A search strategy was developed in collaboration with a librarian. The strategy comprised four different parts: (a) injuries or accidents, (b) prevention, (c) economic evaluation, and (d) demarcation. MeSH terms such as accident prevention, wounds and injuries, prevention and control, primary prevention, protective devices, and cost-benefit analysis were used in different combinations in all databases except in Embase; in Embase similar terms were identified and used by utilising the thesaurus.

Searches were conducted in PubMed, Embase, Cochrane, and NHS EED and they were limited to the last ten years, from 1998 until the beginning of August 2008. Additional searches were also conducted to include the not yet MeSH indexed articles. The searches used the MeSH terms' entry terms and were limited to 2008 or the last 180 days. The search strategies used are available - see additional file 1 and 2.

For inclusion, the study should

• be an economic evaluation of an injury prevention intervention (cost-effectiveness, cost-benefit, or cost-utility analysis);

• include some sort of comparison (randomised controlled trial, quasi-experimental, longitudinal cohort, or case-control); a judgement was made if the comparison groups were comparable;

• evaluate an intervention that could be conducted by municipalities; and

• be published in English.

In addition, there is a need for a clarification of which interventions that might be included or excluded on the basis of the "could be conducted by municipalities" criterion, because municipalities are organized differently across the world. Accordingly, interventions including legislative reforms, interventions targeting workers, and interventions including modifications of vehicles have been excluded in this review. On the other hand, interventions targeting nursing home residents have been included. Furthermore, interventions do not need to have been undertaken by municipalities to be included; municipalities should, however, have the possibility to implement the intervention at hand.

A study was excluded if

• the intervention included any form of medication or drug use;

• it was assessed as not being relevant to the general context; and

• it had an unacceptable quality, appraised by using a checklist previously used by The Swedish Council on Technology Assessment in Health Care 7.

Review articles were excluded, although they were later revised to see if they would add any valuable information.

Sifting was systematically conducted in three different steps and was carried out by one assessor. In the first sift titles and abstracts were screened; in the second sift papers were obtained and screened if they met the inclusion and exclusion criteria; in the third sift papers were critically appraised by using a checklist 7. Only papers with at least an acceptable quality were included.

In the last sift, articles were critically appraised using a modified checklist for evaluating health economic studies 7. The checklist is in turn based on Drummond's checklist 8. The assessment of an article's quality cannot be given with mathematical precision, therefore The Swedish Council on Technology Assessment in Health Care propose three categories to express the quality of a study: High, acceptable, or not acceptable quality, which corresponds to more than 80, 50-80 or less than 50 percent positive answers, respectively, of the applicable checklist questions 7. If an article had previously been reviewed, and was available on the Centre for Reviews and Dissemination's database, that review was used to validate the quality assessment. When there were ambiguities, a revision was conducted.

Findings from the included studies were extracted into pre-developed evidence tables. To facilitate comparisons all cost estimates were converted to US dollars in price year 2007 by first using GDP deflators and then Purchasing Power Parities (PPPs) as suggested by the Campbell & Cochrane Economics Methods Group 9. Apt data were retrieved from IMF and OECD 1011. If the price year was not reported in the article, the year before publication was used when converting cost data. Critical appraisals of the quality of the included studies were also conducted. Furthermore, the interpretation and validity was briefly discussed. In the case of ambiguities a more detailed perusal was carried out.

Due to the research question and to the heterogeneity in economic evaluation methodologies it is neither feasible nor meaningful to conduct a meta-analysis 12.

The systematic literature search yielded 791 potentially relevant papers. Following a sifting process in three steps shown in Figure 1, 20 articles were finally included in the systematic review. Most articles were discarded due to violating the inclusion or exclusion criteria, although many duplicates were also excluded. In the final sift three articles, which did not meet the quality criterion, were excluded.

The included articles comprise different kinds of injury prevention interventions shown in Table 1 by area and by economic result. The table clearly shows that there are few articles in each area except in fall reduction and the use of hip protectors. Furthermore, many studies report a potential of being cost saving and only three studies report no effects.

Studies were also conducted across different settings, compromising eight studies from North America, eight from Europe, and four from Australia and New Zeeland as well as different populations, with the main focus on older people with 15 interventions solely targeting them.

The extracted data from the included studies are delineated and juxtaposed in a table enclosed in Additional file 3. First out are six articles that base their findings on effectiveness data generated within the study, followed by 14 articles using effectiveness data reported elsewhere. The quality and validity of the included studies are appraised and briefly discussed qualitatively in the column furthest to the right in the table. In general, there are great methodological varieties between studies which make comparisons more difficult.

The limited number of studies in each area makes evidence difficult to synthesise in specific areas. However, nine studies examined the use of hip protectors 131415161718192021. Many reported cost saving results although one randomised controlled trial reported no significant effects 21.

Six studies targeted fall reduction in elderly people. One randomised controlled trial of a multidisciplinary program reported no effect 22. Four studies calculated a cost per prevented fall and the sixth study reported a QALY (Quality Adjusted Life Year) gain and a minor saving 2324252627.

Two studies examined smoke alarm give-away interventions in high risk areas 2829. One reported net-savings and one no effects. All other studies investigated interventions not directly comparable and are, hence, only reported in Table 1 and in Additional file 3.

This systematic review identified few economic evaluation studies of injury prevention interventions, of acceptable quality, that might be conducted at the municipal level. Hence, it is shown that there is a paucity of economic evidence - especially in low- and middle-income countries - in this field. Consequently, the evidence based options, including economic information, available for a decision-maker at the municipal level are limited. However, the review reveals that many interventions included in the review are reported to be effective and also cost saving. This shows a potential for societies to obtain good value for societal resources spent.

In general, it seems like interventions targeted at high-risk individuals are more cost-effective. Probably, this is the case if the cost of identification of high-risk individuals does not reduce cost-effectiveness. Furthermore, the results indicate great disparities in methodologies used, which make comparisons more difficult. The outcomes of many interventions are also context dependent. Hence, the results and their transferability should be interpreted with caution.

Before interpreting the studies some caveats should be noted. Different interventions cannot directly be compared because there might be, for instance, contextual, compositional, or methodological differences. Especially, in economic evaluations, with disparities in methods, there are difficulties to synthesise evidence into a coherent whole 30. Consequently, caution should be exercised when interpreting, synthesizing, and generalising results.

Some interpretations could however be made on the basis of the included studies. Many interventions including the use of hip protectors were reported as cost saving. However, the only study that did not import effectiveness data from other sources reported no significant effects 21. This may be due to the low compliance reported and that the study was underpowered. Nevertheless, it is likely that compliance is a key factor for achieving effectiveness; and that cost-effectiveness will be enhanced if compliance could be improved. Moreover, some of the studies suggest that interventions targeting older groups are more cost-effective than interventions targeting younger groups. Some of the studies also suggest that women profit from the use of hip protectors at an earlier age than men do.

Among fall prevention interventions in elderly, most were reported as being effective. In four interventions the cost-effectiveness was reported as a cost per fall prevented. If we do not know the value of a prevented fall this information is of limited value for decision-makers. One study reported that the investigated program was very likely to be cost-effective and encourage similar programs 24. On the other hand, another study found no significant effects of the program, and thus did not recommend it 22. There are also contextual and compositional differences between interventions, which influence outcome and make comparisons and synthesis difficult. However, Salkeld et al. concludes that their program is likely to be more cost-effective among older people with a history of falls 26. Hence, it may be the case that interventions targeted at high risk individuals could give more value for money than interventions targeted at medium and low risk individuals.

The evidence seems conflicting in smoke alarm give-away interventions in high risk areas. It seems that many smoke alarms stopped functioning 2829. Functioning smoke alarms is likely to be a crucial part of successful smoke alarm interventions, and if the life time of smoke alarms could be enhanced this might improve the outcome of similar programs. All remaining studies are too heterogeneous to synthesis, and are thus reported solely in the table in Additional file 3 21313233.

To conclude, analysing all the interventions, it seems that many interventions are targeted towards high-risk individuals and that some interventions show better cost-effectiveness for high-risk groups. It is plausible that interventions targeted towards high-risk individuals have a potential to be more cost-effective than interventions targeted to the general population as long as identifying high-risk individuals is relatively inexpensive.

First, the review is systematic, including a comprehensive search strategy in several major databases. Hence, most of the relevant published studies should have been captured. Second, the quality and validity of the included studies are reviewed, which simplify the interpretations and conclusions of the studies. Third, the review provides a better decision basis by including economic evidence, when deciding about implementing injury prevention interventions. Fourth, the review guides future research by giving an overview of the existing economic evidence in injury prevention.

This study is not without limitations. First, only peer-reviewed literature in English has been considered. Accordingly there may be a risk of publication bias; that is, there is a risk of that negative findings might not have reached publication to the same extend as positive findings 34. Possibly important non peer-reviewed literature could also have been omitted. Second, even though the assessor used established checklists in the sifting and quality assessment processes there are always elements of subjectivity. Also, using only one assessor enhances the risk of error. To minimise these risks a validation of the quality appraisal was conducted with the help from abstracts, written by health economists, in the databases of Centre for Reviews and Dissemination and Euronheed, when available; 15 out of 20 articles were found in these databases and the unanimity was high between the appraisals. Third, this review only included identified studies with an economic perspective - see Anderson 12 for a discussion about the limitations. This fact disregards evidence including only effectiveness data. To obtain better knowledge, it would have been desirable to conduct comprehensive systematic reviews on the health related effectiveness, besides collecting information about the costs, of each intervention of interest. That would however be more resource demanding. Fourth, economic models, which are included in this review, often rely on assumptions and data collected from various sources. This makes the results more prone to uncertainty and therefore the results should be interpreted with caution. Finally, the interventions included in the review do not give a fair picture of the populations and settings where the burden of injury is the greatest; thus the results are biased in the following ways: 1. The interventions are limited to high-income countries although 90% of injuries occur in low- and middle-income countries 35. 2. Falls account for only 8% of all injuries, yet most of the included interventions relate to falls 35. 3. The review is mainly focused on elderly even though almost 50% of the world's injury-related mortality occurs in people aged between 15-44 years 36. 4. Suicide, homicide, drowning, and poisoning are causes that account for approximately 29% of all injury deaths 35; despite this, no interventions targeting these causes have met the inclusion criteria.

To the author's knowledge this is the first systematic review with a health economic perspective in this area. Other reviews, with economic perspectives, have studied more specific topics. For instance, Smith and Fordham have reviewed the cost-effectiveness of "interventions in reducing the risk of new falls, or modifying the harm caused in the event of a fall" among elderly. One of their conclusions is that there are very few economic studies relating to fall prevention programs 4. The scarcity of economic studies is in line with the conclusions in this study. Miller and Levy have made estimates from literature reviewed concerning cost-outcome analyses in injury prevention and control 5. From 84 estimates made in the United States they conclude that injury prevention interventions often can reduce medical costs and save lives. This is also in line with the findings in this study.