How to increase earthquake and home fire preparedness: the fix-it intervention

1.1 Effective intervention studies on earthquake and fire preparedness

Despite their scarcity, there are a number of examples of successful interventions and campaigns regarding community preparedness that provide a transparent methodology. Findings from these studies inform the present intervention. Overall, disaster preparedness interventions have proven successful in affecting adjustment measures when including hands-on training (Miller et al. 2014), drills (Simpson 2002), face-to-face interactions as opposed to sending a pamphlet or internet-based information, targeting social empowerment, a process of increasing personal, interpersonal or political power so that individuals can take action to improve their life situations (Paton et al. 2003) and community cohesion (Eisenman et al. 2009; Glik et al. 2014), defined as a “set of attitudes and norms that include trust, a sense of belonging, and the willingness to participate and help, as well as their behavioural manifestations” (Chan et al. 2006).

Based on additional evidence, other factors must also be built into interventions to make them effective. Proven predictors of preparedness are: perceived self-efficacy, that is, beliefs regarding personal capacity to act effectively, which influences the number and quality of plans, effort and perseverance in risk reduction behaviours (Paton et al. 2003); collective efficacy, defined as people’s shared beliefs in their collective power (Paton and Johnston 2015); perceived outcome expectancy, the perception of whether ones’ actions will reduce the problem or threat (Paton et al. 2003); behavioural intention, which refers to the motivation to act and precedes adjustment adoption (Paton et al. 2000, 2010); critical awareness, defined as the frequency of talking and thinking about a specific hazard (Paton and Johnston 2015; Paton et al. 2005); sense of community, that is, feelings of belonging and attachment to people and places (Bishop et al. 2000; Paton 2000); community participation, which refers to individuals who form groups to collaborate with management agencies (Karanci and Aksit 2000; McGee and Russell 2003; Paton and Johnston 2015; Paton et al. 2005; Tierney et al. 2001); and trust in authorities, defined as trust in various sources of information (Dillon and Phillips 2001; Paton and Johnston 2015). In addition to targeting these predictors in this intervention, sociocultural and emotive variables, such as anxiety and fatalism, were added as mediators of the relationship between awareness and actual preparedness behaviours, as well as that between awareness and personal responsibility (Joffe et al. 2013; Johnston et al. 2013; Paton 2008; Paton et al. 2010).

1.2 Theoretical models and preparedness intervention studies

In order to be effective, behaviour change interventions should be based on theoretical models that explain the behaviour change process (Michie et al. 2008). Various theoretical models have been used in natural hazard intervention studies, such as the person-relative-to-event model (Duval and Mulilis 1999), the precaution adoption process model (Weinstein and Sandman 1992) and the emotion-focused coping model (Lazarus and Folkman 1984). In the health field, the prediction of preparedness behaviours has been modelled according to the theory of planned behaviour and protection motivation theory (Abraham et al. 1998; Bennett and Murphy 1997). The inclusion of largely cognitive factors, as found in many of these models, is not sufficient for understanding the relationship between risk awareness and hazard preparedness. More holistic studies have attempted to develop models to predict the adoption of natural hazard preparedness with good success (Paton et al. 2003, 2005).

Thus, the study reported in this paper utilised a modified version of Paton et al.’s model of natural hazard risk reduction preparedness (Paton et al. 2003, 2005), which proposes that motivation, outcome expectancy and self-efficacy beliefs are the main predictors of preparedness. Paton et al. add a fourth variable to their model: intentions to prepare. In addition, drawing from Paton et al.’s model, empowerment and social cohesion are added to the intervention as potential mediators of preparedness. In addition to drawing on Paton et al.’s model of preparedness, this study is also rooted in and draws on previous work on social representations of earthquakes in Seattle and Izmir (Joffe et al. 2013). Like the Paton et al. model, this work showed self-efficacy to be important, but added anxiety, trust, sense of responsibility and sociocultural variables such as fatalism versus sense of control as potential determinants of preparedness. Thus, two key models are drawn on in shaping this intervention with the former giving more emphasis to cognitive and empowerment variables and the latter to emotive and sociocultural factors.

This paper is divided into two parts. The first provides a detailed description of the methodology and design of the intervention study so that it can serve as a basis for the development of future community intervention studies. The second part describes results of the intervention conducted in Seattle in September 2015.

2.1 Study design and assessment periods

Monetary incentives are given for participation in the study. In the study conducted in Seattle,¹ participants in the intervention group were paid $250 for completing baseline and post-intervention assessments, as well as a 6-h workshop divided over 2 days (3 h each day). Participants in the control group were given $100 for completing the baseline and post-intervention assessments. To ensure completion of the pretest and posttest assessments, participants in both groups were paid only after completion of the second assessment. Participants are paid $40 per completed follow-up assessment, at 3 and 12 months after the intervention. Study incentives for Izmir are equivalent to the monetary incentives in Seattle.

2.2 Sample

In order to define sample size, an a priori power analysis using G*Power statistical software, version 3.1 (Faul et al. 2009) was conducted. Equal-sized sample groups were planned, meaning that the allocation ratio of N1 to N2 is 1 (Bredenkamp 1969; Orme et al. 1990). To perform a two-group comparison of a primary outcome with a standard alpha of 0.05, 64 individuals per group, or a total of 128 individuals, provide 80 % power to detect a significant difference of 5 % between the intervention group and the control group in a two-tailed analysis. Consequently, we planned to recruit 100 participants per group to account for dropouts, in order to have a minimum of 64 participants per group left in the final longitudinal follow-up. The post-intervention sample in Seattle consisted of 157 people (85 intervention group and 72 control group), which decreased at 3 months to 145 (73 intervention group and 72 control group).²

The sample in Seattle consisted of adult residents (aged 18–80) of North Seattle, with a mean age of 50 (SD = 13). Sixty per cent of residents were female, 77 % identified as Caucasian, 47 % identified as Christian, while 36 % reported not having any religion; 59 % were married or living with their partner and 75 % were home owners. More than half of the sample reported being employed (52 %).

2.3 Area selection

Since sense of community and community participation have been found to be variables that affect preparedness and adoption of mitigation measures (Bishop et al. 2000; Paton 2000), intervening in an existing community seemed appropriate. Consequently, for the Seattle study, one area was selected for the intervention group and the intervention was conducted there. A geographically separate, equally large area served as the field for those in the control condition. In February 2015, the team of researchers travelled to Seattle to explore potential areas to conduct the study and ultimately selected two of them.³ Additional aims for the scoping trip included establishing contact with a recruitment agency and arranging meetings with Seattle emergency management experts, such as the Red Cross, the Seattle Fire Department and the Seattle Office for Emergency Management for information and collaboration in the intervention. This first trip to Seattle proved to be highly important in the design of the intervention and in making connections with experts and survivors of earthquakes, who agreed to serve as motivational speakers in the intervention. A similar trip was conducted for Izmir to explore and select the intervention areas as well as to establish contact with the respective emergency management authorities and trusted informants.

2.4 Participant recruitment and allocation

The sample constitutes adults (aged 18–80), one representative per household, who are residents of Seattle and Izmir. In each of these cities, the study areas constituted two geographically separated areas of equivalent demographic mix assigned to be the intervention and control groups, respectively. A recruitment agency from each city was hired to recruit participants. Area selection in the two cities was based on the same criteria.

In each of the two cities, the agency is instructed to recruit 100 participants from the intervention area and 100 from the control area. Specifically, they are asked, for both groups, to recruit evenly across each age group (18–35, 36–54, 55–80), 50 % male and 50 % female within each age group, two-thirds home owners and one-third renters and 50 % below $45,000 and 50 % above $45,000 annual household income (and the equivalent respective salary in Izmir). All recruiting in Seattle was completed over the phone by the recruitment agency.

One week before the intervention in Seattle, participants were called by the agency to arrange a specific date and time to visit their homes to complete the baseline assessment. After completion, the intervention group participants were called and reminded of the day, date, time and location for their workshops. One week after completion of the workshops, participants in both groups were called by the agency to schedule a specific date and time to complete the post-workshop assessments and to receive their respective incentives for participation.

2.5 Intervention design and delivery

The intervention is named fix-it, as it focuses on fixing and securing items in the home. Fix-it is a name that is easy to remember and has positive/neutral connotations, as opposed to threat and fear-inducing names with negative and catastrophic connotations. According to the literature on campaigns, in order to be effective in facilitating action, names and messages of preparedness campaigns should be positive in tone, promote a sense of empowerment, as well as being short, clear, consistent and motivating of action (Jones 2012; Kocher 1993). The intervention name received good feedback in the pilot study, with participants stating that the name was, “easy to remember”, “fine”, “concise and it implies action”.⁴ Fix-it was also accepted as a good name for the intervention in Izmir, as a similar Turkish term exists.

Fix-it includes the key ingredients of previous preparedness interventions that have proven successful, such as hands-on training, face-to-face interactions, empowerment and community cohesion. In addition, determinants of behaviour such as self-efficacy, outcome expectancy and motivation are targeted through different behavioural techniques (Michie et al. 2008). The intervention design was influenced by the work of Michie et al. (2011) concerning effective behaviour change techniques to challenge specific determinants of behaviour. To address motivation, the use of rewards, incentives (gift cards for group challenge winners, magnets and mugs with acronyms and the name of the intervention), social encouragement and support (engaging with the fix-it Facebook group page, taking pictures of their homes) were used, as was persuasive communication. Self-efficacy was addressed using self-monitoring (homework review), rehearsal (videogame play), coping skills learning, social encouragement and support and feedback (homework review); all behaviour change techniques have proven effective in changing self-efficacy (Michie et al. 2008). In order to affect outcome expectancy and increase motivation, persuasive communications were given by experts who were trusted sources, such as the American Red Cross, the Seattle Fire Department and the Seattle Office of Emergency Management (Michie et al. 2008). AFAD, the equivalent trusted source for emergency management in Turkey, and their experts were used as communicators in the Izmir intervention.

The intervention consisted of two 3-h interactive, face-to-face workshops for each participant in the active group and focused on securing items in the household. The workshops were led by two professionals, experts in emergency management training and psychologists.⁵ The workshops included hands-on training, as well as interactive tools, such as playing a video game, reviewing photographs of participants’ homes and engaging in a Facebook group page. With the purpose of intervening in an existing community, the intervention was delivered in a community centre (as in the case of Seattle) or space where communities gather to participate in activities. The intervention was carefully designed by the psychology researchers of the project who, in addition to reviewing the literature on natural disaster preparedness intervention studies and that on behaviour change, consulted three experts in the field of interventions, epidemiological studies and behaviour change techniques to review the intervention. This, together with the pilot study conducted in June, shaped the fix-it intervention.

2.6 Consent

Ethical permission was obtained to conduct the study (UCL Ethics Project ID Number: 1392/001). Informed written consent was obtained from each participant before completing the baseline survey. A member of the recruitment agency collected all signed consent forms from participants in both groups. Both participants and recruitment agency members were blind to group allocation.

2.7 Measures

A 25-min computer-assisted survey and a checklist comprised the assessment that participants in both groups completed at the following time points in their homes⁶: (1) at baseline, 1 week before the intervention, (2) 1 week after the intervention, (3) 3 months after and (4) 12 months after the intervention. At the first home visit, a member of the recruitment agency met with each participant and asked them to complete the baseline survey online. Once finished, the agency member proceeded with the second part of the assessment, which was to conduct a tour around the house and examine with a checklist whether the participant had or had not adopted each of the nine preparedness measures. This constituted the observational measure of the study. Home visits to observe and document participants’ preparedness behaviours, rather than sole reliance on self-reports, offer a novel approach to this field.

2.8 Interrater reliability

Interrater reliability on 5 % of the checklists in all countries, using two independent raters, produced a satisfactory result (kappa = 0.65).

2.9 Data analysis and management

The SPSS statistical software package, version 20 (Nie et al. 2011), was used for conducting the statistical analysis. Behaviour change was measured in both groups, and t tests were conducted to evaluate differences between means of behaviour change between groups. Logistic regression analysis of the outcome measure on the active group was conducted, with adjusted confidence intervals and p values. In addition, cross-tabulation and χ ² statistics were carried out to evaluate correlations between self-report and observational preparedness measures to determine the extent of agreement between them. Moderated mediation analysis among relevant variables was carried out to explore mediators of preparedness. Levels of anxiety in relation to preparedness behaviours were also examined. Although preparedness is measured as an overall score, including earthquake and fire measures, separate analyses for earthquake and home fire preparedness were conducted in order to explore differences in preparedness for the more every day risk (i.e. fires) versus the more long return period risk (i.e. earthquakes). It must be noted that it is the difference in scores between assessment times that is the key outcome: that is, the study measures behaviour change.⁷

3.1 Baseline analyses (Tables 3, 4, 5, 6)

Both the intervention and control groups reported barriers to earthquake and fire preparedness. The most common barrier reported was “other things/problems to think about”, followed by “the time required to do it” and “the skill or knowledge required”. The least reported barrier for preparing for earthquakes was “the financial cost” (Table 4).

Predictors of preparedness before the intervention are listed in Table 5. It is important to note that all results are given first for overall preparedness, which refers to an average of the five earthquake and the three fire preparedness items.⁸ Results show that empowerment (F(1,154) = 9.1; p = 0.003, R ² = 0.05), trust (F(1,154) = 4.0; p = 0.046, R ² = 0.02), corruption (F(1,155) = 6.2; p = 0.01, R ² = 0.03) and social cohesion (F(1,155) = 1.6; p = 0.04, R ² = 0.23) predicted overall preparedness in a univariate regression analysis at baseline. Predictors of earthquake preparedness, in particular, at baseline were empowerment (F(1,154) = 4.5; p = 0.034, R ² = 0.02), trust (F(1,154) = 3.9; p = 0.04, R ² = 0.02), corruption (F(1,155) = 5.6; p = 0.01, R ² = 0.03) and social cohesion (F(1,155) = 9.8; p = 0.00; R ² = 0.05). Empowerment was the only predictor for fire preparedness at baseline.

Correlations between self-reported preparedness in the survey and observational preparedness measures in the checklist are described in Table 6. All items correlated significantly with each other except “computer is secured”, “all exits are clear of obstruction” and “no objects are hanging above sofas/beds”.

3.2 Changes in preparedness levels after the intervention (Table 7; Figs. 1, 2, 3)

One week post-intervention, the intervention group had significantly improved its overall preparedness (t(72) = 3.8, p < 0.001), as well as its earthquake (t(72) = 2.4, p = 0.018) and fire preparedness (t(72) = 3.7, p < 0.001) compared to baseline. Those in the control group improved their overall preparedness compared to baseline, though not significantly. The control group’s levels of earthquake preparedness decreased slightly though not significantly from baseline, but they had significantly increased their fire preparedness (t(71) = 2.4, p = 0.019). Between-group differences of preparedness change means 1 week after the intervention were significant for overall preparedness change (p = 0.001) and for earthquake preparedness change (p = 0.019), with the intervention group showing significantly more change (Table 7).

Table 7

Changes in preparedness levels 1 week after the intervention and at 3-month follow-up

Post-interventiona	Intervention (n = 73)				Control (n = 72)				Between groups
	M	SD	t test	p-level	M	SD	t test	p-level	t test	p-level
Overall preparedness change	1.48	3.30	3.83	<0.001	0.18	2.90	0.52	0.60	2.51	0.013
Earthquake preparedness change	0.85	2.99	2.42	0.018	−0.22	2.42	−0.77	0.40	2.36	0.019
Fire preparedness change	0.62	1.44	3.65	<0.001	0.38	1.32	2.39	0.019	1.05	0.296
3-month follow-up
Overall preparedness change	0.93	2.28	3.31	0.001	0.91	3.44	2.18	0.032	0.27	0.978
Earthquake preparedness change	0.07	1.98	0.30	0.80	0.47	2.82	1.37	0.20	−0.94	0.348
Fire preparedness change	0.88	1.29	5.55	<0.001	0.38	1.65	1.90	0.061	1.94	0.054

^aOne week after the end of the intervention

Open image in new window

Open image in new window

Open image in new window

Three months after the intervention, both the intervention (t(66) = 3.3, p = 0.001) and control groups (t(67) = 2.2, p = 0.032) had significantly improved their overall preparedness levels compared to baseline. Although both groups had improved their earthquake preparedness 3 months after the intervention, the scores did not reach statistical significance. Fire preparedness improved for both groups at 3-month follow-up compared to baseline, but only the intervention group’s fire preparedness reached statistical significance (t(66) = 5.6, p < 0.001) (Table 7).

Multivariate analysis of variance was conducted to explore levels of preparedness change between groups across the two points in time, 1 week after the intervention and at the 3-month follow-up. Results show a significant interaction of time and group (p = 0.013), with the intervention group significantly decreasing their overall preparedness from time 1 (post-intervention) to time 2 (3-month follow-up) compared to the control group (Fig. 1). A similar pattern can be found regarding earthquake preparedness (p = 0.002) (Fig. 2). There was no significant interaction for group or time in fire preparedness between the two time points (Fig. 3).

3.3 Predictors of preparedness change (Table 8)

In terms of the specific predictors of earthquake and fire preparedness change at the 3-month follow-up, only anxiety (F(9,58) = 2.7; p = 0.043, R ² = 0.297) and trust (F(9,58) = 2.7; p = 0.009, R ² = 0.297) were significant predictors of earthquake preparedness, for the control group only (within a multivariate analysis of self-efficacy, outcome expectancy, fatalism, collective efficacy, anxiety, trust, corruption, empowerment and social cohesion). The only predictor for fire preparedness change at 2 months was self-efficacy (F(9,58) = 1.49; p = 0.045, R ² = 0.188), and this was only for the control group.