Roles Conceptualization, Formal analysis, Methodology, Project administration, Resources, Writing – original draft, Writing – review & editing Affiliation School of Rehabilitation Science, McMaster University, Hamilton, ON, Canada ⨯
Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Supervision, Writing – original draft, Writing – review & editing * E-mail: gnazari@uwo.ca Affiliations School of Physical Therapy, Health and Rehabilitation Science, Western University, London, ON, Canada, Collaborative Program in Musculoskeletal Health Research, Bone and Joint Institute, Western University, London, ON, Canada
Roles Conceptualization, Formal analysis, Funding acquisition, Methodology, Software, Supervision, Visualization, Writing – original draft, Writing – review & editing Affiliations School of Physical Therapy, Health and Rehabilitation Science, Western University, London, ON, Canada, Collaborative Program in Musculoskeletal Health Research, Bone and Joint Institute, Western University, London, ON, Canada, Roth McFarlane Hand and Upper Limb Centre, St. Joseph’s Hospital, London, ON, Canada ⨯
Roles Supervision, Writing – original draft, Writing – review & editing Affiliation Burlington Fire Department, Burlington, ON, Canada ⨯
Roles Supervision, Writing – original draft, Writing – review & editing Affiliation Burlington Fire Department, Burlington, ON, Canada ⨯
To assess the effectiveness of Home Fire Safety (HFS) interventions versus other interventions/no interventions/controls on HFS knowledge and behaviour at short-, intermediate- and long-term follow ups.
Systematic review and meta-analysis of randomized controlled trials.
MEDLINE, EMBASE and PubMed databases were searched from January 1998 to July 2018, and studies retrieved.
Toddlers, children (primary or secondary school), teenagers or adults.
HFS interventions compared to other interventions / no interventions / controls.
HFS knowledge and behaviour.
The limited evidence supports the use of HFS interventions to improve HFS knowledge and behaviour in children, families with children and adults.
Citation: Senthilkumaran M, Nazari G, MacDermid JC, Roche K, Sopko K (2019) Effectiveness of home fire safety interventions. A systematic review and meta-analysis. PLoS ONE 14(5): e0215724. https://doi.org/10.1371/journal.pone.0215724
Editor: Wisit Cheungpasitporn, University of Mississippi Medical Center, UNITED STATES
Received: January 22, 2019; Accepted: April 9, 2019; Published: May 20, 2019
Copyright: © 2019 Senthilkumaran et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the manuscript and its Supporting Information files.
Funding: The author(s) received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Every day, 7 people die from home fires in the United States (US) [1]. Residential fires remain a major public health burden [2–4]. Between 2011 and 2015, U.S. fire departments responded to an average of 358,500 home structure fires per year, which resulted in an average of 2,510 fatalities annually [1]. In 2016, the rate at which U.S. home structure fires were reported was 1.1 per thousand population [1]. In United Kingdom, it was estimated that at least 500 deaths and 15,000 injuries were due to residential fires in 1998 [5–6].
Fire prevention requires multiple strategies. One strategy is to identify and target risk factors. Several systematic reviews have identified factors associated with higher rates of fires high number of residents, male homeowner, children under age of 5 years, smoking, low-income, buildings in poor conditions, frailty/disability, young and old age tenants, as the distinguishing risk factors associated with such incidents [7–9].
Another approach to fire safety is early detection of fire initiation in the homes, to prevent progression. To date, two meta-analyses examined smoke alarm coverage interventions by comparing the intervention to no interventions or to usual care [10–11]. A later network meta-analysis evaluated the effectiveness of smoke alarm interventions, but instead, compared several types of interventions, and found that the most effective method was the most intensive (includes education, low cost equipment fitting and in-home safety inspection) [12]. Home Fire Safety (HFS) knowledge and behaviour outcomes were also examined in a 1999 review [13]. The results concluded that there is a need for program evaluation especially among school-based education programs.
While the review by Warda et al. (1999) provides valuable insights, it has important limitations. For example, the review is outdated, included no critical appraisal or meta-analyses. Furthermore, since 1999, numerous studies have emerged [14–27], which warrants the need for a systematic review and meta-analysis. Therefore, the aims of this review were:
We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and Cochrane collaboration guidelines [28–29] (S1 Checklist) PROSPERO registration number: CRD42018106866.
Studies were included in this systematic review if the below criteria were met [30–33]:
Reports, conference abstract and posters were excluded from this systematic review [30–33].
We conducted systematic electronic searches to identify relevant studies in MEDLINE, EMBASE and PubMed from January 1998 to July 2018. Several different combinations of keywords were used, such as: “home fire safety”, “home fire safety knowledge”, “home fire safety behavior”, “effectiveness of home”, fire safety, “residential fires”, “fire prevention programs”, “fire prevention programs adults”, “fire prevention programs children”, “fire prevention”. (S1 File). In addition, we carried out a manual search of the reference lists of the identified studies.
Two independent reviewers (MS and GN) performed the systematic electronic searches in each database. We then identified and removed the duplicate studies. In the next stage, we independently screened the titles and abstracts and retrieved in full text any article marked include or uncertain by either reviewer. Lastly, we carried out an independent full text review to assess final eligibility. In case of disagreement, a third reviewer, (JM), provided a consensus through discussion.
Two independent researchers (MS and GN) extracted the data from the eligible studies. In case of disagreement, a third reviewer (JM), provided a consensus through discussion. Data extraction included the author, year, study setting, study population, sample size, age, intervention/comparison groups, follow up periods and the primary and secondary outcomes. When insufficient data were presented, GN contacted the authors by email and requested further data.
Two independent review authors (JM and GN) assessed the RCTs and non-randomized studies for risk of bias. The risk of bias assessment in the included RCTs was performed using the Cochrane Risk of Bias tool [29]. The Cochrane Risk of Bias tool is based on 7 items, random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting and other bias [29]. We defined the other bias category as trials that did not include statements on sources of funding/potential sources of conflicts of interest. The adequacy of each of the seven risk of bias domains was rated as “low”, “unclear” or “high” risk according to criteria provided in the Cochrane Handbook for Systematic Reviews of Interventions [29].
We used the GRADE approach for systematic reviews, to determine the quality of evidence related to each outcome to summarize the extent of our confidence in the estimates of the effect [31–36]. The GRADE approach considers the risk of bias, publication bias, consistency of findings, precision, and the applicability of the overall body of literature to provide a rating of quality of evidence (high, moderate, low, or very low) per outcome [34–39].
To quantify and interpret our data, a Minimal Clinically Important Difference (MCID) of 0.5 standard deviation points for HFS knowledge and behaviour was used [40]. Timing of outcome assessment were categorised as short-term (3–4 months), intermediate-term (6 months) and long-term (12 months).
In the presence of heterogeneity, we planned to perform the following subgroup analyses (a priori): trials at low risk of bias (low risk of bias in allocation concealment and blinding of outcome assessor), type of HFS intervention used. An I 2 estimate of at least 50% and a statistically significant Chi 2 statistic (P = 0.10) were used to indicate evidence of a substantial problem with heterogeneity [41].
We performed 7 meta-analyses of studies comparing HFS interventions to other interventions/no interventions/controls, using the outcomes HFS knowledge or behaviour, at short-, intermediate- and long-term follow ups. We used the Review Manager 5.3 (RevMan 5.3) software to conduct our review and used the standardized mean difference (SMD) with a random-effects model to pool outcomes.
Initially, our search identified 510 publications. After removal of the duplicates, 455 articles remained and were screened using their title and abstract, leaving 44 studies for full text review. Of these, 10 studies were eligible (8 RCTs and 2 prospective cohort) [18–27]. The flow of studies through the selection process is presented in Fig 1.
PowerPoint slide larger image original image Fig 1. Selection of studies for inclusion in the systematic review.The 8 eligible RCTs were conducted between 2003–2017 and included 1962 participants. Study size ranged from 76 to 499 participants. Trials were conducted in Canada, USA and UK [18–20, 22–26]. Only two out of the eight trials were registered in a clinical trials register [20,22]. In addition, three trials did not include statements on sources of funding or potential sources of conflicts of interest [24–26]. A summary description of all the included RCTs is displayed in Table 1. The 2 eligible prospective cohort studies were conducted in 2003 and 2017, and included 1491 participants (study sizes were 671 and 820). Studies were conducted in UK and Australia. A summary description of all the included studies is displayed in Table 2.
PowerPoint slide larger image original image Table 1. Study characteristics of the included randomized controlled trials. PowerPoint slide larger image original image Table 2. Study characteristics of the included prospective cohort studies.The risk of bias assessment is presented in Fig 2. Performance bias (lack of or inadequate blinding of participants who could influence how interventions, including co-interventions are performed/administered) was rated at high risk in all the included trials (n = 8) [18–20, 22–26]. Detection bias [18, 22–25] (lack of or inadequate blinding of participants who could influence the measurement or interpretation of outcomes) and attrition bias [20, 22–24, 26] were rated at high risk in five trials. Selection bias [18–28, 23–25] and selective reporting bias [18–19, 23–26] (significant or imbalanced missing outcome data) were rated at high risk in six trials. Other biases (RCTs with no statements on sources of funding/conflicts of interest) were rated at high risk in three trials [24–26]. Overall, all eight included RCTs were rated at high risk of bias.
PowerPoint slide larger image original imageFig 2. Risk of bias summary: Review authors’ judgements about each risk of bias item for each included study.
The EP (Table 3) displays a detailed quality assessment and includes a judgment of each factor that determined the quality of evidence for each outcome. The SoF tables (Tables 4–6) include an assessment of the quality of evidence for each outcome.
PowerPoint slide larger image original image Table 3. GRADE evidence profile: Intervention vs no intervention/control. PowerPoint slide larger image original image Table 4. Summary of findings.Intervention vs No Intervention.
PowerPoint slide larger image original image Table 5. Summary of findings.Intervention vs Control.
PowerPoint slide larger image original image Table 6. Summary of findings.Computer-based vs Instructor-based.
Among the eligible RCTs, four recruited parents/caregivers of children [20–23,26], three included primary school children [18–19, 24], and one recruited adult participants [25]. Among the eligible prospective cohort studies, one included school children [21], and the one recruited adult participant [27].
Home fire safety knowledge was assessed in 7 RCTs and 2 prospective cohort studies [18–19, 22–26]. Home fire safety behaviour was examined in 6 RCTs [18–20, 22–23, 25]. The follow-up period ranged from immediate to 12 months post-intervention.
Two studies were pooled to examine the effects of interventions (Risk Watch and Great Escape) vs no interventions on home fire safety knowledge at short-term (up to 4 months) follow up. The pooled results, demonstrated significant difference between groups (very low quality, 2 RCTs, 535 participants, standardized mean difference (SMD) 0.38, 95% CI: 0.21 to 0.55, p < 0.001, Fig 3; Analysis 1.1.1). Heterogeneity was absent. Given that an MCID is approximately 0.5 SD, the pooled results were not clinically important. However, more data are required to make a definitive conclusion.
PowerPoint slide larger image original imageFig 3. Analysis 1.1.1 forest plot of comparison: Intervention vs no intervention, up to 4 months–primary school children, outcome: Home fire safety knowledge, 2 RCTs. Analysis 1.1.2 Forest plot of comparison: Intervention vs No Intervention, up to 4 months–Primary School Children, outcome: Home Fire Safety Behaviour, 2 RCTs.
Higher values indicate better/improved outcome.
Two studies were pooled to assess the effects of interventions (Risk Watch and Great Escape) vs no interventions on home fire safety behaviour at short-term (up to 4 months) follow up. The pooled results, displayed no significant difference between groups (very low quality, 2 RCTs, 609 participants, SMD 0.34, 95% CI: -0.21 to 0.89, p = 0.23, Fig 3; Analysis 1.1.2). Heterogeneity was high, and we were not powered to conduct sub-group analysis. Given the MCID of 0.5 SD, the pooled results were not clinically important. However, more data are required to make a definitive conclusion.
Fig 4. Analysis 4.1.1 forest plot of comparison: Intervention vs no intervention, immediate–primary school children, outcome: Home fire safety knowledge, 1 study.
Analysis 4.1.2 Forest plot of comparison: Intervention vs No Intervention, 3 months–Primary School Children, outcome: Home Fire Safety Knowledge, 1 study. Higher values indicate better/improved outcome.
One study assessed the effects of home fire safety intervention vs control (minimal intervention) on home fire safety knowledge at short-term (2 months) follow up. The results, displayed significant difference between groups (very low quality, 1 RCT, 96 participants, SMD 0.66, 95% CI: 0.25 to 1.07, p = 0.002, Fig 5; Analysis 2.1.1). Given that an MCID is approximately 0.5 SD, the results were clinically important. However, more data are required to make a definitive conclusion.
PowerPoint slide larger image original imageFig 5. Analysis 2.1.1 forest plot of comparison: Intervention vs control, 2 months–families with children, outcome: Home fire safety knowledge, 1 rct.
Analysis 2.1.2 Forest plot of comparison: Intervention vs Control, 6 months–Families with Children, outcome: Home Fire Safety Behaviour, 1 RCT. Analysis 2.1.3 Forest plot of comparison: Intervention vs Control, 12 months–Families with Children, outcome: Home Fire Safety Behaviour, 1 RCT. Higher values indicate better/improved outcome.
One study examined the effects of home fire safety intervention vs control (minimal intervention) on home fire safety behaviour at intermediate-term (6 months) follow up. The results demonstrated significant difference between groups (very low quality, 1 RCT, 277 participants, SMD 0.35, 95% CI: 0.09 to 0.60, p = 0.007, Fig 5; Analysis 2.1.2). We found similar results at long-term (12 months) follow up, (very low quality, 1 RCTs, 277 participants, SMD 0.36, 95% CI: 0.11 to 0.61, p = 0.005, Fig 5; Analysis 2.1.3). Given the MCID of 0.5 SD, the results were not clinically important. However, more data are required to make a definitive conclusion.
One study assessed the effects of different modes of home fire safety interventions (computer-based vs instructor-led) on home fire safety knowledge immediately post-intervention. The results, displayed no significant difference between groups (very low quality, 1 RCT, 68 participants, SMD -0.02, 95% CI: -0.50 to 0.45, p = 0.92, Fig 6; Analysis 3.1.1). Given that an MCID is approximately 0.5 SD, the results were not clinically important. However, more data are required to make a definitive conclusion.
PowerPoint slide larger image original imageFig 6. Analysis 3.1.1 Forest plot of comparison: Computer-based vs instructor-led, immediate–adults, outcome: Home fire safety knowledge, 1 RCT.
Analysis 3.1.2 Forest plot of comparison: Computer-based vs Instructor-led, Immediate–Adults, outcome: Home Fire Safety Behaviour, 1 RCT. Higher values indicate better/improved outcome.
One study assessed the effects of different modes of home fire safety interventions (computer-based vs instructor-led) on home fire safety behaviour immediately post-intervention. The results, displayed no significant difference between groups (very low quality, 1 RCT, 68 participants, SMD 0.06, 95% CI: -0.41 to 0.54, p = 0.79, Fig 6; Analysis 3.1.2). Given that an MCID is approximately 0.5 SD, the results were not clinically important. However, more data are required to make a definitive conclusion.
This review identified and synthesized the most rigorously designed intervention studies, finding that there is a small number of studies examining diverse HFS interventions on knowledge and behaviour. In fire prevention research a major challenge is how researchers can ascertain whether a fire was prevented. Hence, they rely on test of knowledge of fire prevention strategies. The limitation, which is substantial, is that this may not insure these strategies are implemented. However, promising results were found in the small pool of studies in that statistically and clinically important improvements in HFS knowledge were found when different interventions were compared to the control or no intervention groups, in primary school children and families with children at up to 4 months follow up. We also found that there was no immediate difference in HFS knowledge and behavioural improvements between two ways of delivering HFS programs (instructor-led vs. computer-based).
Warda et al. (1999) review concluded that there is a need for intensive program evaluation, especially among school children demographic. In our review, we identified 3 RCTs and 1 prospective cohort study that examined the effectiveness of HFS interventions in this population. However, the magnitudes of intervention effects were different between the two study designs. In the Lamb et al. (2006) prospective cohort study (interventions vs no intervention groups), SMDs of 1.64 (95% CI: 1.44–1.84) and 0.86 (95% CI: 0.68–1.04) were reported for improvements in HFS knowledge and behaviour, respectively. These values were much higher than those reported in the 3 included RCTs (Kendrick et al. 2007; Morrongiello et al. 2012; Hwang 2006). It is likely that the magnitude of intervention effects was over-estimated by Lamb et al. (2006).
All eight trials identified in this review were rated at high risk of bias. The rating of very low-quality evidence per outcome across trials was based on the judgement of serious limitations (risk of bias), very serious imprecision and likely publication bias in all the outcomes across trials. This can be challenging area to conduct RCTs, and it likely that cluster-randomized trials may be needed to evaluate group interventions on a larger scale. Given that multiple approaches are likely to reach and benefit different target audiences, it will require a much larger pool of studies to define the optimal approaches. Despite the limitations in current research, it is reassuring that the methods evaluated have had positive effects on knowledge, and this suggests that the methods that are currently being used at least have a positive effect on this precursor to behaviour change.
We focused on RCTs and prospective cohort studies and did not include conference papers, posters or abstracts. Therefore, there might be a source of publication bias within our search strategy.
The limited evidence supports the use of HFS interventions to improve HFS knowledge and behaviour in children, families with children and adults. Large-scale well-designed randomized controlled trials that consider the unique nature of prevention research and look at behavioural or fire rates as outcomes in larger scale implementation are needed to further assess the effectiveness of HFS interventions.
For more information about PLOS Subject Areas, click here.
We want your feedback. Do these Subject Areas make sense for this article? Click the target next to the incorrect Subject Area and let us know. Thanks for your help!
Is the Subject Area "Fire safety" applicable to this article? Yes No Thanks for your feedback.Is the Subject Area "Medical risk factors" applicable to this article? Yes No Thanks for your feedback.
Is the Subject Area "Schools" applicable to this article? Yes No Thanks for your feedback.Is the Subject Area "Randomized controlled trials" applicable to this article? Yes No Thanks for your feedback.
Is the Subject Area "Wildfires" applicable to this article? Yes No Thanks for your feedback. Is the Subject Area "Children" applicable to this article? Yes No Thanks for your feedback.Is the Subject Area "Preventive medicine" applicable to this article? Yes No Thanks for your feedback.
Is the Subject Area "Systematic reviews" applicable to this article? Yes No Thanks for your feedback.
PLOS is a nonprofit 501(c)(3) corporation, #C2354500, based in San Francisco, California, US