and social relations)." Similarly, Weinman and colleagues describe the consequences component of illness perceptions as encompassing "physical, social and psychological functioning" (Weinman et al., 1996). Other researchers specify that intrapersonal threats should also be included, such as threat to self-esteem (Rogers, 1983). However, one key difference between theories is whether the concept of severity should include measures of emotional response, i.e., fear and worry.
Health Belief Model. The Health Belief Model appears to differ from other theoretical frameworks by including emotional arousal in its definition of severity. Rosenstock (1974) says that: "The degree of seriousness may be judged both by the degree of emotional arousal created by the thought of a disease as well as by the kinds of difficulties the individual believes a given health condition will create for him". Hence in the HBM, fear/worry forms part of perceived severity and consequently also forms part of the motivation to act. However, few modern researchers using the HBM include these emotional elements in their definitions or assessments of severity.
Other theories. In contrast to the HBM, other theories have treated anticipated or actual emotional reactions to disease as conceptually distinct from beliefs about disease severity. For example, the Parallel Response Model (Leventhal, 1970) separated emotional representations from cognitive ones. This division was considered necessary because emotions (physical symptoms/arousal in particular) did not appear to be associated with behaviour, whilst cognitive representations (e.g. beliefs about severity) did.
Consistent with the empirical findings, the revised Protection Motivation Theory (Rogers, 1983) also proposed that emotional response was not part of severity and that emotional arousal played an indirect role in protection motivation through its effect on cognitive representations (perceptions of severity and vulnerability). Although Witte also argues that fear does not play a direct role in danger-control responses, she does state that fear plays a direct role in fear-control responses.
Dimensionality. Because the potential consequences of a health threat or hazard can be so diverse, it is not surprising that severity appears to be a multidimensional concept (Milne et al., 2000). For example, Champion's measure of breast cancer severity, based on the HBM, contains 3 factors: physical symptoms of fear, long-term effects of breast cancer, and financial/career problems. In addition, Milne and colleagues point out that some researchers have focused on the dimension of fatality, whereas others have measured dimensions relating more to psychosocial severity such as the effect a disease would have on life goals (Milne et al., 2000). But they also highlight other potential dimensions of severity, such as whether the disease is likely to have a rapid or gradual onset and how visible the symptoms of the disease are likely to be. These aspects of severity have rarely been explored.
Specificity/ content validity. A number of severity measures are generic, assessing broad judgments about the seriousness of the disease. This means that the same questions have been used to assess the impact of different health threats. For example, although a variety of disease-specific questionnaires exist for assessing illness perceptions, the severity items are the same - the name of the particular disease is just inserted in the appropriate place (see Appendix). However individual researchers have sometimes chosen to supplement these with additional items. For example, following interviews with participants about perceptions of breast cancer, Anagnostopoulos and Spanea (2005) added severity items to the IPQ to assess social isolation, physical exhaustion, pain, financial consequences, and family conflicts (although not all of these made it into the final questionnaire).
The decision about whether to use general or specific items depends partly on the aims of the research. If the research aims to assess the impact of health messages that seek to alter particular beliefs about the consequences of a disease, then the measurement of disease-specific consequences may be advisable. Some guidelines for the development of items to assess severity are offered by Witte and are reproduced in the Appendix. Fishbein et al. (2001) also offer guidelines for assessing the perceived consequences associated with a given behaviour, and their suggestions could be applied to the perceived consequences associated with a disease.
The likelihood and valence of different outcomes. Fishbein et al. (2001) distinguish between the likelihood of anticipated outcomes occurring and the value attached to those outcomes (positive or negative), and they state that these constructs need to be assessed separately. This proposal raises two issues: whether each possible outcome (e.g. job loss, impact on social relationships) should be assessed for likelihood of occurrence as well as severity, and the need to pay close attention to valence.
Usually, disease severity items are summed and then combined with the likelihood of getting the disease. But predictive utility may be improved if each possible outcome were weighted according to the likelihood of its occurring. For example, job loss may be viewed as severe but unlikely to occur, whereas the impact on social relationships may be viewed as more likely to occur but less severe, and it may be wrong to assume that the former should contribute more to total perceived severity than the latter. However very little research has examined the performance of different measures of severity and such work would help answer this kind of question.
In terms of valence, while there might be universal agreement that certain outcomes tend to be viewed as negative (e.g. death), other consequences, such as the avoidance of pregnancy, will be valued differently by different people. Consequently, one of the problems with some of the existing measures of severity is that some items are ambiguous with respect to valence, such as: 'My illness strongly affects the way others see me' (from the revised Illness Perceptions Questionnaire (Moss-Morris et al., 2002).
Self-reference. Another issue that relates to question wording is the use of self-referencing. As Rosenstock (1974) states, the orientation adopted by the social psychologists involved in the development of the HBM held that "it is the world of the perceiver that determines what he will do". Some researchers have explicitly defined severity as concerned with how serious the outcome would be for the individual concerned (Brewer et al., 2007; Champion, 1984), and others have noted that perceptions of severity are likely to vary widely between individuals (Rosenstock 1974; Janz and Becker 1984). However, not all measures of severity are phrased accordingly. The absence of self-referencing is a major weakness in a number of severity measures. This situation contrasts somewhat with the measures used to assess perceived vulnerability, where the individual is usually asked to give an assessment of their own personal chances of experiencing a particular event, rather than the likelihood of the event happening in general (e.g. Brewer et al., 2004 but deviations from this practice still occur-see Brewer et al., 2007 for a discussion of measure quality in this area).
Conditioned perceived severity. As with perceived vulnerability, the question of obtaining a rating that takes into account behavioral plans deserves consideration. In theories such as the HBM and PMT, vulnerability and severity relate to a threat that would arise if there were no change in behaviour (e.g. no preventive action were taken). However, measures assessing severity rarely make this explicit. Perceptions of disease severity can depend on whether people intend, or already engage in, actions that are likely to reduce the severity of the target disease. For example, adhering to colorectal cancer screening can reduce disease severity because it can result in the detection of the cancer at an earlier stage when it has a much better prognosis. As a result, people who adhere to colorectal cancer screening, or intend to do so, may reasonably perceive colorectal cancer as being less severe than someone who does not intend to adhere to screening. Ideally, questions about the severity of a condition should therefore specify 'a behavioural context' (see Brewer et al., 2004), particularly if disease severity is expected to vary with the relevant behavior.