The stress rating was associated not only with disruption but also with having to leave home, worry about future flooding and with health effects. This rating therefore appeared to capture many of the most severe impacts of flooding on the lives of households. Having to leave home was most strongly linked to the rating of getting the home back to normal and to damage to the property structure, with the stress of the flood event itself also a strongly associated effect.
The overall rating of the seriousness of the effects of flooding was highly correlated with the stress rating. Other effects that were closely associated with the overall rating were disruption, having to leave home, health effects and damage to the house. Two of the subjective severity measures, the stress of the flood event rating and the overall rating of the effects have been selected for further examination in the following section because they appear to be associated with a number of the most serious specific effects of flooding.
Table 6.2: Correlations between the subjective severity scores: Intangibles Survey
6.2 Resilience and vulnerability This section of the report sets out to examine in detail the concepts of vulnerability and resilience through further detailed analysis of the Intangibles data for the flooded population. We currently know least about the social aspects of vulnerability, partly due to the fact that it is not always easy to quantify socially created vulnerabilities (see Task 11 Social Indicator Set report, Tapsell et al., 2005). There has also been much discussion in recent years on whether certain individuals or groups within communities such as the elderly, the very young, women, the disabled, ethnic minorities etc. are likely to be more vulnerable or resilient to the effects of hazards and disasters than the population in general (Morrow, 1999; Fordham, 1998; Buckle et al., 2000). Moreover, the vulnerability of human beings in the community has emerged as the least known element in the disaster literature as hazard-proof building structures and prediction of hazard impact and warning systems have been improved (King and MacGregor, 2000).
There is still a limited understanding of what the terms vulnerability and resilience include (Buckle et al., 2000, and see Section 1.2.1). We hypothesise that resilience will be a function of vulnerability plus other factors, which include:
material resources: e.g. insurance, income, car ownership;
personal/household resources: e.g. skills, knowledge, experience, time;
community resources: e.g. help or support; and
evacuation, disruption, problems with builders or insurers.
These (other) factors affect: a) the ‘coping capacity’ or the ‘means by which people or organisations use available resources and abilities to face adverse consequences that could lead to a disaster’’ (FLOODsite, 2005), and b) the ‘available resources’. However, not all the above information can be obtained from the existing FHRC datasets as these data were collected for a different purpose.
According to their vulnerability, people or households will experience different degrees of tangible and intangible damages due to flooding. Moreover, depending on their resilience, some people or households will recover better than others. The question is, can a very vulnerable household, person or community also be very resilient?
There is some overlap in the definitions. Characteristics that determine vulnerability can also affect resilience or the capacity to recover. This is one of the reasons that make separating the two concepts difficult. Resilience is affected by vulnerability, but can increase or decrease independently. For instance, having insurance against flooding will increase the resilience of a household; it will help them recover better from (at least the financial) effects of the flood. Insurance will not reduce the damages to the house or its contents but will help the recovery, so it does not affect the vulnerability. Flood proofing measures or flood warnings, on the other hand, reduce vulnerability by potentially reducing damages to the household.
Some people are affected more severely by a flood than others. In order to see what factors affect the vulnerability and resilience three variables have been chosen as dependent variables and are outlined below.
The GHQ (General Health Questionnaire). This is a self-completion questionnaire that was designed as a screening test for detecting psychiatric disorders (Goldberg and Williams, 1988). The GHQ12 (12 questions) is regularly used in annual health surveys in England8. However, the GHQ only takes into account symptoms experienced in the past few weeks and does not focus on the health impacts from a specific event. To overcome this, the questionnaires were administered twice: once to measure ‘current’ health and secondly to focus on the ‘worst period’ after the flood event (RPA/FHRC, 2004). Current GHQ12 scores can be used as a measure of resilience but also can be studied as a function of the GHQ12 in the worst period and why some people recover better than others.
Overall subjective severity. Respondents were asked to rate the overall effects of the flood on the household using a 1 (no effects) to 10 (extremely serious) scale. Unlike the GHQ12, this is not an ‘individual’ measure as respondents were asked to consider the effects on the household.
Subjective stress on the household. Respondents were asked to rate the effects of the stress of the event on the household using the same 1 to 10 scale. This is another household measure.
The dependent variables (overall severity, subjective stress and GHQ12 current and worst) will be explained by several factors. The variables that are most likely to account for the vulnerability of the respondents are:
People/ household characteristics: e.g. age, gender, prior health, social class, household composition (e.g. presence of children and people over 75 in the household).
The flood event characteristics: i.e. depth, number of rooms affected, duration, speed of onset, contamination, damages, warnings.
Type of dwelling: people living in ‘vulnerable housing’ (e.g. single storey dwellings or basement flats) will be more vulnerable to the consequences of the flood event, that is, there are no upper floors to store furniture or to seek refuge.
Thus the reanalysis is based on a model that considers that vulnerability and resilience to flooding depend on a series of factors: flood event characteristics, social characteristics including prior health, dwelling characteristics and post-flood factors or intervening factors, see Figure 6.1.
Figure 6.1: Model of factors affecting vulnerability and resilience to flooding
The following variables will be investigated in order to explain resilience.
Money, income and savings used to repair the effects of the flood event. Having savings can help reduce the effects of the flood event and increases resilience. The amount spent by respondents (uninsured losses) can be normalised by level of income and/or social class. Having insurance is another key variable that will increase people’s resilience. The type of contents insurance (e.g. new for old) can be another key variable.
Demographics such as age, gender, social grade, number/presence of children, and single parents can affect resilience. Moreover, older people living alone may be less resilient, whilst people with higher education, as reflected in social grade, may be more articulate and more able to get help, and therefore more resilient.
There are no specific questions regarding community cohesion or social capital in the data sets. Length of residence and help received may give some indication of levels of support received from within the community. For instance, people that have been living at the same address for a number of years will be, in theory, more integrated in the community. The sources of help, such as neighbours or friends, can also indicate level of support within the community. People that have been living at the same address for a long time will be expected to have received more help from their neighbours and perhaps be more active in flooding groups. People with more social capital may rate the overall severity, health effects or stress as lower than people who do not receive support from their neighbours or have been living in the area for a shorter period. Another aspect that can be investigated is whether people that have been more active and participated in flood related groups/events/letters report better health.
General disruption to the life of the household, having to take days off work, having to leave the home and the length of evacuation, are all variables that are likely to increase the stress and severity of the event. Time to get back to normal is another measure of disruption.
As well as the above, there are also other intervening factors that may occur after the flood event and that may also affect resilience:
awareness of the risk of flooding and preventative measures (including insurance).
Thus the main hypotheses are:
Resilience (as measured by the GHQ, subjective severity and subjective stress) is a function of vulnerability (determined by flood event characteristics, dwelling characteristics and people/household characteristics) and the coping capacity (resources and how they are utilised).
Within vulnerable groups, if some people are shown to recover better from flooding than others, resilience can then be increased on its own.
Mediating factors such as problems with insurers and builders, evacuation, worry, awareness and preventative measures affect the coping capacity and thus resilience.
6.2.1 Measures of vulnerability and resilience
The four variables used in the analysis as measures of vulnerability and resilience are:
The General Health Questionnaire 12 (GHQ12) current score.
The General Health Questionnaire 12 (GHQ12) worst time score.
Overall severity of the effects of the flood on the household.
The effect of the stress of the flood event itself on the life of your household.
For conciseness, on occasions these variables are referred to jointly as ‘vulnerability variables’:
Unlike the GHQ12 questionnaire, which is a measure of the individual’s health, the subjective measures refer to the effects of the flood on the household. Therefore they are not measuring the same thing. While the GHQ12 scores and the factors that influence them have been examined in detail (Tunstall et al., 2006), no detailed analysis has previously been undertaken for the subjective severities. It is useful to present a comparative analysis of the two variables here.
As well as ‘objective’ measures of the impacts of flooding such as depth, number of rooms flooded, days spent away from the home, insured and uninsured costs, visits to doctors, etc. respondents were asked to rate the effects of several of those effects on the household. These ‘subjective’ variables were scored on a scale of 1 (indicates "no effect") to 10 (indicates "extremely serious effect").
Background to the General Health Questionnaire
As mentioned above, the General Health Questionnaire (GHQ) is a self-administered screening test aimed at detecting psychiatric disorders in community and non-psychiatric clinical settings. The GHQ was designed to be easy to administer, acceptable to respondents, fairly short and objective. It focuses on the psychological components of ill-health (Goldberg and Williams, 1988). The GHQ12 is a shorter version of the original GHQ60 questionnaire, and has been widely used with disaster victims (e.g. Reacher et al., 2004), and is regularly used in annual health surveys in England. The GHQ12 consists of twelve questions concerning general level of happiness, depression, anxiety and sleep disturbance over the past few weeks (Sproston and Primatesta, 2004).
The GHQ was designed for London and was intended to be culture-specific. However, the test has been translated into 38 languages and seems to work just as well in India, China or south London and also in very different settings: from rural communities to university students and general practice clinics. This seems to indicate that psychological distress has certain common features in widely different settings. Symptoms such as not being able to sleep due to worry or the inability to face up to one’s problems appear to be common to the human condition rather than being country-specific (Goldberg and Williams, 1988).
The GHQ focuses on changes in normal functions rather than upon long term disorders. The questionnaire focuses on two main classes of phenomena:
The inability to continue to carry out one’s normal ‘healthy functions’.
The appearance of new phenomena of a distressing nature (Goldberg and Williams
Scoring the GHQ12
Each item in the questionnaire consists of a question on whether the respondent has been experiencing a particular symptom on a 0-3 scale ranging from ‘less than usual’ to ‘much more than usual’. Each of the 12 questions thus has four possible responses. The GHQ12 is very simple to score and also has the advantage of eliminating any errors due to ‘end users’ or ‘middle users’ (Goldberg and Williams, 1988). Scoring can be by one of two methods: the GHQ method (score 0-12) and the Likert method (score 0-36).
To use the GHQ method, the first two response categories for each question are both given a zero score (no symptoms) and the third and fourth response categories are given a score of one (some symptoms). This method simply differentiates between those respondents within a sample who display symptoms of impaired mental health (cases) compared with those that do not (non cases). It does not take into account the degree of impaired health effects. The standard threshold for diagnosis of impaired mental health is a score of four or more out of the possible score of 12; this is referred to as a ‘high GHQ score’ (Sproston and Primatesta, 2004).
Using the second scoring method - the Likert scale - responses to questions are scored either 0, 1, 2 or 3, depending upon whether the respondent had experienced the symptom (e.g. 'Have you recently felt constantly under strain?') either not at all, no more than usual, rather more than usual or much more than usual. This system is preferred when the GHQ score is to be analysed as a continuous outcome. Total scores will be ranged between 0 and 36. Research suggests that 11-12 is the most effective threshold for identifying cases in Likert scored GHQ-12s. This means that respondents who score between 0 to 10 are not classified as cases, but those who score 11 and above are. However, thresholds may be varied to being higher or lower than 11/12, depending upon the particular population sample in question (Goldberg and Williams, 1988).
Current and Worst GHQ12
The GHQ12 was administered twice in the survey. First, respondents were told ‘We would like to know how your health has been in general over the past few weeks.’ They were then asked to complete the 12 item General Health Questionnaire. The respondents were not asked how long they had been experiencing the symptoms but to focus on how they had been feeling over the last few weeks, and this may thus result in the short and medium-term affects of an event not being captured. The GHQ is thus sensitive to very transient disorders which may remit without treatment (Goldberg and Williams, 1988). Thus these responses cover the health of respondents at around the time of the interview referred to as the ‘Current GHQ12’. Very few of those interviewed had been flooded within the last year; most (58%) had experienced flooding between two and two and a half years earlier including the autumn 2000 floods; a substantial minority (26%) had experienced flooding four or five years earlier including the Easter 1998 floods. Thus, for most respondents years had elapsed since the flooding giving their health time to recover or, as in a few cases, to deteriorate.
Second, after flooded respondents had completed the GHQ12 with reference to their current health, they were asked ‘to think back to how your health was when the health effects from the flooding were at their most severe’ i.e. the worst time and to complete the GHQ12 again with reference to that time (RPA, FHRC, 2004). Earlier in the interview after questions on health effects, respondents had been asked, ‘At what stage during or after the flooding were the health impacts most severe or worst for you personally? Please think about health in the broadest sense to include physical, mental and social well-being’. Respondents were asked to refer to this ‘worst time’ in completing the GHQ12 for a second time referred to as the ‘Worst GHQ12’. This required respondents to think back and recall how they felt in most cases several years earlier.
The ‘worst ever episode’ approach has been validated by Power (1988 and undated) using the longer GHQ28 questionnaire. It was not possible to validate the approach in the Intangibles study, but it was concluded that the GHQ12, if applied retrospectively to the ‘worst time’ following the flood, provided a reasonable measure of the short-term psychological effects (RPA, FHRC, 2004). A version of the GHQ12 questionnaires used in the Intangibles Survey is included in Appendix 2.