Prevention of childhood unintentional injuries in low- and middle-income countries: A systematic review

Injuries are a leading cause of death and disability among children. Numerous injury prevention strategies have been successful in high-income countries, but the majority of unintentional injuries happen to children living in low- and middle-income countries (LMICs). This project aims to delineate the childhood injury prevention initiatives in LMICs. For inclusion, peer-reviewed articles needed to address unintentional injury, include children <18, assess a prevention-related intervention, contain a control group, and be published after 1988. Two pairs of reviewers evaluated articles independently to determine study eligibility. 74 articles were included. 30 studies addressed road traffic injuries, 11 drowning, 8 burns, 3 falls, 8 poisonings, and 21 an unspecified injury type. The findings show positive effects on injury outcome measures following educational interventions, the need for longer follow-up periods after the intervention, the need for effectiveness trials for behavior change, and the need for an increase in injury prevention services in LMICs. This is the first systematic review to summarize the prevention initiatives for all types of childhood unintentional injuries in LMICs. Increased attention and funding are required to go beyond educational initiatives with self-reported measures and little follow-up time to robust interventions that will reduce the global burden of unintentional injuries among children.


Introduction
Five million deaths are attributed to injury globally every year, and 12% of these are among children [1]. Globally, injuries are a leading cause of death and disability among children [2]. Over 900,000 children under the age of 18 die every year due to unintentional injuries [2]. With this review, we identified the current state of childhood injury prevention programs in low and middle-income countries (LMICs), including seemingly effective intervention methods as well as challenges and gaps in current research efforts.

PLOS ONE
PLOS ONE | https://doi.org/10.1371/journal.pone.0243464 December 29, 2020 1 / 46 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 [21]. There is a great need for coordination between and among countries facing this burden to create solutions that are scalable and context appropriate. A systematic review by Vecino-Ortiz et al. assessed effective interventions for unintentional injuries among the world's poorest billion [22]. While this review identifies interventions to reduce mortality, our systematic review included a variety of outcomes and focused exclusively on interventions for children. Rather than assessing the poorest billion from all countries, our review examined interventions from all countries classified as LMIC, allowing for the consideration of geopolitical structures and opportunities for implementing interventions. This systematic review aggregates and summarizes the prevention initiatives for all types of childhood unintentional injuries in LMICs and is inclusive of primary research and additional injury types. With this review we will present what strategies have been proven to be most effective for 5 different injury types and different geographical locations, as well as current gaps in the knowledge of injury prevention strategies.

Protocol and registration
This systematic review is reported in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) Statement (S3 Appendix) and is registered in the PROSPERO database (International Prospective Register of Systematic Reviews) under the number CRD42018091453 [23].

Eligibility criteria
Our main criteria for article inclusion were assessment of a prevention initiative for unintentional injuries in children in LMICs. LMIC status and categories of unintentional injuries (RTIs, drowning, burns, falls, and poisonings) were decided according to World Bank and WHO criteria, respectively [24,25]. As of the 2021 fiscal year, countries with a gross national income (GNI) lower than $1,035 were classified as low-income countries (LICs), and LMICs include countries with a GNI between $1,036 and $4,045 [24]. For inclusion, articles needed to be related to unintentional injury, target or include children under the age of 18, assess a prevention-related intervention, contain a control or comparison group (including pre-post designs), and be peer-reviewed and published after 1988. 1988 was chosen as the cut-off date to capture as many as studies as possible within a reasonable timeframe (30 years). Injuries resulting from self-harm behavior were not included. Articles were excluded if they were abstracts, literature or systematic reviews, meta-analysis, unpublished theses, or commentaries.

Study selection and data collection
We found a total of 3960 articles in our initial search. Two reviewers independently screened the titles and abstracts. Abstracts not providing sufficient information concerning the eligibility criteria were accessed for full-text review. Two pairs of reviewers then evaluated full-text articles independently to determine study eligibility in the original study language. Reference and citation analysis were done on the articles meeting inclusion criteria. Sixty articles were included from this search (Fig 1).
We updated the same search criteria to include studies published after the initial search (between February 2018 and April 2019. This resulted in an additional 344 articles. These articles went through the same inclusion and screening process as the initial search. Nine articles were included from this search. A third search was conducted for studies published between April 2019 and May 2020, resulting in an additional 397 articles to be screened. Following the same inclusion and screening process, five articles were added from this search.

Quality of studies
To assess data quality we used the Cochrane RoB 2 tool (A1 in S1 Appendix). The Cochrane RoB 2 tool [26] assesses risk of bias by asking questions about the study design, aim of the study, randomization process, deviation from intended interventions, missing outcome data, measurement of the outcome, and selection of reported result. The randomized control trials RCTs were also assessed using the Consolidated Standards of Reporting Trials (CONSORT) (A2 in S1 Appendix). However, in order to have a standardized classification of bias for all studies, we classified studies as low, moderate, or high risk of bias as outlined in the Cochrane handbook [27]. No studies were excluded from data extraction based on their assigned quality.

Data extraction
Two pairs of reviewers conducted data extraction on articles independently, and disagreements were resolved through discussion. The data extraction included year of publication, geographic region of author, location of study, objective, study design, setting, intervention type, sample size, participant characteristics and inclusion criteria, data collection and analysis methods, reported outcomes and values, results, and main conclusions. For non-English articles, data extraction was conducted by a bilingual researcher that was a fluent or native speaker in English and the language of the article.

Data analysis
Upon screening the articles for this review, it was concluded that a meta-analytical approach of all of the articles would not be feasible given the high level of variability in study designs. We thus conducted a qualitative metasummary. Thematic analysis was done by aggregating the main outcomes of the articles by categories of injury and type of intervention.

Study characteristics
In total, 74 articles were included in this review (Table 1). Fifty articles were in English, 13 in Chinese, 5 in Portuguese, and 6 in Spanish (Table 1).
Although we only included studies conducted in LMIC, the geographic regions of first authors included Argentina, Bangladesh, Brazil, China, Colombia, Cuba, Denmark, Germany, Indonesia, Iran, Kenya, Mexico, The Netherlands, Pakistan, Peru, Saudi Arabia, South Africa, Sweden, Uganda, the United Kingdom, the United Republic of Tanzania, the United States, and the West Indies. The LMIC countries in which the studies were conducted included Argentina, Bangladesh, Brazil, Cambodia, China, Colombia, Cuba, Grenada, India, Indonesia, Iran, Kenya, Mexico, Pakistan, Peru, Serbia, South Africa, Sri Lanka, Turkey, Uganda, the United Republic of Tanzania, and Zambia.
The study characteristics, including the geographic regions of the interventions, the risk of bias, and the location of the first authors are shown in Fig 2.

Qualitative summary of results
In our review, 30 studies dealt with RTIs, 11 with drowning, 8 with burns, 3 with falls, 7 with poisonings, and 21 did not specify the injury type (Fig 3). Some studies addressed multiple injury types or used more than one intervention category. A widely used framework to reduce accidents and unintentional injuries is the 5 E's-engineering, education, encouragement,

PLOS ONE
Prevention of childhood unintentional injuries in low-and middle-income countries enforcement, and evaluation [1]. We built upon the 5 E framework by classifying interventions as one or more of the following categories Education interventions were classified as skillsbased education (e.g. driving course, swimming lessons) and theory-based education and provision of information (e.g. lectures, videos, pamphlets without practical skill-based component). Engineering interventions. included environmental change (e.g. pedestrian overpass, water barriers) and the provision of safety devices (e.g. helmets, pesticide storage boxes). Enforcement interventions included law enforcement (e.g. roadside sobriety checks) and legislation (e.g. new laws). Finally, Encouragement interventions included community campaigns or awareness programs (e.g. television or radio messages) as well as supervision (e.g. observing adult presence). Many studies assessed more than one intervention type. Seven studies used skills-based educational interventions, 5 changed environmental factors, 3 examined the effectiveness of law enforcement, 11 raised community awareness, 3 observed adult supervision as an intervention, 14 provided safety devices, 3 evaluated legislation, and 65 studies used an educational intervention. The follow-up time after each intervention ranged from 0 to 3 years, with most interventions (n = 42) having a follow-up period of less than 6 months, no follow-up at all, or not specified (Table 1).
A summary of the outcomes of these interventions, stratified by injury type, can be found in Tables 2-6. For RTIs, the studies used interventions in skills-based education, environmental change, law enforcement, community awareness, provision of safety devices, legislation, and education ( Table 2). The majority of these studies found an increase in road safety knowledge and selfreported safety behaviors. All studies that reported injury related outcomes (incidence, risk, severity, mortality) showed a decrease in numbers, except one study from Tanzania [56] that reported a 3% increase in RTI incidence on the intervention site with no change in incidence at the control site.
For drowning, the studies used interventions in skills-based education, environmental change, community awareness, supervision, provision of safety devices, and education ( Table 3). The majority of these studies found an increase in knowledge of the prevention of drowning. Most studies found a reduction in the incidence of non-fatal and fatal drowning [58,67,92]. One study did not find a significant decrease in injury rates (11.1% to 11.0%), but did find a positive increase in knowledge following a community awareness program and education materials [59].
For burns, the studies used interventions in the categories of community awareness, provision of safety devices, legislation, and education (Table 4). A reduction in the incidence of burns and hospitalizations as a result of burns was found following the provision of safety devices [71,76]. While one study did not find statistically significant intervention effects on electrical and paraffin burn-related hazard reduction, they did report a significant change in burn-related safety practices and behaviors [80]. An increase in knowledge of the prevention of burns was reported in the studies using educational interventions [68-70, 75, 76, 78, 79].
In the fall category, the studies used interventions including the provision of safety devices and education, combined with interventions about poisonings and burns. There was no significant decline in hazards or injury risk [76,77,80].

PLOS ONE
Prevention of childhood unintentional injuries in low-and middle-income countries Table 2. Summary of intervention outcomes for injury types: Road traffic injuries.

Intervention Type (Component of the 5 E Framework)
Intervention Description

Injury-Related Outcomes
Other Outcomes

Road Traffic Injuries
Skills-based education (Education) One-week driving course Increase in RTIs perhaps due to changing climate conditions [56] Practical training of safe streetcrossing behaviors Observed improvement of safe street crossing behaviors [55] Environmental change (Engineering) A pedestrian overpass was constructed Fewer fatalities but more pedestrian injuries and traffic crashes [47].
Infrastructure enhancements designed to lower vehicle speeds and separate pedestrians from traffic Reduced incidence of RTI [50] Law enforcement (Enforcement) Additional sobriety checkpoints Increased awareness of road safety measures [41] Confiscation of motorcycle for riders not wearing helmets Decrease in hospital admissions, decrease in ICU admission due to head trauma (significant) increase in helmet use (significant), decreased mortality (not significant) (combined with other interventions) [37] Traffic Act that increased arrests and surveillance No changes in injury severity pre-and post-enforcement [46] Community awareness (Encouragement) Radio messages, banners and posters Decrease in hospital admissions, decrease in ICU admission due to head trauma (significant) increase in helmet use (significant), decreased mortality (not significant) (combined with other interventions) [37] Increase in level of knowledge [40].

Social marketing campaign
Reported awareness of road safety messages [41] Media resources, videos, t-shirts Reduction in injury severity, mainly for traumatic brain injuries [52] Provision of safety devices.
(Engineering) Provision of motorcycle and bicycle helmets Decrease in hospital admissions, decrease in ICU admission due to head trauma (significant) increase in helmet use (significant), decreased mortality (not significant) (combined with other interventions) [37] Increased observed helmet use [33] Provision of child safety seats Increased self-reported use of child safety seats [45] Provision of reflector vests and motorcycle helmets Increase in RTIs perhaps due to changing climate conditions [56] Legislation (Enforcement) Zero blood-alcohol limit and higher penalties for drinking and driving Decrease in overall and pedestrian mortality [28] Law requiring helmets be worn on motorcycles Decrease in hospital admissions, decrease in ICU admission due to head trauma (significant) increase in helmet use (significant), decreased mortality (not significant) (combined with other interventions) [37] Child restraint legislation Reduction in the rate of child injuries [48] (Continued )

Injury-Related Outcomes
Other Outcomes

Education (Education)
Storybooks Increase in knowledge of road safety [29].
Pamphlets and videos for child safety seats Increase in knowledge and purchase of child safety seat [31] Lectures Decrease in incidence of bicycle injuries [42]; decrease in hospital admissions, decrease in ICU admission due to head trauma (significant) increase in helmet use (significant), decreased mortality (not significant) (combined with other interventions) [37] Increase in knowledge of drink driving risk; [32] change in knowledge about brain and spinal cord injuries but no change in attitudes toward prevention; [36] increased knowledge on helmet usage only; [38] no change in reported attitudes about injury prevention; [35] increase in knowledge and safety practices; [54] change in self-reported attitude; [41] decrease in self-reported drink driving; [30] reduced incidence of RTI; [50] change in risk perception [51] In-school training on the importance of helmets and proper fit Increase in observed helmet use [33] Promotional videos Increase in knowledge and self-reported use of booster seats [34] Education materials No change in traffic violations or accidents; [43,44] Change in knowledge; [43] increase in knowledge; [44] education alone did not increase use of child safety seats; [45] observed improvement of safe street crossing behaviors [55] Lectures, posters and guidebooks Decrease in incidence of injury [87] Peer education Increase in knowledge of safety [53] Educational program using the educational therapeutic method Increased level of knowledge and maintained the preventive attitudes and practices of traffic accidents [39] Seminars and videos Decrease in injury [96] Increase in knowledge [96] The outcomes reported may be a summary of more than one intervention type. https://doi.org/10.1371/journal.pone.0243464.t002 For poisonings, the studies used interventions in the categories of community awareness, provision of safety devices, and education (Table 5). Most studies reported an increase in observed safety behaviors [73,76,78], and an increase in self-reported knowledge of prevention methods [72]. Krug et al. reported a decrease in the incidence by 47.4% of poisonings by ingestion after the distribution of child restraint containers for paraffin [74]. Swart (2008), who also provided safety devices and conducted several home visits, did not find a significant intervention effect for poisoning and falls, but for burn safety practices (p-value 0.021 intervention effect -0.41 (-0.76 to -0.07)) [80].
The intervention types used for non-specified injury categories included skills-based education, environmental change, community awareness, supervision, and education (Table 6).

Injury-Related Outcomes Other Outcomes
Drowning Skills-based education (Education) Swimming lessons Reduction in relative risk of drowning death [61] Lifeline throwing skills Increase in skill level [64] Environmental Change (Engineering)

Elimination of water reservoirs
Reduced incidence of fatal drowning [58] Addition of barriers to unsafe water areas Decreased incidence rate of non-fatal drowning [67] Community Awareness (Encouragement) Messages on local TV and radio stations Decreased risk of drowning [58] Radio messages and newsletters No significant decrease in non-fatal drowning [59] Increase in knowledge [59] Media collaboration with health department Decrease in incidence of non-fatal drowning [67] Supervision (Encouragement)

Observed supervision
Higher percentage of children supervised [57] Increased life guard and rescue stations Decrease in risk and probability of drowning [58] Provision of Safety Devices (Engineering)

Provision of door barrier and playpen
Tools accepted by parents in community [57] Education (Education) Training of community health workers Decrease in risk of drowning [58] Education materials No change in non-fatal drowning incidence [59] Increase in knowledge; [59] increase in knowledge; [60] increase in perception of drowning risk [66] Workshops Increase in water safety knowledge; [63] increase in water safety knowledge [64] Video testimonials Improved knowledge and simulated water behaviors [62] Lectures Reduction in mortality rates; [65] decrease in incidence of non-fatal drowning [67] Change in parental knowledge [65] The outcomes reported may be a summary of more than one intervention type. https://doi.org/10.1371/journal.pone.0243464.t003 These studies mainly reported an increase in caregiver knowledge of injury prevention methods. Nine studies reported positive outcomes directly related to injury rates, incidence, mortality, and severity [86,91,92,96,97,99,100,102]. Interestingly, 2 studies conducted in LICs evaluated implemented environmental changes, as recommended by the WHO, and found varying results. While infrastructure enhancements to reduce vehicle speed and create spatial separation from pedestrians and vehicles resulted in a reduction of injuries [50] the construction of a pedestrian overpass resulted in an increase of pedestrian injuries and traffic crashes, albeit reducing fatality rates [47].
Among the 19 RCTs included, 84% (n = 16) used educational interventions, with the remaining studies using a combination of education and other interventions (Fig 4A).

Discussion
This is the first systematic review evaluating all types of childhood unintentional injury prevention initiatives in LMICs published within the past 30 years, building on the 2008 WHO report on Child Injury Prevention. It is also the first to summarize the available evidence in English and non-English studies on LMIC childhood prevention interventions by injury, Provision of kettle strap Device is acceptable and affordable [75] Education (Education) Educational materials Increase in knowledge; [68] increase in knowledge and belief of prevention; [69] reduction in hazards [76] Presentation and coloring book Increase and decrease in knowledge for different concepts [70] Informative film and verbal instruction Decrease in burns [71] Comic books Increase in knowledge [79] Home education No significant decline in injury risk [80] The outcomes reported may be a summary of more than one intervention type. https://doi.org/10.1371/journal.pone.0243464.t004 geographic location, and intervention type. This systematic review confirms that despite having the highest global burden of childhood unintentional injuries, LMICs have a disproportionately limited amount of research in this area compared to HICs [2,18]. The findings reveal that 1) there is an unequal distribution of research regarding each injury type and a lack of injury-specific research, 2) there is an uneven geographical coverage, and 3) the general quality of the included studies was low, often due to the study design and the failure to be sensitive to or relevant for local cultures.

Unequal distribution of injury types
The number of studies per injury type varied greatly, revealing an unequal distribution of research regarding each injury type. The highest burden of RTI in children is found in Africa, with 19.9 deaths per 100,000 people, followed by LMICs in the Eastern Mediterranean with 17.4 deaths per 100,000 people [105]. The majority of the studies focused on RTI, which may be warranted given their high burden among children in LMICs. However, we found that less than 15% of the included studies took place in SSA, representing only 3 out of the 48 SSA countries [46,47,50,56]. Additionally, only one study was conducted in the Mediterranean [55], which demonstrates the need to focus future intervention programs on countries with the highest burden of RTI. The number of studies conducted for the other injury types were generally representative of the disease burden in the geographical locations. There is very little research on falls among children in LMICs, with the majority of this research targeted towards the elderly population in HIC [13]. Compared to other injury types, there is a low burden of fall injury among children in LMICs and this was consistent with the small number of studies focusing on falls included in this review. The included drowning prevention interventions were mostly conducted in Asian countries, where it is most prevalent, surpassing the highest reported injury rates for RTI (19.9 per 100,000 in Africa), with 30 per 100,000 in China, Philippines, Bangladesh, Vietnam, and Thailand. While the above findings show studies that mostly Training community workers to deliver materials Increase in self-reported knowledge and observed safety practices [78] Provision of Safety Devices (Engineering) Child-proof, lockable storage containers Decrease in incidence of paraffin ingestion [74] Increase in pesticides kept safe from children; [73] reduction in poison hazards; [76] no significant decrease in hazards [80] Education (Engineering) Seminars Increase in knowledge [72] Printed materials Reduction in paraffin ingestion [74] Reduction in hazards [76] Counselling Observed change in risk factors and knowledge and attitude scores of injury prevention methods [77] Home education No significant decline in injury risk [80] The outcomes reported may be a summary of more than one intervention type. https://doi.org/10.1371/journal.pone.0243464.t005 corresponded with the associated disease burden, 33% studies focused on "all" or "non-specific" injury types. The WHO recommends research specific actions to decrease the burden of childhood injuries, namely to "enhance the quality and quantity of data for child injury Table 6. Summary of intervention outcomes for injury types: Non-specific/all injuries.

Intervention Type (Component of the 5 E Framework)
Intervention Description

Skills-based education (Education)
Emergency response training Incidence rate and severity of injury decreased [97] First aid training Reduction in injury rate among boys [101] Increase in caregiver knowledge [101] Environmental change (Engineering)

Upgrade school environment
Reduction in incidence of injury [87] Community Awareness (Encouragement)

Change in community regulations
Decrease in incidence of injuries [97].

Supervision (Encouragement)
Evaluation of supervision Unintentional injury mortality 3x higher when unsupervised-mainly for drowning [86] Education (Education) Posters, letters, lectures, videos Increased knowledge [103] Training of health workers Increase in knowledge [84] Lectures, posters, guidebooks Decrease in injury incidence [87] Counselling Increase in parental knowledge and safety behaviors [88] Lectures Decrease of injury rate; [99] decrease in injury incidence rates [96] Increase in knowledge, decrease in risks in household; [83] increase in knowledge; [89] increase in knowledge; [99] increase in knowledge [94] Communication from doctor Decrease in accidents [91] Decrease in risk factors for injury [91] Education, conversation circles, dialogical relationship framework Increased knowledge of risk factors [81] Brochures Increase in knowledge [104] App-based parenting education No change in unintentional injury incidence among preschoolers [90] Significantly improved caregivers' safety behaviors [90] Guidebooks, peer education Decrease in incidence rate and severity of injury; [97] decrease in injury rate (including RTI-specific) [98] Increase in knowledge [98] Games and posters Decrease in the rate of injury (including RTI specific) [98] Increase in injury prevention knowledge [93,98] Education of parents, teachers, and children Decrease in injury rate among boys [101] Increase in knowledge among parents; [101] increase in mothers' awareness of prevention methods for pediatric injuries [85] The outcomes reported may be a summary of more than one intervention type https://doi.org/10.1371/journal.pone.0243464.t006 prevention" and "define priorities for research, and support research on the causes, consequences, costs and prevention of child injuries" [19].

Uneven geographical coverage
The geographical distribution of the studies also reveals uneven coverage as seen in Fig 2. In terms of authorship, 16 of the 74 included articles had first authors from HIC, and 51 out of 74 had first authors from upper middle-income countries (UMICs), leaving only 7 articles with first authors from low andlower middle income countries. Therefore, conducting interventions in LMICs does not imply that local researchers were consulted or included in research design and application. 58 of the 74 (78%) included studies were conducted in 13 UMICs). However, 34 of the 58 (59%) were conducted in either Brazil or China. Latin America does not represent the highest burden of injury compared to Africa and Asia, yet is represented by 30% of the included studies, with more than half of the studies conducted in Brazil. While 22% of all UMICs are included in this review, 13% of all LMICs and only 6% of low-income countries (LICs) were the setting for childhood injury prevention program evaluations. LICs were not only the most underrepresented group, but also the least diverse group in terms of injury types and geographical locations, including only RTI interventions in SSA. While LICs face the highest injury burden [19], the limited number of interventions conducted could be due to a lack of available funding to invest in expensive prevention strategies that have been successful in HICs. However, in many LICs, the roads are shared by pedestrians, animals, cars, buses, motorcyclists, and bicyclists.

Lack of high-quality study designs and relevance to local culture
Lastly, the quality of the included studies varied greatly, and our findings support the need for implementing high-quality interventions that are culturally sensitive, relevant, and welcomed by the local culture [106]. The insufficient funding for research on injury prevention in LMICs has resulted in few randomized evaluations and few studies with control groups or significant follow-up periods [107]. In this review, only 19 RCTs were included, and 47% (n = 35) of the 74 studies had a moderate or high risk of bias. In many studies, a self-reported survey tool was used to determine injury rates. Among the RCTs included, the majority used educational interventions and addressed all or non-specific injury categories, leading to a lack of clearly defined, injury-related outcomes. Importantly, 6 of the 9 RCTS since 2008 have been conducted in China almost exclusively investigating the impact of education on all or non-specific injuries within the school setting. We only included studies that included a comparison group to assess the impact of the prevention intervention. Rothman et al. recently conducted a review on the study designs used to assess the quality of child injury prevention interventions that were published between 2013 and 2016 [108]. Their findings also suggest a lack of high quality, hypothesis-driven study designs. The evaluation of all of the studies in this review mainly showed a positive or desired self-reported change following educational interventions, a need for longer follow-up, the need for effectiveness trials to access behavior change, and the need for an increase in injury prevention services in low-and middle-income settings. While those recommendations are critical and in line with what other organizations have called for, they do not provide a strong indication on which of those recommendations will be most important to focus on going forward. Beyond improved study designs, higher quality interventions will be sensitive to the needs of specific populations, environments, and available resources [106]. A validation study by Kohrt et al. also points out the risk of doing more harm than good by providing interventions that have not been validated in comparable settings [109]. For example, although overpasses have been effective in other settings, one study that implemented an overpass for a busy street saw an increase in RTI, due to individuals' perceptions that the overpass increased walking distance and its low visibility created a new space for crime [47].

Limitations
This review has certain limitations. Due to our inclusion of various methodologies, conducting a quantitative meta-analysis was not possible. There were 19 RCTs, but they reported non-similar outcomes for different injuries, and had varying study follow up times, making them not suitable for a meta-analysis. Additionally, including only a single type of study design would not have produced a thorough representation of the present literature on the prevention of unintentional injuries of children in LMICs. Our inclusion criteria included a control group and peer-review, which adds to the cost of the studies, excludes unpublished theses and nonindexed journals that are common in LMICs, and does not account for researchers falling victim to predatory journals that do not lead to publication. Grey literature was not included in our analysis, which could lead to possible exclusion of presented intervention programs, but we ensured that only data of sufficient rigor was included in this review by making peerreviewed publication part of our inclusion criteria. In order to categorize the injury types, we have followed the major categories of injury presented by WHO. While this may have resulted in the exclusion of some other types of injury, such as animal bites or sunburns, those injury types, while important to address in specific hotspots and high-risk populations, do not account for a significant burden of injuries globally. Additionally, some of the reported injuries could have been a result of violence against children and falsely identified as unintentional, which could impact the effectiveness of the investigated prevention methods for unintentional injuries. We have included countries based on their current income status rather than at the time of the publication of the study to identify the needs and intervention possibilities for current LMICs. While this likely does not exclude any countries in the low-income category, we may have excluded a few countries that moved from UMIC to HIC over time.

Conclusion
Childhood unintentional injury contributes to a significant amount of global mortality and morbidity. Children living in LMICs are especially vulnerable to injury. There have been numerous effective and low-cost solutions for injuries, but there is a lack of dispersal of these initiatives into the most affected settings. This requires significant political will and increased funding to go beyond educational initiatives with self-reported measures and little follow-up time, to robust research and interventions that will reduce the global burden of unintentional injuries among children. Low or non-existing funding is a significant and ongoing barrier for researchers from LMIC. International donors providing research funding should focus on LMIC, and particularly LIC, and provide assistance in writing grant proposals to allow for more rigorous study designs, thereby improving the quality of research relevant to LMIC. To significantly reduce the rates of injury and the associated negative health and social outcomes, future studies should focus on high-quality trials to assess targeted intervention strategies for areas with a high injury burden that are specifically tailored for the needs of specific cultural and geographical settings.
Supporting information S1 Appendix. Data quality assessments.