Eligibility criteria

Overviews of reviews, systematic reviews and meta-analyses of experimental (RCTs, non-RCTs and CBAs) and controlled observational (cohort and case–control) studies were eligible for inclusion. Systematic reviews were defined using the Cochrane reviewer’s handbook definition.³⁴² Reviews were eligible if they reported:

1. Non-legislative interventions aimed at the prevention of fire-related injuries, scalds, falls or poisonings among children aged 0–19 years.
2. The use of home safety equipment or other injury prevention practices for the primary, secondary or tertiary prevention of fire-related injuries, scalds, falls or poisonings. Primary prevention referred to preventing the injury-producing events (e.g. fires), secondary prevention referred to preventing an injury occurring during the event (e.g. a smoke alarm does not prevent the fire but can prevent an injury by alerting people and enabling escape from the fire) and tertiary prevention referred to minimising the impact of the injury through the provision of first aid.
3. Interventions within the scope of activities undertaken by children’s centres in England. The remit of children’s centres included the provision of ‘advice on accident and injury prevention’ (p. 12)²¹⁰ (contains public sector information licensed under the Open Government Licence v3.0); interventions beyond this remit, such as legislative interventions, World Health Organization Safe Community programmes or complex home visiting programmes, were excluded.

Primary studies, of the designs described above and published following the date of the most recent comprehensive systematic review, were eligible for inclusion. We searched from the date of the most recent review that included all study designs, interventions and outcomes eligible for our reviews. For fire prevention, these were the reviews published by Towner et al. in 1996³⁵² and Warda et al. in 1999.^353,354 For the prevention of scalds, falls and poisoning we chose the most recent review, which was the 2001 review by Towner et al.³⁶ For first aid interventions only one systematic review was found, which included only four studies;³⁵⁵ hence, searches for primary studies for first aid interventions were run from the date of inception of the databases.

Information sources

We searched MEDLINE, EMBASE, CINAHL, ASSIA, PsycINFO and Web of Science. The dates for which searches were run for each overview are shown in Table 69.

TABLE 69

Dates for running searches for reviews and primary studies for each overview

Searches

Search terms for MEDLINE are provided in Appendix 5, Search terms for the overviews of reviews and primary studies for study H for each overview and systematic review, with strategies adapted as necessary for the other databases. Other sources searched are listed in Appendix 5, Other sources searched for overviews of reviews and primary studies for study H. Searches were not restricted by language or publication status. Articles were translated when necessary.

Study selection

Titles and abstracts of articles were scanned independently by two reviewers to identify articles to retrieve in full. When articles appeared eligible but no abstract was available, full articles were retrieved. Disagreements between reviewers were dealt with by consensus-forming discussions and referral to a third reviewer.

Data collection process

Data were extracted on study design, participants, interventions, comparator groups and outcomes independently by two reviewers using a standardised data extraction form. Disagreements between reviewers were dealt with by consensus-forming discussions and referral to a third reviewer.

Assessment of risk of bias

Assessments of risk of bias were carried out independently by two reviewers. Disagreements between reviewers were dealt with by consensus-forming discussions and referral to a third reviewer.

The risk of bias for reviews was assessed using the Overview Quality Assessment Questionnaire (OQAQ).³⁵⁶ The risk of bias for primary studies was assessed using adequacy of allocation concealment, blinding of outcome assessment and follow-up of at least 80% of participants for RCTs and blinding of outcome assessment, follow-up of at least 80% of participants and balance of confounders between treatment groups for non-RCTs and CBAs. For overviews conducted later in the programme of research (scalds, poisonings and first aid interventions) we assessed risk of bias for primary studies using the criteria specified in the Cochrane reviewers handbook.³⁴² The risk of bias for cohort and case–control studies was assessed using the Newcastle–Ottawa Scale.³⁵⁷

Data synthesis

Data were synthesised narratively. We tabulated the characteristics of included reviews and included primary studies. For each primary study included in a review, we tabulated which reviews it was included in and the outcomes that it reported.

Eligibility criteria

Randomised controlled trials, non-RCTs and CBAs whose participants were children and young people (aged ≤ 19 years) and their families and which evaluated home safety interventions provided by health or social care professionals, school teachers, lay workers or voluntary or other organisations aimed at reducing home injuries or increasing home safety practices were included. To be included, studies had to report injuries, possession and use of home safety equipment or injury prevention practices among their outcome measures. The outcomes of interest for the KCS programme were:

1. fire prevention – possession of a smoke alarm, possession of a functional smoke alarm, checking or changing smoke alarm batteries, possession of fireguards, storage of matches out of reach, possession of a fire extinguisher, possession of a fire escape plan
2. scald prevention – having a safe hot tap water temperature, keeping hot drinks/foods out of reach
3. falls prevention – possession of a fitted safety gate, possession or use of a baby walker, possession of a non-slip bath mat or decals, possession of window safety devices (locks, screens or opening width restrictors, hereafter referred to as window locks), never leaving a child alone on a high surface
4. poisoning prevention – storing medicines out of reach, storing other household products out of reach, storing poisons out of reach, storing plants out of reach, possession of the PCC number.

Information sources

We searched the Cochrane Central Register of Controlled Trials, MEDLINE, EMBASE, PsycINFO, Web of Science, CINAHL, Database of Abstracts and Reviews of Effects, NHS Economic Evaluation Database and Health Technology Assessment database from date of inception to end of May 2009. In addition, a range of websites was searched (to June 2009) and hand searching was carried out of abstracts from the 1st to the 9th World Conferences on Injury Prevention and Control, the journal Injury Prevention (to March 2009) and reference lists of articles included in the review and published systematic reviews. The sources searched are shown in Appendix 5, Other sources searched for study I. Authors of published and unpublished studies were contacted as described in the published review.⁴⁹

Search

The MEDLINE search strategy is provided in Appendix 5, Search strategy for study I. This was adapted as necessary for other databases. Searches were not restricted by language or publication status. Articles were translated when necessary.

Study selection

Titles and abstracts of articles were scanned independently by two reviewers to identify articles to retrieve in full. When articles appeared eligible but no abstract was available, full articles were retrieved. Disagreements between reviewers were dealt with by consensus-forming discussions and referral to a third reviewer.

Data collection processes

Data were extracted independently by two reviewers using a standardised data extraction form. Disagreements between reviewers were dealt with by consensus-forming discussions and referral to a third reviewer. Authors of studies were contacted to supply unpublished data, including individual participant data when this allowed studies to be included in our meta-analyses (e.g. when papers reported composite outcomes such as safety scores but individual items of the composite measure were outcomes in our meta-analyses). When individual participant data were obtained, we formatted these in a standard format.

Assessment of risk of bias

Assessments of risk of bias were made independently by two reviewers. Disagreements between reviewers were dealt with by consensus-forming discussions and referral to a third reviewer.

Data synthesis

Meta-analyses were undertaken when three or more studies reported the same outcome. For injury rates, pooled IRRs were estimated using random-effects models. For CBAs, we estimated follow-up injury rates adjusted for baseline rates. For binary outcome measures (safety equipment possession or use, injury prevention practices), pooled ORs were estimated using random-effect models. Studies were adjusted for clustering as necessary and management of studies with more than one intervention or control arm is described in the published review.⁴⁹ Cases with missing values were excluded from all analyses.

Heterogeneity between effect sizes was described using forest plots, chi-square tests and the I² statistic. Potential explanations for heterogeneity were explored by a priori subgroup analyses, which included whether or not safety equipment was provided, study setting and study quality. Publication and related biases for binary outcomes were investigated using the approach recommended in recent guidelines.³⁵⁸ For meta-analyses of injury rates, when there were ≥ 10 studies we assessed publication bias using Egger’s test. The individual contribution of each study to the pooled result was assessed graphically, and sensitivity analyses were undertaken to assess the effect of removing each study from each analysis. The robustness of the findings with respect to study quality was assessed using subgroup analyses by comparing treatment effects between randomised and non-randomised studies and between RCTs with and without adequate allocation concealment, blinded outcome assessment and follow-up of at least 80% of participants in each study arm.

Study identification

Data were extracted from primary studies identified from the overviews of reviews and systematic reviews of more recently published primary studies undertaken in study H. When individual participant data were used in the PMA for study I, the same data were used in study J.

Statistical methods

Network meta-analyses were undertaken for the following binary outcomes:

1. fire prevention – possession of a smoke alarm, possession of a functional smoke alarm, possession of different types of battery-powered alarms, possession of fireguards, storage of matches out of reach, possession of a fire extinguisher, possession of a fire escape plan
2. scald prevention – having a safe hot tap water temperature, keeping hot drinks/foods out of reach
3. falls prevention – possession of a fitted safety gate, possession or use of a baby walker, possession of a non-slip bath mat or decals, possession of window locks, never leaving a child alone on a high surface
4. poisoning prevention – storing medicines out of reach, storing other household products out of reach, storing poisons out of reach, storing plants out of reach, possession of the PCC number.

Pairwise meta-analyses are usually restricted to calculating a pooled estimate of effectiveness comparing two groups, often an intervention group with a control group. Home safety interventions are often complex and multifaceted interventions, consisting of various combinations of safety education, the provision of free or low-cost safety equipment, fitting of safety equipment and home safety assessments. The control arms used in studies of home safety interventions may include usual care (which can vary between studies), generic safety advice (as opposed to specific or tailored advice) or some, but not all, of the elements of the intervention. Consequently, PMA often involves some ‘lumping’ together of interventions (and control conditions). For example, in PMAs of interventions to promote functional smoke alarms,^33,37,40 interventions that provided education were grouped together with interventions that provided smoke alarms. This included interventions fitting low-cost/free smoke alarms, providing but not fitting low-cost/free smoke alarms, providing home safety inspections and referring to child safety centres. In addition, the educational component of these interventions included very brief face-to-face advice, providing leaflets and videos, providing generic or tailored safety advice or classes or lectures on parenting/child safety. The control conditions included unspecified usual care, well child visits, standard/generic safety advice, leaflets, videos, information about or referral to child safety centres and home safety inspections.

However, commissioners, service providers and parents, among others, need to decide on the ‘best’ intervention for preventing a particular injury. NMA enables comparison of all evaluated interventions simultaneously within a single coherent analysis.^343–345 Suppose we have studies providing effect estimates for a control compared with intervention A and for intervention A compared with intervention B. NMA allows estimation not only of the pooled effects when pairwise evidence exists (direct comparison between control and intervention A and between intervention A and intervention B) but also of effects when interventions are not directly compared but are linked through a connected network of studies (indirect comparison between control and intervention B). Interventions can also be ranked in order of effectiveness. This approach is increasingly being used in health technology assessment when deciding on the optimal intervention strategy for a particular condition.^346,359

Network meta-analysis was used to compare all interventions with one another, using all available data in a connected network of studies and a standard NMA random-effects model with a binary outcome.^344,345 Pooled estimates of intervention effects for all combinations of pairwise comparisons were obtained. Intervention effectiveness was ranked based on absolute intervention effects (derived using an underlying rate based on the usual-care arms) and the probability that each intervention was best for a particular outcome was calculated.³⁴⁴

The between-study SD parameter was used to assess the variability in effect sizes within pairwise comparisons above that expected by chance.^360–362 We assessed and tested the consistency between evidence from studies that directly compared the two treatments under consideration (‘direct’ evidence) and evidence from the remaining studies in the network (‘indirect’ evidence). We also assessed goodness of fit of the model. Analyses were conducted using a Markov chain Monte Carlo (MCMC) method³⁴⁴ and fitted using WinBUGS software [version 1.4.3; Medical Research Council (MRC) Biostatistics Unit, Cambridge, UK].³⁶³ Further technical details of the analysis together with the WinBUGS code are available from the authors on request.

In addition, we have already taken steps to start to address some of the research recommendations arising from the KCS programme to allow for more detailed analysis of the evidence by extending methods for NMAs. These include methods for adjusting for baseline risk, simultaneously incorporating aggregate and individual participant data, exploring effect modifiers and extrapolating evidence across different networks for multiple outcomes. These are not described in this report and the reader is referred to the publications.^364–367

Decision models

Our decision modelling was guided by published principles for good modelling practice and design³⁶⁹ and NICE public health methods guidance.^370,371 We used the software package R (version 2.15.1; R Foundation for Statistical Computing, Vienna, Austria) to construct models and these were assessed using Monte Carlo simulation in R or MCMC simulation in WinBUGS 1.4.3.³⁷² Figure 22 illustrates the three-stage decision model using the example of interventions to promote the prevalence of functional smoke alarms and Figure 23 illustrates the decision model structure within each yearly cycle of the stage 2 (preschool) model.

FIGURE 22

Schematic diagram illustrating the three-stage process for decision modelling using the example of interventions to promote the prevalence of functional smoke alarms. Note: shaded box is an intermediate state that may occur within any one yearly cycle. (more...)

FIGURE 23

Decision model structure within each yearly cycle of the stage 2 (preschool model) model. Note: the oval nodes match to the oval nodes in the schematic diagram in Figure 22.

The first stage of the model, referred to as the intervention model, consisted of a decision tree analysing the effectiveness of interventions to increase the prevalence of functional smoke alarms (informed by NMA undertaken in study J³⁷⁴) and the costs of these interventions. The second stage of the model, referred to as the pre-school model, used outputs of the intervention model as its primary inputs and modelled fire-related injuries (i.e. minor, moderate and severe) and fatalities of children during the preschool period (aged 0–4 years) using a Markov state-transition structure incorporating a decay/repair factor to allow smoke alarms to cease functioning and to be repaired. The third stage of the model, referred to as the long-term model, modelled over an individual’s lifetime, includes costs and health effects of fire-related injuries occurring from birth to 4 years of age, using a Markov state-transition structure. For the Markov models in the second and third stages of the analysis, a yearly cycle duration was used and models were run for 100 years (5 years in stage 2 and 95 years in stage 3). Total costs and QALYs were established for each intervention by attributing costs (inflated to 2012 prices) and quality-of-life weights to each state.

Evidence used to inform the base-case model, together with distribution information when applicable, is presented in Appendix 5, Base-case model inputs for the decision analysis for smoke alarms for study K for smoke alarm interventions, Appendix 5, Base-case model inputs for the decision analysis for safe hot tap water temperatures for study K for safe hot tap water temperature interventions, Appendix 5, Base-case model inputs for the decision analysis for safety gates to prevent stairway falls for study K for safety gate interventions, Appendix 5, Base-case model inputs for the decision analysis for safe storage of medicines for study K for safe storage of medicines and Appendix 5, Base-case model inputs for the decision analysis for safe storage of household products for study K for safe storage of household products. When possible, input parameters were informed by UK-based data. When possible we tried to ensure that input parameters that were common across decision models were consistent but as the models were developed at different times across the project this was not always possible (see Appendix 5). A summary of the base-case methodological assumptions is outlined in Table 70.

TABLE 70

Summary of assumptions for the base-case model for the decision analysis for interventions to promote functional smoke alarms

Main modelling assumptions

We made a series of assumptions during the modelling:

1. The possession of ‘functioning’ smoke alarms was a surrogate/intermediate outcome for household risk of fire-related injury/death. We used previously published evidence to support this assumption.^37,376,377 The evidence used for the decision-analysis models for other interventions is given in the base-case evidence tables in Appendix 5.
2. The probability of a household accepting an intervention was assumed to be the same across all interventions because of a lack of information on the acceptance of different smoke alarm promotion interventions. When this information was available for decision-analysis models for other interventions it was used and is described in the base-case evidence tables in Appendix 5.
3. The benefit of a household having a functioning smoke alarm accrues to a single child aged 0–4 years of age. This ignored the potential benefits for other household members and is therefore a conservative assumption. This assumption also applies to models for interventions for TMVs, safety gates and the safe storage of medicines and household products.
4. As with most Markov models, we assumed that the probability of future fire-related injuries was independent of previous fires or fire-related injuries and remained constant throughout the model time frame (i.e. 5 years for part 2 of the model). This assumption also applies to models for interventions for TMVs, safety gates and the safe storage of medicines and household products.
5. The model allows for only one fire or fire-related injury in a single cycle (i.e. 1 year). This assumption also applies to models for interventions for TMVs, safety gates and the safe storage of medicines and household products.

Uncertainty

Modelling took account of uncertainty around input parameter point estimates through the use of posterior probability distributions from 5000 MCMC simulations for estimates of the effectiveness of interventions from the NMAs and the use of probability distributions based on point estimates and SEs (see Appendix 5) for other model parameters obtained from the literature. The decision model was evaluated by performing 5000 MCMC simulations. Mean costs and mean QALYs were averaged across all 5000 MCMC simulations.

We assessed the robustness of our findings to the assumptions that we made during modelling and to the data that we used to populate models through a series of one-way sensitivity analyses. For interventions to promote functional smoke alarms these included reducing the prevalence of smoke alarms in households prior to interventions from 80% to 50%;³⁴⁷ reducing the probability of accepting the intervention from 90% to 50%;³⁴⁷ reducing the probability of the decay of safety equipment from 0.1 to 0; and increasing the number of children per household from 1 to 1.8 (i.e. the national average).³⁷⁸

For the decision analyses for interventions to promote a safe hot tap water temperature, sensitivity analyses included increasing the probability of accepting non-TMV interventions from 0.74 to 0.9 (as used in the functional smoke alarm model); reducing the probability to 0 of a child incurring a scald when safe hot water is practised, as for the TMV interventions; removing the fixed costs associated with setting up a scheme to promote safe hot water as promotion of safe hot water is likely to be part of a wider home safety scheme; increasing the number of scalds per year to 653 from 582,700 households¹⁴⁰ to reflect the probability of a scald in children in social housing/deprived areas; and including the costs of TMVs and fitting separately (i.e. not part of a refurbishment or rebuild scheme).¹⁴⁰

For the decision analyses for interventions to increase the possession of a fitted safety gate, analyses included reducing the number of safety gates fitted to 1; reducing the cost of education by using the cost of providing a leaflet only; providing low-cost (£5) rather than free safety gates; reducing the fixed costs of the intervention to £40,000; increasing the number of children per household from 1 to 1.8; reducing the probability of possessing a fitted safety gate from 0.56 to 0.45 to reflect the baseline possession of a fitted and used stair gate in families in deprived areas of Nottingham;⁷² and changing the utility deficits by assuming no uncertainty from 0.05, 0.1 and 0.2 for mild, moderate and severe injuries, respectively, to 0.07, 0.19 and 0.34, respectively.¹⁴³

For the decision analysis for interventions to promote the safe storage of medicines, the sensitivity analyses included changing the baseline probability of safe storage from 75% (ascertained from the results of study A) to 93%⁵⁶ or to 50% (assumption); changing the probability that the intervention is accepted from 90% to 50% (assumption); changing the proportion admitted from 63%³⁷⁹ to 83.3% (Philip Miller, ED nurse, Nottingham University Hospitals NHS Trust, 7 April 2014, personal communication); changing the incidence of unintentional poisoning from 18.07 per 10,000 person-years to 44.9 and 48.5 per 10,000 person-years, reflecting the injury rates among those aged < 5 years living in the two most disadvantaged quintiles based on the Townsend deprivation index;⁹¹ changing the number of cupboard locks from one Pop-It™ lock (costing £2.65) (Kid Rapt, Usk, UK) and one magnetic lock (costing £4.80) to two Pop-It locks (costing £5.30) and two magnetic locks (costing £9.60); increasing the number of children in a household from 1 to 1.8;³⁷⁸ and increasing the uncertainty associated with the utility decrements for poisoning injuries from 10% of the utility decrement value to, first, 20% and, second, 50%.

Finally, for the decision analyses for interventions to promote the safe storage of household products, sensitivity analyses included changing the baseline probability of safe storage from 45% (ascertained from the results of study A) to 93%;⁵⁶ changing the probability that the intervention is accepted from 90% to 50% (assumption); changing the proportion admitted from 33.2%² to 83.3% (Philip Miller, personal communication); increasing the incidence of unintentional poisoning from 12.04 per 10,000 person-years to 44.9 and 48.5 per 10,000 person-years, reflecting the injury rates among those aged < 5 years living in the two most disadvantaged quintiles based on the Townsend deprivation index;⁹¹ and increasing the number of children in a household from 1 to 1.8.³⁷⁸

Study H

The overview included nine systematic reviews reporting interventions to prevent fire-related injuries,^{33,36,37,41,353,354,384,387,391} which, between them, drew on data from only four primary studies. No meta-analyses that reported the effect of interventions on fire-related injuries were found. The reviews found two case–control studies reporting the risk of injury or death in homes with and without smoke alarms. Both found a two- to threefold increase in the odds of a fatal house fire compared with a non-fatal house fire in households with smoke alarms.^50,425 The reviews found inconsistent or insufficient evidence about whether or not interventions promoting smoke alarms were associated with a reduction in fire-related injuries, from one RCT²⁶¹ and one CBA.⁴²³ The CBA reported an increased prevalence of smoke alarm use, which was also associated with a reduction in fire-related injuries.⁴²³ This finding was not confirmed by the RCT, but this may be explained by the ineffectiveness of the intervention in the trial, as there was no significant increase in the prevalence of smoke alarm use.²⁶¹

Two case–control studies explored the odds of fatal house fires related to smoking and alcohol use.^50,394 The first found significant increases in the risk of a fatal house fire when smoking was the source of ignition of the fire and when household members were impaired by drugs or alcohol.⁵⁰ The second found a significantly greater risk of a fatal household fire among households with smokers than among those with non-smokers, a dose–response relationship with the number of smokers in the household and an increased risk with smoking 10–19 cigarettes per day or ≥ 20 cigarettes per day compared with non-smokers. The same study found no significant association between the number of alcoholic drinks consumed per household member and risk of fatal house fire.³⁹⁴

There were insufficient studies to undertake PMA or NMA for the effect of home safety interventions on fire-related injuries in studies I and J.

Study H

The overview included three systematic reviews and meta-analyses^33,37,40 and nine systematic reviews^{36,41,42,331,353,382,383,387,391} reporting the effect of interventions on a range of smoke alarm outcomes. Evidence from the meta-analyses indicates that interventions to promote smoke alarm ownership and function are effective. The first meta-analysis combined effect sizes on smoke alarm ownership from five studies (OR 1.74, 95% CI 1.03 to 2.96),⁴⁰ the second meta-analysis combined effect sizes on smoke alarm ownership from 10 studies (OR 1.21, 95% CI 0.89 to 1.64)³⁷ and the third meta-analysis combined effect sizes on having a functional smoke alarm from 13 studies (OR 1.85, 95% CI 1.24 to 2.75).³³ Subgroup analyses indicate that interventions provided in the clinical setting⁴⁰ or as part of routine child health surveillance³⁷ are effective, whereas those delivered in other settings may be less effective.

The majority of systematic narrative reviews concluded that a diverse range of interventions to promote smoke alarm ownership and function were effective.^{36,353,382,383,387,391} One systematic review confined to examining the effect of home safety equipment and risk assessment schemes³⁸⁷ concluded that there is inconsistent evidence from six robust studies, using observed outcomes and a control group, about the presence of functional alarms. Four suggested that the intervention increased functioning smoke alarm presence and two suggested no significant impact on smoke alarms.

A total of 39 primary studies reporting smoke alarm ownership or functioning were identified (29 from reviews^{71,72,257,261,265,267–271,273,275,277,279,281,282,284,285,287,295,297,300,304,305,393,395,403,409,423,426} and 10 from additional searches for primary studies;^{266,286,288,310,397,398,412,418,431,433}). The studies not included in a review suggest that there is conflicting evidence about the effect of type of smoke alarm on smoke alarm functioning.^286,310,433 Two RCTs found that alarms with lithium batteries were more likely to remain functional than those with carbon zinc batteries.^310,433

Seven primary studies reporting the effectiveness of interventions in promoting other smoke alarm outcomes were included in the overview. The outcomes reported include checking or changing smoke alarm batteries, false alarms, reasons for non-functional alarms, redeeming vouchers for free alarms, awakening to alarms and time to awaken or escape. There appears to be inconsistent evidence about the effect of interventions on checking or changing smoke alarm batteries.^{282,287,288,295} One RCT found that photoelectric alarms were significantly less likely to cause false alarms than ionising alarms.³¹⁰ One RCT found that installing smoke alarms was more effective at increasing the prevalence of functional smoke alarms than providing vouchers for families to redeem against alarm purchase, with 47% of vouchers not redeemed.²⁶⁸ Finally, one RCT found that children are significantly more likely to awaken to a parent-voice alarm than to a standard smoke alarm and that time to awaken and time to escape were significantly shortened with parent-voice compared with standard alarms.⁴³⁶

Study I

Pairwise meta-analysis of 17 studies evaluating home safety interventions (education plus provision of smoke alarms and home safety inspections in some studies),^{71,72,261,265,266,269,271,275,277,284,285,297,304,395,398,426,431} which updated the meta-analysis by Kendrick et al.³³ described in the previous section, found that interventions significantly increased functional smoke alarm ownership (OR 1.81, 95% CI 1.30 to 2.52) (Figure 26). Interventions that provided smoke alarms appeared to be more effective (OR 2.49, 95% CI 1.53 to 4.06) than those that did not (OR 1.12, 95% CI 0.87 to 1.45). Seventeen studies reported ownership of alarms (as opposed to functional status)^{71,261,266,267,269,270,273,277,282,287,288,295,305,395,403,418,426} and meta-analysis showed that interventions may be associated with a small increase in smoke alarm ownership (OR 1.17, 95% CI 0.97 to 1.42). Publication bias may have occurred in the functional smoke alarm ownership analysis (p-value for the tests of asymmetry = 0.063 and inspection of the funnel plot indicated the possibility of missing studies with non-significant findings), but findings remained significant after application of the regression bias adjustment method (OR 1.44, 95% CI 1.07 to 1.92).

FIGURE 26

Forest plot of effect sizes for possession of a functional smoke alarm from studies evaluating home safety educational interventions. M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana FA, Cooper (more...)

Four studies that reported checking or changing smoke alarm batteries in the last 6 months were included in a meta-analysis as part of study I.^{282,287,288,295} There was no evidence that home safety interventions were effective in increasing the proportion of families who had checked or changed smoke alarm batteries in the preceding 6 months (OR 1.15, 95% CI 0.63 to 2.08) (Figure 27).

FIGURE 27

Forest plot of effect sizes for smoke alarm batteries being checked or changed from studies evaluating home safety educational interventions. M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana FA, (more...)

Study J

Network meta-analysis was used to determine the effect of component elements of home safety interventions on having a functional smoke alarm in 20 of the studies identified from the overview.^{71,72,261,265,266,268,269,271,275,277,284,285,297,300,304,395,398,409,426,431} Interventions used in these studies were categorised into seven distinct groupings and the NMA estimated the 21 possible pairwise comparisons between the seven interventions reported in the 20 studies. The data used in the NMA from each study for fire prevention outcomes are shown in Table 73. The pooled estimates, 95% CrIs and, when available, direct within-trial estimates are reported in Table 74. The most intensive intervention (home safety education plus equipment provision plus fitting of safety equipment plus home safety assessment) was the most likely to be the most effective (probability = 0.66), with an OR compared with usual care of 7.15 (95% CrI 2.40 to 22.73).

TABLE 73

Summary of studies and their data included in the NMA of the interventions to promote possession of functional smoke alarms

TABLE 74

Pooled ORs (95% CrIs) from NMAs comparing the effect of different interventions on possession of a functional smoke alarm

Three studies evaluated the effect of alarm type and battery type on alarm function.^286,310,433 The data used in the NMA from each study are shown in Table 75 and the pooled estimates, 95% CrIs and, when available, direct within-trial estimates are reported in Table 76. Ionisation smoke alarms with lithium batteries were most likely to be the best type for increasing possession of a functioning alarm (p best = 0.69), although there was considerable uncertainty in these estimates as shown by the wide 95% CrIs in Table 76.

TABLE 75

Summary of studies and their data included in the NMA of types of battery-powered smoke alarms

TABLE 76

Pooled ORs (95% CrIs) from NMAs of types of battery-powered smoke alarms

Study K

The findings from the decision analysis evaluating the cost-effectiveness of different interventions for promoting possession of functional smoke alarms are described below. Four of the seven interventions evaluated in the base-case analysis had higher costs or higher ICERs than more effective interventions (namely education plus free safety equipment and education plus free safety equipment and fitting of equipment plus home safety inspection) and were therefore excluded from further consideration (Table 77). Of the remaining three interventions, education plus free safety equipment had the lowest estimated ICER compared with usual care, at £34,200 per QALY gained.

TABLE 77

Base-case analysis results (probabilistic) for the cost-effectiveness of interventions for promoting possession of functional smoke alarms

Figure 28 shows the probability of the alternative interventions being cost-effective. At a threshold value of £30,000 per QALY gained, usual care has the highest probability of being cost-effective (0.62). However, when this threshold value is increased to £50,000, education plus low-cost/free equipment has the highest probability of being cost-effective (0.69), demonstrating that there is considerable uncertainty in decisions within the £30,000–50,000 threshold range.

FIGURE 28

Cost-effectiveness acceptability curves for interventions promoting possession of functional smoke alarms. E, education; F, fitting; FE, low-cost/free equipment; HSI, home safety inspection; UC, usual care. Adapted from Saramago et al. under the Creative (more...)

Sensitivity analysis

We undertook a series of sensitivity analyses assessing the robustness of the findings to modelling assumptions and data used in the modelling, as described in the methods section (Table 78). Dominated and extendedly dominated interventions have been removed from Table 78. All sensitivity analyses assessed the probability of interventions being cost-effective at a threshold of £30,000. Reducing the prevalence of smoke alarms in UK households from 80% to 50% resulted in ICERs that increased as the intensity of the intervention increased from education to education plus free equipment and fitting plus home safety inspection; however, for any of these interventions to be adopted, decision makers needed to be willing to pay or displace large amounts of funds (i.e. ICERs were ≥ £180,000 per additional QALY).

TABLE 78

Sensitivity analysis results for interventions promoting the possession of functional smoke alarms

As the probability of accepting interventions varied considerably between the trials that informed the effectiveness model input parameters, the effect of reducing the acceptance rate from 90% to 50% was assessed. This resulted in a reduction in the ICER from £34,200 in the base case to £12,701 for education plus low-cost/free equipment compared with usual care. Reducing the probability of decay/repair of the safety equipment from 0.1 to 0 resulted in all interventions being dominated by education, which had an ICER of £80,117 compared with usual care. An increase in the number of children aged < 5 years per household from 1 to 1.8³⁷⁸ resulted in a reduction in the ICER from £34,200 to £4456 for education plus low-cost/free equipment compared with usual care.

Study H

The overview included one systematic review and meta-analysis³³ and three systematic reviews^36,42,387 reporting the effect of interventions on the possession and use of fireguards. The meta-analysis reported some evidence that home safety education was effective in increasing the use of fire guards based on the combined effect sizes from four studies (OR 1.40, 95% CI 1.00 to 1.95), all of which included interventions that provided fireguards.³³ Of the four narrative reviews, one reported weak evidence that a free or discounted supply of fireguards, in conjunction with safety education, increased the use of fireguards and weak evidence that free home safety equipment and installation with safety education had no impact on the fitting and use of fireguards.³⁸⁷ One review reported no significant increase in the use of fireguards³³ and the remaining two reviews did not draw any specific conclusions about the effect of interventions on the use of fireguards.^36,42

Six primary studies reported the effect of interventions on the possession and use of fireguards (all from reviews^{71,72,275,292,295,417}). One study reported a significant increase in the use of fireguards following the provision of safety advice, leaflets and low-cost equipment including fireguards.⁷¹ Five studies reported no significant effect of interventions, all of which included home safety education and four of which also provided free or discounted safety equipment, with two of these including the provision of fireguards^72,275 and two not providing fireguards.^292,295 The one study not providing equipment involved information cards, fridge magnets and checklists in addition to safety education.⁴¹⁷

Study I

The updated searches for study H did not find any additional studies evaluating home safety education, which also included the provision of fireguards. Therefore, the PMA includes the same four studies^{71,72,275,417} reported in Kendrick et al.³³ There was some evidence that home safety interventions were effective in increasing the use of fire guards (OR 1.40, 95% CI 1.00 to 1.95) (Figure 29).

FIGURE 29

Forest plot of effect sizes for use of fire guards from studies evaluating home safety educational interventions. M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana FA, Cooper NJ, Hubbard SJ, Sutton (more...)

Study J

Network meta-analysis was used to determine the effect of component elements of home safety interventions on use of fireguards in four studies identified from the overview.^{71,72,275,417} These studies included five interventions and the NMA estimated the 10 possible pairwise comparisons between the five interventions reported in the four studies. The data used in the NMA from each study are shown in Table 79. The pooled estimates, 95% CrIs and, when available, direct within-trial estimates are reported in Table 80. None of the interventions differed significantly from each other. The intervention with the highest probability of being the most effective was education plus equipment plus home safety inspection (p best = 0.28), but the probabilities were very similar for all interventions (range 0.20–0.28) except for education plus low-cost/free equipment (p best = 0.05).

TABLE 79

Summary of studies and their data included in the NMA of fireguards

TABLE 80

Pooled ORs (95% CrIs) from NMA comparing the effect of different interventions on fireguard use

Study K

Decision analyses were not undertaken for interventions to promote fireguard use as no interventions were found to be effective in NMAs.

Study H

The overview included one systematic review and meta-analysis³³ and one systematic review⁴² reporting the effect of interventions on the possession of fire extinguishers. The meta-analysis, which combined effect sizes from four studies, reported that home safety education interventions, one of which included the provision of fire extinguishers, were not effective in increasing the possession of fire extinguishers (OR 0.95, 95% CI 0.40 to 2.23).³³ One systematic review³³ reported mixed findings from studies not included in the meta-analysis, but effect sizes were not reported. The other review⁴² included one relevant primary study but did not draw any conclusions specific to the possession of fire extinguishers.

Nine primary studies reported the effect of interventions to promote the possession of fire extinguishers; seven of these studies were identified from reviews^{269,277,279,293,304,399,441} and two from additional searches for primary studies.^257,412 One study found a significant effect on the possession of fire extinguishers of an intervention involving home safety education, home hazard assessment and the provision of items of equipment including fire extinguishers.²⁶⁹ Other studies delivering home safety education, hazard identification and assistance with obtaining safety equipment did not report a significant effect on the possession of fire extinguishers.^{257,277,279,304,441} Two studies involving a community injury prevention programme including seminars, workshops, courses and home visits²⁹³ and an intervention involving a home visit from fire service personnel focusing on smoke alarms and fire escape plans⁴¹² both reported no increase in the possession of fire extinguishers. One study did not report the effect of the intervention on the possession of fire extinguishers.³⁹⁹

Study I

Pairwise meta-analysis of five studies evaluating home safety education^{257,269,277,293,304} found a lack of evidence that interventions were effective in increasing the possession of fire extinguishers (OR 0.90, 95% CI 0.53 to 1.51) (Figure 30).

FIGURE 30

Forest plot of effect sizes for possession of a fire extinguisher from studies evaluating home safety educational interventions. M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana FA, Cooper NJ, (more...)

Study J

Network meta-analysis was used to determine the effect of the component elements of home safety interventions on the possession of fire extinguishers in four studies identified from the overview.^{257,269,277,293} These studies included six interventions and the NMA estimated the 15 possible pairwise comparisons between the six interventions reported in the four studies. The data used in the NMA from each study are shown in Table 81. The pooled estimates, 95% CrIs and, when available, direct within-trial estimates are reported in Table 82. None of the interventions differed significantly from each other. The intervention with the highest probability of being the most effective was community campaign plus education plus home visit (p best = 0.63).

TABLE 81

Summary of studies and their data included in the NMA of fire extinguishers

TABLE 82

Pooled ORs (95% CrIs) from NMA comparing the effect of different interventions on the possession of fire extinguishers

Study K

Decision analyses were not undertaken for interventions to promote the possession of fire extinguishers, as none of the interventions was found to be effective in NMA.

Study H

Two meta-analyses^33,40 and one systematic review³⁵⁴ reporting the effect of interventions on the storage of matches out of reach of children were included in the overview. One meta-analysis⁴⁰ reported that there was a modest but non-significant effect on safe storage of matches based on the findings of two studies reporting this outcome (no effect size reported). The second meta-analysis³³ pooled the findings of five studies and found a lack of evidence that home safety inspection was effective in increasing the safe storage of matches (OR 1.23, 95% CI 0.56 to 2.68). The systematic review³⁵⁴ identified only one relevant study, which did not find a significant effect of the intervention on the safe storage of matches.

Six primary studies reporting the effect of interventions on the safe storage of matches were identified from reviews.^{269,273,275,277,304,404} No further studies were identified from additional searches for primary studies. All six studies found no significant effect on the safe storage of matches of interventions involving safety education,^273,404 provision of a safety kit³⁰⁴ and safety education combined with home hazard checks and provision of equipment or assistance with obtaining equipment.^269,275,277

Study I

Pairwise meta-analysis of six studies reporting the effect of interventions on the storage of matches out of reach of children^{269,273,275,277,304,404} found a lack of evidence that home safety education was effective (OR 1.03, 95% CI 0.63 to 1.68) (Figure 31).

FIGURE 31

Forest plot of effect sizes for storing matches out of reach of children from studies evaluating home safety educational interventions. M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana FA, Cooper (more...)

Two primary studies reported other matches-related outcomes, neither of which was included in a systematic review. One study provided training and teaching resources to teachers of children and reported a significant improvement in the number of children never using matches.⁴¹⁸ The other study provided a home visit from fire service personnel and did not report any reduction in the rate of lighting of matches or lighters.⁴¹²

Study J

Network meta-analysis was used to determine the effect of component elements of home safety interventions on the safe storage of matches in five studies identified from the overview.^{269,273,275,304,404} These studies included four interventions and the NMA estimated the six possible pairwise comparisons between the four interventions reported in the five studies. The data used in the NMA from each study are shown in Table 83. The pooled estimates, 95% CrIs and, when available, direct within-trial estimates are reported in Table 84. None of the interventions differed significantly from each other. The intervention with the highest probability of being the most effective was education + equipment + home safety inspection (p best = 0.40).

TABLE 83

Summary of studies and their data included in the NMA of the safe storage of matches

TABLE 84

Pooled ORs (95% CrIs) from NMA comparing the effect of different interventions on the storage of matches out of reach of children

Study K

Decision analyses were not undertaken for interventions to promote the storage of matches out of reach of children, as none of the interventions was found to be effective in NMA.

Study H

Two systematic narrative reviews reported outcomes related to having or practising a fire escape plan.^33,387 The first review³³ found four studies reporting this outcome, two of which reported a significant difference that favoured the intervention group. The second review³⁸⁷ included one relevant study and reported that home risk assessment and free/discounted supply and installation of safety equipment had a positive effect on having a fire escape plan.

A total of six primary studies reporting fire escape plan outcomes were identified, five from systematic reviews^{293,295,300,399,441} and one from additional searches for primary studies.⁴¹² Interventions that were effective included multimedia first aid and home safety training,³⁹⁹ a home visit from fire service personnel,⁴¹² a community programme including safety seminars for parents, workshops with teachers, courses for schoolchildren and weekly home visits²⁹³ and a home visit involving safety education, modification and provision of safety items.³⁰⁰ Two studies that did not demonstrate an effect of interventions on having or practising a fire escape plan both involved home safety education that targeted a range of injuries, one with a safety kit²⁹⁵ and one with hazard identification and free safety equipment.⁴⁴¹

Study I

Pairwise meta-analysis of four studies reporting on having or practising a fire escape plan^{293,295,399,412} found that home safety interventions increased the proportion of families with a fire escape plan (OR 2.01, 95% CI 1.45 to 2.77) (Figure 32).

FIGURE 32

Forest plot of effect sizes for having a fire escape plan from studies evaluating home safety educational interventions. IV, inverse variance. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana FA, Cooper NJ, Hubbard SJ, Sutton (more...)

Study J

Network meta-analysis was used to determine the effect of component elements of home safety interventions on the possession of a fire escape plan in three studies identified from the overview.^293,295,399 These studies included four interventions and the NMA estimated the six possible pairwise comparisons between the four interventions reported in the three studies. The data used in the NMA from each study are shown in Table 85. The pooled estimates, 95% CrIs and, when available, direct within-trial estimates are reported in Table 86. None of the interventions differed significantly from each other. The intervention with the highest probability of being the most effective was usual care (p best = 0.53).

TABLE 85

Summary of studies and their data included in the NMA of a fire escape plan

TABLE 86

Pooled ORs (95% CrIs) from NMA comparing the effect of different interventions on possession of a fire escape plan

Study K

Decision analyses were not undertaken for interventions to promote the possession of fire escape plans, as none of the interventions was found to be effective in NMAs.

Study H

No meta-analyses were found for other outcomes related to fire prevention practices. One narrative review³⁵⁴ reported interventions to teach safer fire responses, based on six primary studies.^{411,414,415,428,429,440} It concluded that school-based programmes using active participation of children in learning fire responses were more effective than programmes using passive methods. It also concluded that skill retention was poor but was improved by periodic repetition and by the addition of fear reduction techniques and teaching the rationale behind behaviours. Two primary studies reporting this outcome were identified from further searches. One of these studies found a significant improvement in the demonstration of the correct action to take in a clothing fire and in knowledge of the correct actions to take in a house fire following delivery of a school-based injury prevention curriculum.⁴¹⁸ The other study reported that a significantly higher proportion of children who visited a learning centre performed the correct response and displayed the correct knowledge about a fire escape routine.⁴²¹

The overview included one review reporting the effectiveness of interventions to prevent fire setting or match play,³³ which included three primary studies.^392,406,419 Two of these studies reported significant reductions in the incidence of fire setting or match play behaviour that favoured the intervention group.^406,419

The overview included one systematic review³³ and six primary studies^{288,292,295,401,412,418} reporting outcomes related to cooking safety. One study found that significantly more intervention arm parents childproofed their boiler and rice cooker.⁴⁰¹ No significant differences were found between treatment arms for other cooking safety outcomes, including children cooking on the stove⁴¹² or without an adult present,⁴¹⁸ for keeping children away from the stove,²⁸⁸ for turning pan handles away from the room^288,295 or for using cooker guards.²⁹²

Three studies reported outcomes related to the safe use of paraffin appliances.^289,301,303 All reported composite scores of paraffin safety practices, with two failing to find a significant difference between intervention arms^289,303 and the third finding significantly safer scores in intervention arm families.³⁰¹ Two studies also reported the individual items that constituted the paraffin safety scores.^301,303 These included a range of safety practices, none of which was found to differ significantly between treatment arms in either study.

Two studies reported candle safety,^295,303 with neither finding a significant difference between treatment groups for leaving burning candles in an empty room,²⁹⁵ leaving candles on unstable surfaces³⁰³ or using candles < 30 cm from flammable materials.³⁰³

Two studies reported electrical safety outcomes.^289,295 One of the studies reported no significant difference between intervention arms in the proportion of families with overloaded electrical sockets²⁹⁵ and the other reported a composite electrical safety score, failing to find a significant difference between intervention arms.²⁸⁹

One study reported two outcomes related to safe smoking practices, finding no significant difference between intervention arms with respect to smoking in bed or safe disposal of ashes.²⁹⁵ The same study also found no significant difference between intervention arms in the proportion of families using an oven to heat the kitchen.²⁹⁵ Finally, one study reported no significant difference between intervention arms in the safe storage of irons.²⁷⁵

It was not possible to undertake pairwise meta-analyses (study I), NMA (study J) or decision analyses (study K) for any of these other fire prevention outcomes because of the small number of studies reporting each outcome.

Study H

Four narrative reviews^{36,49,337,385} reported the effectiveness of interventions for the prevention of scald injuries based on two primary studies.^312,442 One review of home safety interventions concluded that there was a lack of evidence that home safety interventions were effective in reducing rates of thermal (fire and scald) injuries.⁴⁹ One review concluded that there is little evidence that educational approaches alone have achieved any reductions in rates of burn and scald injuries.³⁶ One review concluded that there was a paucity of studies of the effectiveness of community-based injury prevention programmes for preventing burns and scalds in children.³³⁷ The final review drew no conclusions specific to scald injury prevention.³⁸⁵ No meta-analyses reported the effect of interventions on scald injuries. The first primary study was a CBA study and reported a reduction in the number of scalds and the severity of scald injuries following a community campaign and education from public health nurses at home visits and at childhood immunisations promoting lowering of tap water thermostat temperature.⁴⁴² However, the statistical significance of these findings was not reported. The second study, a RCT, found a significant reduction in the occurrence of scalds and burns in an intervention group receiving school-based health education delivered to children and parents.³¹²

It was not possible to undertake PMA, NMA or decision analyses for interventions to prevent scalds because of the small number of studies.

Study H

All 14 reviews reported the effect of interventions on safe hot water temperature,^{36,40,49,331,332,337,338,340,382,383,385–388} two of which included meta-analyses combining effect sizes for having a safe hot tap water temperature.^40,49 Both meta-analyses showed a significant effect favouring the intervention group, with pooled ORs of 2.32 (95% CI 1.46 to 3.68) based on combining effect sizes from four studies⁴⁰ and 1.41 (95% CI 1.07 to 1.86) based on combining effect sizes from 16 studies,⁴⁹ four of which were the same studies as in the DiGuiseppi and Roberts review.⁴⁰ Three reviews concluded that interventions had a positive effect on safe hot tap water temperature.^36,331,382 One review recommended periodic counselling of parents on measures to reduce the risk of unintentional injuries from hot tap water.³⁸⁸

Twenty-nine primary studies reported the effect of interventions promoting safe hot water temperature, 26 of which were identified from reviews^{257,264,265,272,273,275–277,285,287,288,292,295,297,300,304,305,307,395,422,427,431,432,435,439} and three of which were identified from additional searches for primary studies.^99,402,424 Eighteen of the studies gave an explicit definition of safe water temperature, ranging from ≤ 46 °C²⁷⁶ to ≤ 60 °C.³⁰⁷

Interventions were effective at promoting safe hot tap water temperature in six studies.^{257,276,277,287,431,439} In three of these studies, interventions combined safety education, a home safety assessment and items of equipment that targeted a number of hazards.^257,277,431 Of these, one study reported the provision of an item relating to hot tap water temperature (water temperature card).²⁵⁷ The other interventions demonstrating a significant effect included an educational leaflet on bathwater scalds plus the fitting of a TMV²⁷⁶ and a handout targeting burn and scald prevention combined with a 10-minute²⁸⁷ or 1-hour⁴³⁹ safety lecture.

Families in the intervention groups in five studies were significantly more likely to check or test hot tap water temperature compared with control group families.^{257,272,295,422,435} Of these studies, one provided safety education, a home safety assessment and safety items including a water temperature card²⁵⁷ and one delivered safety counselling to prevent a range of injuries and a free home safety kit, which included a bathwater thermometer and bath tap spout covers among other items.²⁹⁵ The other studies reporting a significant effect evaluated interventions providing a bathwater thermometer during paediatric clinic consultations,²⁷² a 1-minute educational message about tap water scalds delivered in maternity wards⁴³⁵ and a national 1-week safety campaign delivered via the media, community partners and retail stores, where free water temperature testing cards were available.⁴²²

The studies described above also reported that significantly more intervention families lowered their hot water temperature⁴²² and significantly more intervention families used spout covers for bath taps.²⁹⁵

Nineteen studies reported no significant effect of interventions on a range of outcomes related to safe hot water temperature.^{99,264,265,273,275,285,287,288,292,297,300,304,307,395,402,424,427,432} They evaluated integrated or individual interventions including home visits, safety checks, safety education, counselling and safety devices.

Study I

The PMA evaluating home safety interventions (education plus provision of home safety inspections and safety equipment in some studies) included the same 16 studies as the meta-analysis by Kendrick et al.⁴⁹ described in study H. Intervention group families were more likely to have a safe hot tap water temperature than control group families (OR 1.41, 95% CI 1.07 to 1.86) (Figure 34).

FIGURE 34

Forest plot of effect sizes for safe hot tap water temperature from studies evaluating home safety educational interventions. M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana FA, Cooper NJ, Hubbard (more...)

Study J

Network meta-analysis explored the effects of component elements of the interventions on safe hot water temperature among 20 studies.^{257,264,265,272,273,275–277,287,288,297,300,304,305,307,395,402,431,432,439} The data used in the NMA from each study are shown in Table 87. Table 88 reports the pooled estimates, 95% CrIs and, when available, direct within-trial estimates. The NMA estimated the 36 possible pairwise comparisons between the nine included interventions. Home safety education plus free or low-cost provision and fitting of TMVs was most likely to be effective (p best = 0.97) with an OR compared with usual care of 38.82 (95% CrI 3.58 to 599.10).

TABLE 87

Summary of studies and their data included in the NMA of the interventions promoting a safe hot water temperature

TABLE 88

Pooled ORs (95% CrIs) from NMA comparing the effect of different interventions on safe hot water temperature

Study K

This section reports the findings from the decision analysis evaluating the cost-effectiveness of different interventions promoting a safe hot tap water temperature. Seven of the nine interventions evaluated in the base-case analysis had higher costs than more effective interventions and were therefore excluded from further consideration (Table 89). Of the remaining two interventions, education had the lowest estimated ICER compared with usual care, at £40,271 per QALY gained.

TABLE 89

Base-case analysis results (probabilistic) for the cost-effectiveness of interventions promoting a safe hot tap water temperature

Figure 35 shows the probability of the alternative interventions being cost-effective. At a threshold value of £30,000 per QALY gained, usual care has the highest probability of being cost-effective (0.75). However, when this threshold value is increased to £50,000, education has the highest probability of being cost-effective (0.54), demonstrating considerable uncertainty in the decisions within the £30,000–50,000 threshold range.

FIGURE 35

Cost-effectiveness acceptability curves for interventions to promote safe hot tap water temperature. E, education; F, fitting; FE, equipment (not scald related); HSI, home safety inspection; Th, equipment (thermometer); UC, usual care.

Sensitivity analysis

A range of sensitivity analyses varying the base-case assumptions and inputs, as outlined in the methods section, were implemented (Table 90). Reducing the probability of a scald to zero for households with a safe hot water temperature for all interventions (SA2), not just TMV interventions, and removing the fixed costs of setting up a safe hot water scheme (SA3) resulted in the ICER for education compared with usual care decreasing from £40,271 per QALY gained to £30,571 and £23,975. The cost-effectiveness results were found to be very sensitive to the probability of a scald. When this parameter was changed to the estimated probability of a scald among children living in social housing (653/582,700 from Phillips et al.¹⁴⁰) to reflect provision of the intervention to families living in social housing, the ICER for education plus TMV and fitting compared with usual care reduced from £68,455 to –£20,828 (i.e. cost saving).

TABLE 90

Sensitivity analysis results for interventions promoting a safe hot tap water temperature

Study H

The overview included one systematic review and meta-analysis⁴⁹ and three systematic reviews^337,340,387 reporting the effect of interventions on the safe handling of hot food and drinks. The meta-analysis estimated the pooled OR for six studies for the effect of home safety education on keeping hot food and drinks out of the reach of children.⁴⁹ Families receiving home safety education were not significantly more likely to keep hot drinks out of the reach of children than control group families (OR 0.95, 95% CI 0.61 to 1.48). No reviews drew conclusions specific to the effectiveness of interventions for promoting the safe handling of hot food or drinks.

A total of 10 primary studies were identified for the overview, eight from reviews^{257,269,275,287,288,295,422,444} and two from additional searches for primary studies.^400,410

One study reported that significantly more families in the intervention group than control families tested the temperature of food prepared in a microwave oven.⁴⁴⁴ The study delivered an intervention that included home visits with active guidance on injury prevention and regular monthly telephone follow-ups. The remaining eight studies evaluated a range of interventions including home safety education, tailored safety advice, home safety assessments, the provision of discounted or free home safety equipment and exposure to a child safety campaign and all reported no significant differences in the safe handling of hot food and drinks between the intervention group and the control group.^{257,269,275,287,288,400,410,422}

Study I

Pairwise meta-analysis of six studies evaluating home safety education, the same six studies as in the meta-analysis of Kendrick et al.⁴⁹ described in study H, found no effect of interventions promoting the safe handling of hot food and drinks (OR 0.95, 95% CI 0.61 to 1.48) (Figure 36).

FIGURE 36

Forest plot of effect sizes for keeping hot food or drinks out of reach of children from studies evaluating home safety educational interventions. M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana (more...)

Study J

Network meta-analysis was used to determine the effect of component elements of home safety interventions explored in the six studies included in the PMA on keeping hot food and drinks out of the reach of children.^{257,269,275,287,288,295} The NMA estimated the six possible pairwise comparisons between the four interventions reported in the six studies, finding no significant difference between any of the interventions. The data used from each study are shown in Table 91 and the pooled estimates and 95% CrIs are shown in Table 92.

TABLE 91

Summary of studies and their data included in the NMA of interventions promoting the safe handling of hot food and drinks

TABLE 92

Pooled ORs (95% CrIs) from NMA comparing the effect of different interventions for promoting the safe handling of hot food and drinks

Study K

As none of the interventions for promoting the safe handling of hot food and drinks was found to be effective in the NMA, decision analyses were not undertaken for this outcome.

Study H

The overview included nine systematic reviews reporting on the effectiveness of interventions to promote kitchen and cooking safety practices.^{36,49,331,332,337,338,340,385,387} The reviews drew no conclusions specific to the effectiveness of interventions to promote kitchen and cooking safety practices and no meta-analyses were identified for these outcomes.

Eight primary studies reported kitchen and cooking safety practices, with six identified from the reviews^{260,288,295,418,422,444} and two identified from additional searches for primary studies.^99,400 One of the eight studies reported that intervention group families were significantly more likely to have childproofed electrical kitchen heating devices (e.g. boiler, rice cooker).⁴⁴⁴ Intervention group families received a programme that involved four quarterly home visits and monthly telephone follow-ups targeting a range of injuries including scalds. Another study found that intervention group homes were significantly more likely to have a ‘child-protected’ cooker (not defined) and to have removed objects that could allow a child to climb and reach the sink following a group scald and burn prevention workshop and a home visit delivering tailored child injury prevention information.⁴⁰⁰ No significant differences between intervention and control groups with regard to kitchen and cooking safety practices were reported in the remaining studies. These included evaluations of the effectiveness of a school-based injury prevention programme for improving practices of children when cooking without an adult,⁴¹⁸ home safety assessments, education plus discount vouchers for safety, equipment on use of cooker guards and on keeping heating devices out of the reach of children,²⁷⁷ an ED-based home safety intervention to promote cooking on the back burners of cookers or turning pan handles towards the back of the cooker,²⁹⁵ tailored home safety education about keeping children away from the cooker or oven or turning pan handles away from the edge of the cooker²⁸⁸ and a scald and burn prevention media campaign to promote using the back burners of cookers, keeping children out of the kitchen when cooking, turning pot handles to the back of the cooker and removing dangling cords of heating devices.⁴²² Finally, there were no significant differences between cases (children who presented with injuries from falls, burns, scalds, ingestions or choking) and controls with regard to having a cooker guard or dangling cords of heating devices.⁹⁹

Pairwise meta-analysis and NMA were not undertaken for interventions promoting kitchen and cooking safety practices as the number of studies reporting each outcome was small.

Study H

Eight reviews were included in the overview evaluating the effect of interventions on other scald-related outcomes.^{36,49,331,332,338,340,385,387} The reviews drew no conclusions specific to the effectiveness of interventions for other scald prevention practices and no meta-analyses were identified for these outcomes. Four primary studies identified from the reviews reported other scald-related outcomes,^{260,295,303,407} with no further studies identified from subsequent searches for primary studies. Of these four studies, two observed significant effects on burn safety scores (representing safer burn prevention practices) of interventions involving home safety education, home safety assessments and free home safety equipment.^295,303 In another study, significantly more families in the intervention group made their home safer after a television campaign, home safety advice, a home safety assessment check and advice on welfare benefits available to purchase safety equipment and the local availability of equipment.²⁶⁰ A multifaceted campaign aimed at reducing the occurrence of scalds in children aged 0–4 years reported no significant effect of the intervention on scald prevention behaviours.⁴⁰⁷

Pairwise meta-analysis and NMA were not undertaken for interventions promoting other scald prevention practices as the number of studies reporting each outcome was small.

Study H

The overview included nine systematic reviews reporting interventions to prevent falls or fall-related injuries,^{33,36,41,331,333–335,383,445} drawing on data from only three primary studies.^267,416,438 Two reviews drew conclusions about the effectiveness of interventions to prevent falls or fall-related injuries, with both concluding that there was a paucity of evidence in this area.^33,333 The three primary studies included in the reviews consisted of one non-RCT,⁴³⁸ one CBA study²⁶⁷ and one cohort study.⁴¹⁶ The cohort study found fewer self-reported fall-related injuries among those receiving home safety information, but the statistical significance was not reported.⁴¹⁶ The CBA study found some evidence of a reduction in fall-related injuries (OR 0.78, 95% CI 0.61 to 1.00) from a community-based injury prevention programme.²⁶⁷ The non-RCT found no significant effect on falls or fall-related injuries of nurse counselling to reduce baby walker use.⁴³⁸ It was not possible to combine effect sizes from these last two studies in PMA or NMA as the numerators and denominators were not published and individual participant data, from which they could be calculated, was provided by only one study.⁴³⁸

Pairwise meta-analyses and NMAs were not undertaken for interventions to reduce falls or fall-related injuries, as the number of studies reporting these outcomes was small.

Study H

The overview included one systematic review and meta-analysis³³ and five reviews^{42,332,338,383,387} reporting the effect of interventions on the possession and use of safety gates. The meta-analysis, which combined effect sizes from 10 studies, reported that home safety education interventions, some of which included the provision of safety gates, were effective in increasing safety gate possession and use (OR 1.26, 95% CI 1.05 to 1.51). There was some evidence that interventions that provided safety gates may have slightly larger effect sizes (OR 1.37, 95% CI 1.15 to 1.62) than those that did not (OR 1.24, 95% CI 0.94 to 1.64).³³ Of the six reviews, that by Lyons et al.³³² and a subsequent update that utilised different eligibility criteria⁴² found no increase in the possession or use of safety gates. The other reviews reported conflicting or unclear evidence on the effect of interventions to increase safety gate possession and use.^{33,338,383,387} Three of these reviews each found only single studies reporting a significant effect of the intervention.^33,338,387 The other review³⁸³ reported that intervention families obtained and fitted safety gates but did not report the significance of this finding.

A total of 16 primary studies reporting safety gate possession and use were identified (14 from reviews^{71,72,257,260,265,275,277,282,287,295,304,413,417,441} and two from additional searches for primary studies^288,431). Three studies reported a significant increase in the possession and use of safety gates on stairs,^72,413,431 two of which provided home safety education and fitted free or low-cost safety gates,^72,431 with the third providing only home safety education.⁴¹³ The remaining 13 studies reported no significant effect of the interventions, only three of which provided low-cost safety gates^71,275 or discount vouchers,²⁷⁷ with none fitting safety gates.

Study I

Pairwise meta-analysis of 12 studies evaluating home safety education, which in some studies also included the provision of safety gates and home safety inspections,^{71,72,265,275,277,282,287,288,295,304,417,431} found that interventions significantly increased the possession of a fitted safety gate (OR 1.61, 95% CI 1.19 to 2.17) (Figure 38). Interventions providing safety gates appeared to be more effective (OR 2.05, 95% CI 1.08 to 3.89) than those that did not (OR 1.26, 95% CI 0.96 to 1.64).

FIGURE 38

Forest plot of effect sizes for having a fitted safety gate from studies evaluating home safety educational interventions (some of which included the provision of safety gates). M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, (more...)

Study J

Network meta-analysis was used to determine the effect of component elements of home safety interventions explored in the 12 studies included in the PMAs for safety gate possession and use.^{71,72,265,275,277,282,287,288,295,304,417,431} These studies included seven interventions and the NMA estimated the 21 possible pairwise comparisons between the seven interventions reported in the 12 studies. The data used in the NMA from each study for each of the falls prevention outcomes are shown in Table 93. The pooled estimates, 95% CrIs and, when available, direct within-trial estimates are reported in Table 94. The most intensive intervention (home safety education plus equipment provision plus fitting of safety equipment plus home safety inspection) was the most likely to be the most effective (p best = 0.97), with an OR compared with usual care of 7.80 (95% CrI 3.18 to 21.30). This combination of interventions resulted in significantly more households having fitted safety gates than any of the other combinations of interventions, with effect sizes being between five and eight times greater with the most intensive intervention.

TABLE 93

Summary of studies and their data included in the NMA of interventions to prevent falls injuries in children aged < 5 years

TABLE 94

Pooled ORs (95% CrIs) from NMA comparing the effect of different interventions on having a fitted safety gate

Study K

This section describes the findings from the decision analysis evaluating the cost-effectiveness of different interventions to increase possession of fitted safety gates to prevent stairway falls. In the base-case analysis, seven interventions were evaluated (Table 95), of which four were found to have higher costs or a higher ICER than more effective interventions (namely education plus free or low-cost safety equipment, education plus free or low-cost safety equipment plus home safety inspection, education plus free or low-cost safety equipment plus fitting of the equipment and education plus home safety inspection). Of the remaining three interventions, education had the lowest estimated ICER compared with usual care, at £284,068 per QALY gained. Figure 39 shows the probability of the alternative interventions being cost-effective. At a threshold value of £30,000 per QALY gained, usual care had the highest probability of being cost-effective (0.999).

TABLE 95

Base case cost-effectiveness results for safety gates to prevent stairway falls

FIGURE 39

Cost-effectiveness acceptability curves for the base-case analysis indicating the probability that each intervention is the most cost-effective for a range of willingness–to-pay ratios. E, education; F, fitting; FE, low-cost/free equipment; HSI, (more...)

Sensitivity analysis

A range of sensitivity analyses varying the base-case assumptions and inputs was carried out (Table 96). All assessed the probability of interventions being cost-effective at thresholds of £30,000 and £50,000. The sensitivity analyses carried out involved reducing the number of safety gates fitted to one; reducing the cost of education by using the cost of providing a leaflet only; providing low-cost (£5) rather than free safety gates; halving the fixed cost of an intervention programme; changing the number of children per household from 1 to 1.8;³⁷⁸ reducing the probability of possessing a fitted safety gate from 0.56 to 0.45 to reflect the baseline possession of a fitted and used stair gate by families in deprived areas of Nottingham;⁷² and changing the utility deficits from 0.05, 0.1 and 0.2 for mild, moderate and severe injuries, respectively, to 0.07, 0.19 and 0.34.¹⁴³

TABLE 96

Sensitivity analysis results for intervention promoting the use of safety gates to prevent stairway falls

The results were not particularly sensitive to any of the changes.

Study H

The overview included one systematic review and meta-analysis³³ and three reviews^332,338,387 reporting the effect of interventions on the possession and use of non-slip bathroom items. The meta-analysis, which combined effect sizes from three studies, reported that home safety education interventions, some of which included the provision of bath mats or decals, were not effective in increasing the possession of non-slip bathroom items (OR 1.16, 95% CI 0.51 to 2.63).³³ Of the three reviews, one³³² found a significant increase in the possession of non-slip bath mats and two^338,387 reported no significant increase in possession of non-slip bathroom items.

Five primary studies reported the effect of interventions to promote the possession and use of non-slip bathroom items (four from reviews^{293,295,304,441} and one from additional searches for primary studies⁴³¹). One study reported a significant increase in the use of non-slip bath decals following home safety education and the provision of a home safety kit that included bath decals.²⁹⁵ The other four studies reported no significant effect of the interventions. All included home safety education, with two including a home safety assessment^431,441 and one providing non-slip bath mats.³⁰⁴

Study I

Pairwise meta-analysis of four studies evaluating home safety education,^{293,295,304,431} which in some studies included the provision of non-slip bathroom items and home safety inspections, found that interventions were not effective in increasing the possession of non-slip bathroom items (OR 1.10, 95% CI 0.68 to 1.79) (Figure 40).

FIGURE 40

Forest plot of effect sizes for possession of non-slip bathroom items from studies evaluating home safety educational interventions (some of which included provision of non-slip bathroom items). M–H, Mantel-Haenszel. Originally published in Kendrick (more...)

Study J

Network meta-analysis was not possible for possession of non-slip bathroom items as there were only two unconnected networks of three interventions.

Study H

The overview included one systematic review and meta-analysis³³ and five reviews^{42,332,338,383,387} reporting the effect of interventions on the possession and use of window safety devices (locks, restrictors or screens). The meta-analysis, which combined effect sizes from five studies, found that home safety education interventions, some of which included the provision of window locks or restrictors, were not effective in increasing the possession of window safety devices (OR 1.16, 95% CI 0.84 to 1.59).³³ Of the reviews, two reported significant improvements in the use, frequency of use and location of window locks,^42,338 and one reported a significant increase in the use of window locks.³⁸⁷ Two reviews reported conflicting findings^33,332 and one review reported findings in which the significance level was not reported.³⁸³

Ten primary studies, nine identified from reviews^{71,72,260,269,275,277,292,304,444} and one identified from additional searches for primary studies,⁴³¹ reported the effect of interventions to promote the possession and use of window safety devices. Two studies reported a significant effect, one on fitted window locks⁷² and one on childproofed window frames.⁴⁴⁴ One of these studies provided home safety education and supplied and fitted window locks for free to low-income families, with free delivery of low-cost window locks to other families.⁷² The second study provided only home safety education.⁴⁴⁴ Eight studies did not report a significant increase in the possession and use of window safety devices. All included home safety education and five included the identification or provision of free or low-cost window safety devices.

Study I

Pairwise meta-analysis of six studies evaluating home safety education,^{71,72,269,275,277,431} which in some studies included the provision of window safety devices and home safety inspections, found that interventions were not effective in increasing the possession of window safety devices (OR 1.10, 95% CI 0.68 to 1.79) (Figure 41).

FIGURE 41

Forest plot of effect sizes for possession of window safety devices from studies evaluating home safety educational interventions (some of which included the provision of window safety devices). M–H, Mantel-Haenszel. Originally published in Kendrick (more...)

Study J

Network meta-analysis explored the effects of component elements of the interventions among the six studies included in the PMA for possession of window safety devices.^{71,72,269,275,277,431} These included six interventions as listed in Table 97 (see also Table 93) The NMA estimated the 15 possible pairwise comparisons between these interventions. Education plus home safety inspection was most likely to be effective (p best = 0.26), but there was very little difference between any of the interventions.

TABLE 97

Pooled ORs (95% CrIs) from NMA comparing the effect of different interventions on the possession of window safety devices

Study K

Decision analyses were not undertaken for interventions to promote the possession of window locks as none of the interventions was significantly better than any other in the NMA.

Study H

The overview included five systematic reviews^{33,332,338,383,387} reporting the effect of interventions promoting the possession and use of furniture corner covers. No meta-analyses were found for this outcome. Two reviews reported interventions associated with significant reductions in sharp-edged bench tops,^332,383 two reviews reported conflicting evidence of the effect of interventions on furniture corner cover use^33,387 and one review included a study reported elsewhere as significantly increasing the use of furniture corner covers although the review did not specifically report this finding.³³⁸

Four studies reported the effect of interventions to promote the possession and use of furniture corner covers.^{257,273,292,304} Two studies found a significant effect of the intervention on use of protective devices on table corners or bench tops.^292,304 One of these studies provided the item to intervention families as part of a free home safety kit³⁰⁴ and the other study provided assistance with identifying retail outlets selling the item in addition to a home safety assessment and home safety education.²⁹² Two studies found no significant effect of the intervention, with one study delivering home safety education, which included the provision of furniture corner covers,²⁵⁷ and the other study providing only home safety education.²⁷³

It was not possible to undertake a PMA or NMA for possession and use of furniture corner covers as only two studies reported numerators and denominators in all arms of the study.^273,304

Study H

The overview included three systematic narrative reviews^33,332,338 reporting the effect of interventions on the possession and use of high-chair harnesses. No meta-analyses were found for this outcome. All three reviews found no evidence that interventions were effective in increasing the use of high-chair harnesses.

Two studies reported interventions to promote the use of high-chair harnesses, neither of which found evidence that interventions were effective at increasing use of this item. The interventions involved home safety education, a home safety assessment and advice on where to purchase equipment,²⁹² and home safety education plus a free home safety kit that did not include a high-chair harness.³⁰⁴ It was not possible to undertake PMA or NMA for this outcome because of the small number of studies included.

Study H

The overview included one meta-analysis³³ and three reviews^332,338,387 reporting the effect of interventions on baby walker possession or use. The meta-analysis, which combined effect sizes from six studies, found some evidence that interventions were effective in reducing baby walker possession or use (OR for having or using a walker 0.66, 95% CI 0.43 to 1.00). One review³³⁸ reported conflicting evidence on the effect of interventions to prevent baby walker use and two reviews found no evidence of a reduction in baby walker use.^332,387

Nine studies reported interventions to reduce baby walker use.^{257,277,287,288,295,304,417,431,438} Two studies, focusing solely on preventing baby walker-associated injuries, reported a significant reduction in ownership and use of baby walkers.^417,438 The first study provided education to reduce walker use from nurses during child visits for immunisation⁴³⁸ and the second provided an educational package delivered by health visitors and midwives before and after childbirth.⁴¹⁷ The remaining studies all provided education about baby walkers as part of interventions aimed at preventing a range of childhood injuries.

Study I

Pairwise meta-analysis of the nine studies of home safety education for the reduction of baby walker use^{257,277,287,288,295,304,417,431,438} found that interventions were effective in increasing the proportion of families who did not have or use a walker (OR 1.57, 95% CI 1.18 to 2.09) (Figure 42).

FIGURE 42

Forest plot of effect sizes for not having or using a baby walker from studies evaluating home safety educational interventions. M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana FA, Cooper NJ, (more...)

Study J

Network meta-analysis explored the effects of component elements of the interventions from the nine studies included in the PMA for baby walker possession or use.^{257,277,287,288,295,304,417,431,438} The data for these studies are provided in Table 93. The studies included seven interventions, which are listed in Table 98. The NMA estimated the 21 possible pairwise comparisons between the seven interventions. The education-only intervention was the most effective (p best = 0.65), with families in the education-only intervention group being less likely to possess or use a baby walker than those in the usual-care group (OR 0.48, 95% CrI 0.31 to 0.84).

TABLE 98

Pooled ORs (95% CrIs) from NMA comparing the effect of different interventions on baby walker possession or use

Study K

Baby walker use may impact differentially on different mechanisms of falls.⁴⁴⁷ For example, using baby walkers has been associated with an increased risk of head injuries from stair falls and their use may also increase the risk of falls on one level from tipping over. However, baby walkers may reduce the risk of falls from furniture if they prevent infants from reaching furniture to climb on or reduce the risk of falls on one level if infants spend less time walking and more time in the baby walker. Using baby walkers has been identified as a risk factor for poisonings and burns as they can allow infants to reach hazards that they would not be able to reach if they not using a baby walker. Furthermore, since these risks were described, some countries, for example Canada, have banned the sale of baby walkers⁴⁴⁷ and design changes introduced by the 2005 European standard⁴⁴⁸ may have altered the risk of walker-associated injuries. More complex decision analyses are therefore required to evaluate the cost-effectiveness of interventions to reduce baby walker use and this is included in the recommendations for research resulting from the KCS programme.

Study H

The overview included two systematic narrative reviews^33,387 reporting the effect of interventions on preventing children being left unattended on high surfaces. No meta-analyses were found for this outcome. Neither review reported a significant effect of interventions.

Three primary studies included in the systematic reviews reported this outcome. There was no evidence of the effectiveness of interventions to prevent children being left unattended on a high surface. The studies delivered interventions involving home safety education and a free safety kit^257,295 and home safety education alone.²⁸⁸

Study I

Pairwise meta-analysis of three studies of home safety education for preventing children being left unattended on high surfaces^257,288,295 found no evidence that education was effective (OR for does not leave child alone on high surfaces 0.84, 95% CI 0.58 to 1.20) (Figure 43).

FIGURE 43

Forest plot of effect sizes for preventing children being left unattended on high surfaces from studies evaluating home safety educational interventions. M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas (more...)

Study J

Network meta-analysis estimated the six possible pairwise comparisons between the four interventions listed in Table 99 (data for these studies are shown in Table 93). There was very little difference between the interventions, but education only was the least likely to be effective in preventing children being left on high surfaces (p best = 0.10), with an OR of 0.56 (95% CrI 0.06 to 4.65) compared with education plus low-cost/free equipment and an OR of 0.50 (95% CrI 0.03 to 8.76) compared with education plus low-cost/free equipment plus home safety inspection.

TABLE 99

Pooled ORs (95% CrIs) from NMA comparing the effect of different interventions on preventing children being left unattended on high surfaces

Study K

Decision analyses were not undertaken for interventions to prevent children being left unattended on high surfaces, as none of the interventions was significantly better than any other in the NMA.

Study H

The overview included four systematic narrative reviews^{33,36,332,383} reporting the effect of interventions on stairway safety. No meta-analyses were found for this outcome. The reviews all found no evidence that interventions were effective in promoting safe indoor stairways. One review reported mixed findings regarding the effect of interventions on outdoor stair safety.³³

Six primary studies reported interventions to promote stairway safety practices, five identified from reviews^{269,277,292,293,300} and one identified from additional searches for primary studies.⁴¹⁸ The interventions in these studies included home safety education plus home safety assessments^269,277,292 and an injury prevention curriculum including falls prevention education delivered to children in a school setting.⁴¹⁸ There was no evidence that these interventions were effective in preventing children playing on stairs or in promoting safe indoor stairways, including the presence of handrails, railings or adequate lighting on stairs, a reduction in the number of railings or stairs in disrepair, a reduction in the ease of opening doors to cellars or basements and stairs properly designed in terms of safety features. There was a significant improvement observed in the modification of safety features of outdoor stairs²⁹³ following a multifaceted community intervention and a significant reduction in missing or loose porch railings after home safety education, a home safety assessment and modification of hazards.³⁰⁰ It was not possible to undertake PMA or NMA or decision analyses for these stairway safety outcomes because of the small number of studies.

Study H

The overview included five systematic narrative reviews^{33,36,332,338,383} reporting the effect of interventions on tripping hazards. No meta-analyses were found for this outcome. Four reviews reported the effect of interventions on the fixing of rugs or carpets.^{33,36,338,383} One review found a significant reduction in tripping hazards from rugs or floor coverings.³⁸³ Another review included the same primary study but did not report the relevant outcome.³⁶ Two reviews did not find a significant effect of interventions on the fixing of rugs or carpets.^33,338 Four reviews reported the effect of interventions on other tripping hazards.^{33,36,332,338} Two reviews^36,332 found a reduction in tripping hazards from electrical cords but not tripping hazards from floors in need of repair, although the reviews did not specifically report these outcomes. One review³³ reported conflicting evidence regarding interventions to reduce tripping hazards from floors in disrepair and cables or leads likely to cause falls. One review³³⁸ found no evidence that interventions were effective in reducing other tripping hazards.

Four primary studies, all included in the reviews described above, reported on the effect of interventions on tripping hazards.^{269,275,300,304} One study included home safety education, a home safety assessment and modification of hazards and found a significant reduction in tripping hazards from loose floor coverings.³⁰⁰ Other outcomes related to tripping hazards reported by this and the other studies did not differ significantly between treatment groups, including tripping hazards posed by electrical cables, floors in disrepair and unsafe rugs and carpets. The interventions in the other studies included home safety education and home safety assessment^269,275 and a free safety kit.³⁰⁴ It was not possible to undertake PMA or NMA or decision analyses for tripping hazard outcomes because of the small number of studies.

Study H

The overview included two systematic narrative reviews^33,332 reporting the effect of interventions on other falls prevention practices. No meta-analyses were found for this outcome. Both reviews reported on interventions to reduce the accessibility of roofs,^33,332 with both including evidence demonstrating a significant effect favouring the intervention group. Both reviews also reported on the effect of interventions on balcony safety,^33,332 with both finding no evidence that interventions were effective in promoting balcony safety. One review³³ reported the effect of interventions on safe furniture layout, reporting that significantly more intervention group families rearranged furniture to avoid staggering the layout. One review³³² reported an intervention that demonstrated a significant improvement in lighting in corridors. Both reviews^33,332 reported an intervention to reduce the amount of climbable fencing, with neither review reporting a significant effect.

One primary study demonstrated a significant effect of an intervention to reduce the accessibility of roofs,²⁹² one study reported that significantly more intervention group families rearranged furniture to avoid staggering the layout⁴⁰¹ and one study²⁹³ demonstrated a significant improvement in lighting in corridors. Other studies found no significant effects of interventions with regard to balcony safety or climbable fencing or gates.

It was not possible to undertake PMA or NMA as the number of studies reporting each outcome was small.

Study H

The overview included nine systematic reviews reporting interventions to prevent poisoning-related injuries.^{33,36,334–336,383,387,388,445} The reviews included a total of seven different primary studies reporting this outcome.^{263,267,275,309,434,437} In addition, one study did not report poisoning rates⁷² but did provide individual participant data for inclusion in a meta-analysis³³ combining poisoning rates from three studies. The meta-analysis found a lack of evidence that interventions reduced poisoning rates (rate ratio 1.03, 95% CI 0.78 to 1.36). The seven primary studies reporting poisoning-related injuries included in the reviews consisted of two RCTs,^309,437 two non-RCTs^263,275 and three CBA studies.^267,434,437 The study that did not report poisoning rates but that did provide IPD was a RCT.⁷² One further primary study, a RCT reporting poisoning rates, was identified by systematic review searches.³¹²

Two of the eight primary studies reported significantly lower rates of medically attended or self-reported poisonings in the intervention groups. The studies evaluated child-resistant aspirin containers,⁴³⁴ reporting a reduction in the proportion of medically attended aspirin poisonings in the intervention area, and a school-based educational intervention targeting a range of injuries and involving teachers, parents and pupils.³¹²

The remaining six studies reported no significant effect of interventions on medically attended poisonings. These interventions included the provision of ‘Mr Yuk’ stickers (depicting a green-faced grimacing man with a protruding tongue) for alerting children to poisoning hazards,²⁶³ the provision of safety items [telephone stickers, coupon for syrup of ipecac (emetic agent), cabinet lock, checklist for ‘poison proofing’ the home and pamphlets],³⁰⁹ safety education and safety equipment provision covering a range of injuries^275,437 and community injury prevention programmes.^267,437 The study not reporting poisoning rates but providing IPD did not find a significant effect of education and the provision and fitting of free safety equipment.⁷²

Study I

Meta-analysis of four studies evaluating home safety interventions (education plus in some studies the provision of home safety inspections and safety equipment)^{72,263,275,312} found a lack of evidence that interventions reduced the rate of poisoning (IRR 0.93, 95% CI 0.65 to 1.32) (Figure 45).

FIGURE 45

Forest plot of effect sizes for poisoning injury rates from studies evaluating home safety educational interventions. IV, inverse variance. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana FA, Cooper NJ, Hubbard SJ, Sutton (more...)

It was not possible to undertake NMA for interventions to prevent poisonings because of the small number of studies.

Study H

The overview included one meta-analysis³³ and 10 systematic reviews^{33,40,42,332,338,383,385,387,388,445} reporting the effect of interventions on the safe storage of medicines. The meta-analysis found evidence that education, with or without the provision of safety equipment, was effective in increasing the safe storage of medicines (OR 1.57, 95% CI 1.22 to 2.02).³³

A total of 25 primary studies reporting on interventions promoting the safe storage of medicines were identified, 18 from reviews^{71,72,257,260,273,282,287,292,295,297,300,304,306,396,399,405,427} and seven from additional searches for primary studies.^{99,266,288,303,398,431,432} Of the 25 studies, seven reported that significantly more intervention than control group families stored medicines safely,^{71,72,260,266,292,300,396} all of which evaluated interventions targeting multiple injury mechanisms. The interventions in these studies consisted of GP safety advice with access to low-cost safety equipment (including cupboard locks) for families receiving means-tested state benefits,⁷¹ a home visit with safety checks and tailored safety advice including assistance in obtaining home safety devices,^72,260,292, a home safety inspection, education and modification,³⁰⁰ a standardised safety consultation and the provision and fitting of safety equipment (including cupboard locks) free to low-income families,²⁹² a personalised safety report (including the promotion of the safe storage of poisons) printed at a computer kiosk in an ED²⁶⁶ and prevention lessons delivered by teachers to children in schools with take-home materials and posters displayed for parents.³⁹⁶

The remaining 18 studies, evaluating a range of interventions including safety education, tailored safety education, or safety education plus equipment, demonstrated no significant difference between treatment groups in the safe storage of medicines.

Study I

Meta-analysis of 13 studies evaluating home safety interventions (education plus the provision of home safety assessments and safety equipment in some studies),^{71,72,257,266,273,282,287,288,295,300,303,304,404} which updated the meta-analysis described in study H,³³ found that interventions were effective in increasing the storage of medicines out of reach (OR 1.53, 95% CI 1.27 to 1.84) (Figure 46). Storing medicines out of reach was defined as stored in locked cupboards, drawers or cabinets; stored at or above adult waist level; or stored so that they are inaccessible to a child.

FIGURE 46

Forest plot of effect sizes for storage of medicines out of reach from studies evaluating home safety educational interventions. M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana FA, Cooper NJ, (more...)

Study J

Data from studies included in the NMAs of poisoning outcomes are shown in Table 100. NMA was used to estimate the 21 possible pairwise comparisons between seven interventions to promote the safe storage of medicines from 13 studies (Figure 47). Home safety interventions with education and low-cost or free equipment were the most likely to be effective (p best = 0.39), with an estimated OR compared with usual care of 2.51 (95% CrI 1.01 to 6.00).

TABLE 100

Summary of studies and their data included in the NMA of the interventions to prevent poisonings

FIGURE 47

Network meta-analysis and PMA results for the safe storage of medicines. Heterogeneity: between-study variance = 0.06 (95% CrI 0.000 to 1.087). Key: NMA results in black; PMA results in green. Interventions are displayed in the order that (more...)

Study K

Seven interventions were evaluated, of which three were excluded from further consideration because they had higher costs than more effective interventions (Table 101). Home safety education had the lowest estimated ICER compared with usual care, at £41,330 per QALY gained, followed by education and the provision of free equipment, with an ICER of £90,615 compared with usual care. Figure 48 shows the probability of the alternative interventions being cost-effective. At a threshold value of £30,000 per QALY gained, usual care had the highest probability of being cost-effective (0.83) followed by education (0.17).

TABLE 101

Base-case cost-effectiveness results for interventions to promote the safe storage of medicines

FIGURE 48

Cost-effectiveness acceptability curves for interventions to promote safe storage of medicines. E, education; FE, low-cost/free equipment; pCE, probability cost-effective; UC, usual care.

Sensitivity analysis

A range of sensitivity analyses varying the base-case assumptions and inputs, as outlined in the methods section, was implemented (Table 102). All assessed the probability of interventions being cost-effective at a threshold of £30,000 and £50,000. The results were mainly sensitive to the baseline incidence of unintentional injuries; when this was increased to reflect a higher incidence rate among children aged < 5 years living in the two most disadvantaged quintiles (SA9 and SA10), the ICER for education compared with usual care reduced from £41,330 to < £20,000 per QALY gained.

TABLE 102

Sensitivity analysis results for interventions promoting the safe storage of medicines

Study H

The overview included two systematic reviews and meta-analyses^33,40 and nine narrative reviews^{36,42,332,338,383,385,387,388,445} reporting the effect of interventions on the safe storage of household and other products (defined as the use of safety catches or locks on cupboards/drawers, the use of CRCs and storage out of the reach of children). One meta-analysis reported evidence that education, with or without the provision of safety equipment, was effective in increasing the safe storage of household products (OR 1.63, 95% CI 1.22 to 2.17).³³ The other meta-analysis of similar interventions delivered in a clinical setting reported that intervention families were 1.8 times more likely to store cleaning agents safely.⁴⁰

A total of 31 primary studies reporting the safe storage of household and other products were identified (24 from reviews^{71,72,257,260,263,265,269,273,275,277,282,287,292,295,297,304,306,309,396,399,405,413,427} and seven from additional searches for primary studies^{99,266,288,303,398,431,432}). Six studies reported that significantly more intervention group families stored household and other products safely.^{72,260,269,292,295,309} One of these studies provided safety education plus equipment,³⁰⁹ four provided safety education, equipment and home safety inspections,^{72,260,269,292} and one delivered home safety counselling and safety equipment with specific injury-focused instructions.²⁹⁵ The remaining 25 studies found no significant differences between groups in the safe storage of household and other products, evaluating a range of interventions including general or tailored safety education, home safety equipment and home safety inspections.

Study I

Pairwise meta-analysis of 15 studies,^{71,72,266,269,273,275,277,282,287,288,295,303,304,309,404} which updated the meta-analysis described in study H,³³ found that home safety interventions (education plus the provision of home safety inspections and home safety equipment in some studies) increased the safe storage of cleaning products (defined as for the safe storage of medicines) (OR 1.55, 95% CI 1.22 to 1.96) but there was significant heterogeneity between effect sizes (Figure 49). Interventions providing locks appeared to be more effective (OR 1.87, 95% CI 1.28 to 2.72) than those providing education without locks (OR 1.13, 95% CI 0.92 to 1.40). Interventions delivered at home also appeared to be more effective (OR 2.14, 95% CI 1.06 to 4.32) than those provided in clinical settings (OR 1.29, 95% CI 1.10 to 1.51).

FIGURE 49

Forest plot of effect sizes for storage of household products out of reach from studies evaluating home safety educational interventions. M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana FA, Cooper (more...)

Study J

Network meta-analysis estimated the 21 possible pairwise comparisons between the seven interventions promoting the safe storage of household products other than medicines in the 15 studies listed in Table 100. The most intensive home safety intervention (education plus low-cost/free equipment and fitting plus home safety inspection) was most likely to be effective (p best = 0.37), with an estimated OR compared with usual care of 2.59 (95% CrI 0.59 to 15.16), but no interventions were significantly more effective than usual care (Figure 50).

FIGURE 50

Network meta-analysis and PMA results for the safe storage of household products. Heterogeneity: between-study variance = 0.3 (95% CrI 0.018 to 1.562). Key: NMA results in black; PMA results in green. Interventions are displayed in the (more...)

Study K

Seven interventions were evaluated (Table 103) but all interventions were more costly and less effective than usual care. This is likely to reflect the OR being < 1.0 for the safe storage of household products comparing children with a poisoning to community controls in study A (i.e. OR 0.77, 95% CI 0.59 to 0.99). In sensitivity analyses, all interventions remained more costly and less effective than usual care (results available from the authors on request).

TABLE 103

Base-case cost-effectiveness results for interventions promoting the safe storage of household products

Study H

Interventions promoting the safe storage of poisons (i.e. when the type of poisonous product was not specified) are reported with those promoting the safe storage of medicines and the safe storage of household products.

Study I

Studies reporting the safe storage of unspecified poisons were analysed separately from those reporting the safe storage of medicines or household products in study I. PMA of five studies reporting interventions promoting the safe storage of poisons^{265,266,297,398,431} found some evidence that home safety education (plus the provision of locks and home safety inspections in some studies) was associated with poisons being stored more safely (OR 2.07, 95% CI 0.92 to 4.66) (Figure 51).

FIGURE 51

Forest plot of effect sizes for the storage of poisons out of reach from studies evaluating home safety educational interventions. M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana FA, Cooper NJ, (more...)

Study J

Network meta-analysis estimated the 10 possible pairwise comparisons between five interventions promoting the safe storage of poisons in the nine studies listed in Table 100. The most intensive home safety intervention (education plus low-cost/free equipment and fitting plus home safety inspection) was most likely to be effective (p best = 0.78), with an estimated OR compared with usual care of 11.10 (95% CrI 1.60 to 141.50) (Figure 52).

FIGURE 52

Network meta-analysis and PMA results for interventions promoting the safe storage of poisons. Heterogeneity: between-study variance = 0.24 (95% CrI 0.005 to 2.647). Key: NMA results in black; PMA results in green. Interventions are displayed (more...)

Study K

A decision analysis for the safe storage of poisons was not conducted as the exposures studied in study A were specific types of medicine or household products as opposed to non-specific ‘poisonous substances’. Consequently, data were not available for decision analyses for this outcome.

Study H

Six systematic narrative reviews reporting the use of CRCs were included in the overview.^{33,42,332,383,387,445} No meta-analyses reporting this outcome were found.

Four primary studies reported use of CRCs, two identified from the included reviews^277,300 and two identified from additional searches for primary studies.^99,303 Of the four studies, one reported that significantly more intervention group families stored paraffin in containers with CRCs.³⁰³ This study evaluated the effect of four home visits providing safety education on several injury mechanisms, with home safety inspections and provision of safety devices including childproof locks and CRCs. The remaining three studies evaluated interventions including safety education, safety equipment and home safety inspections, but no significant effects were reported for this outcome.^277,300

It was not possible to undertake PMA or NMA as the number of studies was small.

Study H

The overview included one systematic review and meta-analysis³³ and eight systematic narrative reviews^{36,40,332,338,383,387,388,445} reporting the effect of interventions on the possession and use of syrup of ipecac. The meta-analysis³³ found evidence that education, with or without the provision of safety equipment, was effective in increasing possession of syrup of ipecac (OR 3.34, 95% CI 1.50 to 7.41).

Fifteen primary studies reporting possession or use of syrup of ipecac were identified from the included reviews.^{265,271,273,274,279,282,287,292,293,300,308,309,420,437,441} Searches for additional primary studies identified no further eligible studies reporting this outcome. Eight of the 15 studies reported that significantly more families in the intervention group possessed or used syrup of ipecac.^{271,279,282,292,293,300,308,420} Of the eight studies, two focused specifically on poisoning prevention. One evaluated the provision of counselling about poisoning treatment methods, a leaflet on poison prevention, a PCC number sticker and free syrup of ipecac with instructions delivered to parents during consultations at a children’s hospital emergency clinic.³⁰⁸ The other assessed the impact of a community education programme aimed predominantly at school pupils and involving information on the risks of products, methods of poison prevention and the correct use of a poison centre.⁴²⁰ The other six studies reporting a positive effect evaluated a home visit, safety inspection, educational materials and safety equipment including syrup of ipecac,²⁷¹ safety equipment and safety counselling by a physician,²⁷⁹ home safety checks and a tailored education booklet plus assistance in locating and obtaining home safety devices,²⁹² a tailored safety report generated by an interactive computer kiosk in a well-child clinic, information on safety equipment savings at a child safety centre and a feedback report for the paediatrician to encourage safety counselling,²⁸² and multifaceted community programmes providing home safety inspections and education with the provision of safety items including syrup of ipecac³⁰⁰ or home safety inspections with the discussion of specific home safety issues.²⁹³

The remaining seven studies evaluated interventions involving safety education, tailored safety education, the provision of syrup of ipecac and community programmes providing safety education, but reported no significant difference in the possession or use of syrup of ipecac favouring the intervention group.

Study I

Pairwise meta-analysis of 10 studies evaluating home safety education (including the provision of syrup of ipecac in some studies)^{265,271,273,274,282,287,293,300,308,309} found that interventions were effective in increasing syrup of ipecac possession (OR 3.34, 95% CI 1.50 to 7.44) (Figure 53). Interventions providing syrup of ipecac appeared to be more effective (OR 10.41, 95% CI 2.40 to 45.09) than those not providing it (OR 1.77, 95% CI 1.08 to 2.91). Interventions provided at home appeared to be more effective (OR 5.45, 95% CI 1.22 to 24.32) than those provided in clinical settings (OR 2.02, 95% CI 1.08 to 3.75).

FIGURE 53

Forest plot of effect sizes for the possession of syrup of ipecac from studies evaluating home safety educational interventions. M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana FA, Cooper NJ, (more...)

Network meta-analysis or decision analyses were not carried out for interventions to promote the possession of syrup of ipecac, as the use of syrup of ipecac by lay people has never been recommended in the UK and is no longer recommended for managing poisoning in children in the USA.⁴⁵²

Study H

The overview included one systematic review and meta-analysis³³ and seven systematic narrative reviews^{36,40,42,338,387,388,445} reporting the effect of interventions on the use of PCC stickers and telephone numbers. The meta-analysis³³ found evidence that education, with or without the provision of safety equipment, was effective in increasing the availability of PCC telephone numbers (OR 3.67, 95% CI 1.84 to 7.33).

Thirteen primary studies (11 identified from reviews^{257,263,269,274,287,293,295,308,309,413,441} and two identified from additional searches for primary studies^288,431) reported the effect of interventions promoting the use of PCC stickers and/or telephone numbers. Six of the 13 studies reported a significant effect favouring the intervention group.^{269,274,308,309,413,431} These studies evaluated safety education, the provision of PCC stickers and telephone numbers, home safety inspections and the Healthy Steps child development and behaviour programme, in which one intervention group received the Healthy Steps programme and another intervention group received the programme and antenatal home visits. The study found a significant effect only for the Healthy Steps programme compared with usual care.⁴¹³ The remaining seven studies did not report a significant improvement in the use of PCC stickers and telephone numbers in the intervention groups, having evaluated a range of interventions including education, tailored safety education, the provision of PCC stickers and home safety inspections.

Study I

Pairwise meta-analysis of nine studies evaluating home safety education (including the provision of PCC number stickers in some studies),^{269,274,287,288,293,295,308,309,431} which updated the meta-analysis reported in study H,³³ found that interventions were effective in increasing the proportion of families who had the PCC number accessible (OR 3.30, 95% CI 1.70 to 6.39) (Figure 54). Interventions providing PCC stickers may be more effective (OR 4.44, 95% CI 2.08 to 9.49) than those not providing stickers (OR 2.66, 95% CI 0.93 to 7.67). Interventions delivered at home (OR 5.99, 95% CI 2.08 to 17.26) may be more effective than those delivered in clinical settings (OR 2.10, 95% CI 0.85 to 5.15).

FIGURE 54

Forest plot of effect sizes for having a PCC sticker available from studies evaluating home safety educational interventions. M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana FA, Cooper NJ, Hubbard (more...)

Study J

Network meta-analysis estimated the 21 possible pairwise comparisons between six interventions to promote having a PCC number available in 10 of the studies included in the PMA, as listed in Table 100. Interventions delivering education, low-cost or free equipment and home safety inspection were more likely to be effective in increasing possession of a PCC number (p best = 0.76; OR 38.82, 95% CrI 2.19 to 687.10) (Figure 55).

FIGURE 55

Network meta-analysis and PMA results for interventions to promote having a PCC number available. Heterogeneity: between-study variance = 1.35 (95% CrI 0.328 to 3.709). Key: NMA results in black; PMA results in green. Interventions are (more...)

Study K

Decision analyses were not undertaken for interventions to promote having a PCC number available because the exposures studied in study A did not include having a PCC number available as PCCs are not provided for public use in the UK. Consequently, data were not available for decision analyses for this outcome.

Study H

Twelve reviews^{33,36,42,332,334–336,338,383,387,388,445} reporting other poisoning prevention outcomes were included in the overview.

A total of 13 primary studies reporting other poisoning prevention practices were identified, nine from the reviews^{257,263,267,271,295,304,408,437} and four from additional searches for primary studies.^{289,303,418,432} Of the 13 studies, two evaluated the effect of education, provision of safety equipment and home safety inspections on poisoning hazards scores, with both finding significant effects favouring the intervention group.^289,303 One reported a significant improvement in intervention group poison safety scores after a school safety fair⁴⁰⁸ and one observed significantly safer storage of beauty products and paraffin properly labelled and stored in tightly closed non-glass containers.³⁰³ The remaining studies evaluated a range of interventions including community injury prevention programmes, safety education, tailored safety education and the provision of safety equipment, but reported no significant effects favouring the intervention groups.

Study I

Meta-analysis of three studies reporting storing plants out of reach,²⁵⁷ plants not being accessible²⁹⁵ or not having any toxic plants in the home³⁰⁴ found a lack of evidence that home safety education was effective in promoting the safe storage of plants (OR 1.18, 95% CI 0.40 to 3.48) (Figure 56).

FIGURE 56

Forest plot of effect sizes for storage of plants out of reach from studies evaluating home safety educational interventions. M–H, Mantel-Haenszel. Originally published in Kendrick D, Young B, Mason-Jones AJ, Ilyas N, Achana FA, Cooper NJ, Hubbard (more...)

Study J

Network meta-analysis estimated the 10 possible pairwise comparisons between five interventions to promote the safe storage of poisonous plants in three studies as listed in Table 100. There was no evidence that any one of the interventions was more likely to be effective than the others in promoting safe storage of poisonous plants (Figure 57).

FIGURE 57

Network meta-analysis and PMA results for the safe storage of poisonous plants. Heterogeneity: between-study variance = 1.00 (95% CrI 0.003 to 3.818). Key: NMA results in black; PMA results in green. Interventions are displayed in the (more...)

Study K

Decision analysis was not undertaken for interventions to promote the safe storage of plants, as none of the interventions in the NMA was any more effective than any other.