Risk factors for injuries in young children in four developing countries: the Young Lives Study


Corresponding author Laura Howe, Department of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, 49/51 Bedford Square, London WC1B 3DP, UK. Tel.: 0044(0)207 958 8177; Fax: 0044(0)207 299 4632; E-mail: laura.howe@lshtm.ac.uk


Objective  To assess the occurrence of child injury in four developing country settings and to explore potential risk factors for injury.

Methods  Injury occurrence was studied in cohorts of 2000 children of age 6–17 months at enrolment, in each of Ethiopia, Peru, Vietnam and India (Andhra Pradesh). Generalized estimating equation models were used to explore potential risk factors for child injury.

Results  Occurrence of child injury was high in all countries. Caregiver depression emerged as a consistent risk factor for all types of injury measured (burns, serious falls, broken bones and near-fatal injury) across all countries. Other risk factors also showed consistent associations, including long-term child health problems, region of residence and the regular care of the child by a non-household member.

Conclusions  This report provides further evidence of the importance of childhood injury in developing countries and emphasizes the importance of including infants in injury research and prevention strategies. It provides strong evidence of an association between caregiver mental health and child injury risk and contributes to the limited knowledge base on risk factors for child injury in developing countries.


Objectif  Etudier la survenue de blessure infantile dans quatre pays en voie de développement et déterminer des facteurs de risques potentiels pour les blessures.

Méthodes  La survenue de blessures était étudiée dans une cohorte de 2000 enfants âgés de 6 à 17 mois à l'inclusion en Ethiopie, Pérou, Vietnam et Inde (état de Andhra Pradesh). Des modèles d’équation d'estimation généralisée ont été utilisés pour étudier les facteurs de risques potentiels de blessure chez l'enfant.

Résultats  La survenue de blessure infantile était élevée dans tous les pays. La dépression de la personne chargée d’élever l'enfant ressortait comme facteur de risque important pour tous les types de blessures étudiées (brûlures, chutes, fractures, blessures mortelles) dans tous les pays. D'autres facteurs de risques étaient associés comme les problèmes de santéà long terme, la région de résidence, et le fait que l'enfant soit élevé par une personne autre qu'un membre de la famille.

Conclusions  Cette étude fournit des preuves de l'importance des blessures infantiles dans les pays en voies de développement et souligne l'importance d'inclure les nourrissons dans les stratégies de prévention et dans les études de recherche sur les blessures. Elle montre la forte association entre la santé mentale de la personne chargée d’élever l'enfant et le risque de blessure infantile et montre les connaissances limitées des facteurs de risque de blessures infantile dans les pays en voie de développement.


Objetivo  Establecer la incidencia de lesiones infantiles en cuatro escenarios en países en vías de desarrollo, y explorar los potenciales factores de riesgo de lesiones.

Métodos  La incidencia de lesiones fue estudiada en una cohorte de 2000 niños entre 6 y 17 meses de edad en el momento de enrolamiento en Etiopía, Perú, Vietnam e India (estado de Andhra Pradesh). Se utilizaron modelos de ecuaciones de estimaciones generalizadas para explorar los potenciales factores de riesgo de lesiones infantiles.

Resultados  La incidencia de lesiones infantiles fue alta en todos los países. La depresión de los cuidadores surgió como un factor de riesgo consistente en todo tipo de lesiones medidas (quemaduras, serias caídas, huesos rotos, y lesiones casi fatales) en todos los países. Otros factores de riesgo que también mostraron asociaciones consistentes incluyeron problemas de salud de largo tiempo, región de residencia, y el regular cuidado del niño por otras personas no miembros de su familia.

Conclusiones  Este estudio proporciona más evidencia de la importancia de las lesiones infantiles en los países en vías de desarrollo, y enfatiza la importancia de incluir a los niños en las investigaciones sobre lesiones y en las estrategias de prevención. Proporciona una fuerte evidencia de una asociación entre la salud mental de los cuidadores y el riesgo de lesiones infantiles, contribuyendo al limitado conocimientos basado en factores de riesgo de las lesiones infantiles en países en vías de desarrollo.


Injury is the leading cause of child death in high-income countries, accounting for approximately 40% of deaths in children aged 1–14 (UNICEF 2001). Despite the rising proportion of child deaths attributable to injuries, child injury death rates have been falling since the 1970s because of a coordinated programme of prevention measures including legislative measures, public education, environmental modification, and improvements in accident and emergency services (UNICEF 2001). The burden of disease attributable to child injury in developing countries, however, is increasing in both absolute and relative terms, as infectious diseases decrease (Meyer 1998; Deen et al. 1999; Guastello 1999; Krug et al. 2000). An estimated 98% of child injury deaths occur in low- and middle-income countries (Murray & Lopez 1996). Furthermore, mortality is often described as the ‘tip of the injury iceberg’, because for every child who dies many more will suffer non-fatal injuries; a proportion of these will be left with varying degrees of disability. Hazardous living conditions; heavy traffic with no separation of vehicles from pedestrians; and lack of safe play spaces, childcare options and health care facilities compound the problem for many children in these settings.

The availability and quality of data on childhood injuries in developing countries are extremely limited. The majority of existing studies are hospital-based, which fail to capture the high numbers of injured children who do not access medical care and generally lack data to explore risk factors. There have been relatively few community-based and population-based studies of child injury in such settings, but they suggest potentially high burdens (Reichenheim & Harpham 1989; Janson et al. 1994; Forjuoh et al. 1995a; Ahmed et al. 1999; Kozik et al. 1999; UNICEF/TASC 2004).

The term ‘accident’ implies a random event beyond control. Thinking of injuries as inevitable may in part be responsible for the lack of global effort in injury prevention. There is a wealth of data; however, that suggests injuries are both predictable and preventable. Research in developed countries has identified many environmental and social risk factors for child injuries. The few studies exploring child injury risk factors in developing countries have identified male sex (Reichenheim & Harpham 1989; Celis et al. 2003; Villalba-Cota et al. 2004), large family size (Janson et al. 1994; Bang et al. 1997; Ahmed et al. 1999; Rahman et al. 2005), low maternal education (Bangdiwala & Anzola-Perez 1990), child's main caregiver not being the mother (Rahman et al. 2005), maternal illiteracy (Rahman et al. 2005), care of the child by older siblings (Janson et al. 1994), ‘pre-existing impairments’ in the child (Forjuoh et al. 1995b) and maternal depression (Reichenheim & Harpham 1989) as risk factors for child injuries. Injuries also appear to be both more prevalent and more severe in rural areas, perhaps because of the hazards of agricultural machinery, chemicals and exposed bodies of water (Rahman et al. 2005).

We present here data from the enrolment phase of Young Lives (YL), a longitudinal study of childhood poverty conducted in four countries – Ethiopia; Peru; Vietnam and Andhra Pradesh, India. Occurrence of, and risk factors for, different injury types are explored for children aged on average 12 months.

Data and methods

Full details of the YL Study are available at http://www.younglives.org.uk. The data from the first round of YL are now available from the UK Public Archive, Study Number SN5307 (http://www.esds.ac.uk). In each of the four countries, 2000 children of age 6–17 months (average 12) were recruited in 2002. YL employed a clustered sampling strategy with semi-purposive sampling of 20 ‘sentinel sites’ within each country selected by local experts to represent a range of regions, ‘policy contexts’ and living conditions, with oversampling of sites covering ‘poor’ areas (Wilson & Huttly 2004). Within sites, 100 children were selected by an equivalent of random sampling; the exact sampling procedure varied between sites because of topographical and administrative differences within and between countries, but was carefully documented to ensure a sample indistinguishable from the one drawn at random from qualifying households, with reasonable control of bias (Wilson & Huttly 2004). The response rate was above 90% in all the four countries. Although not nationally representative samples, for ease of presentation, country samples will be referred to by that country's name.

Data collection was by standardized, interviewer-administered questionnaire with the child's main caregiver. Training was given to all interviewers based on common guidelines and aimed at measuring items in the same way, as far as possible, across the four countries. Interviews, lasting 1–2 h, were conducted in the local language.

Ethical clearance was obtained from the participating research institutions in the United Kingdom and each YL country. Prior to interview, informed consent was obtained from participants. The funding agencies had no direct role in the design, implementation, data analyses or decision to publish the study.


Caregivers were asked whether the child had ever had a burn that left a scar, broken a bone or had a serious fall. Caregivers were also asked whether the child had ever had an illness or injury so serious they thought the child might die. Details of this illness/injury were recorded, making it possible to estimate the prevalence of near-fatal injuries. Responses were categorized as injuries according to ICD-10. Ambiguous cases were not included.

Risk factors

Putative risk factors were selected for analysis on the basis of existing literature and are presented in Table 1. A wealth index (categorized into quintiles in each country) was based on work conducted by the World Bank and Macro International and included housing quality, consumer durables and infrastructure such as water and sanitation. To assess probable cases of common mental disorders (CMD) among caregivers, a screening tool was used (the SRQ-20), which has been widely used in the developing country settings and has acceptable levels of reliability and validity (Division of Mental Health WHO 1994). The tool is not diagnostic and cannot separate anxiety from depression. It consists of 20 questions and has a 30-day reference period. Cut-off scores to determine how many ‘yes’ answers constitute a probable case have been validated against clinical assessments in each of the YL countries (Division of Mental Health WHO 1994; Harpham et al. 2003; Tuan et al. 2004). Accordingly, a cut-off of 7/8 to separate probable non-cases/cases of CMD was used.

Table 1.   Description of cohorts
No. of children1999201120521999
Males (%)52.853.850.051.5
Age at interview (%)
 6–8 months17.415.217.711.9
 8–14 months47.947.246.956.4
 14–17 months34.837.635.431.7
% From each region
 Region 1Addis Ababa (15)Coastal Andhra (34.8)Mountain (50.4)Central coastal (40.0)
 Region 2Amhara (20)Rayalaseema (30.1)Coast (34.6)Northern uplands (20.2)
 Region 3Oromia (20)Telangana (35.1)Jungle (15.0)Red River delta (20.0)
 Region 4SNNP (25)  Mekong River delta (20.0)
 Region 5Tigray (20)   
% Urban residents3525.166.320.0
% Main caregiver is not biological mother3.
% Child's biological father doesn't live in household16.
 No. of under-5s in household (median)1–5 (1)1–7 (1)1–5 (1)1–3 (1)
 No. of school-age children in house (median)0–8 (1)0–6 (0)0–6 (1)0–3 (0)
% Single-parent households6.00.352.70.55
% Single-room households65.550.328.741.5
Caregiver education (%)
Wealth index quintiles and range
% Regularly cared for by non-household member16.66.421.332.6
% Regularly left alone or with children under-512.
% Caregivers probable case of CMD33.630.130.521.3
% Children with long-term health problem10.14.620.74.2

Data analysis

Analyses were performed in the statistical package Stata 8.0 (StataCorp 2003; Stata Corporation, College Station, TX, USA). Prevalence of each injury type and an overall prevalence of any of the reported injuries were calculated. To allow for the clustered sampling, generalized estimating equations (GEE) were used in risk factor analysis. Multivariable GEE models were constructed by adding variables sequentially in order of significance from univariable analysis. Variables were retained in the final models if Wald tests gave a P-value of <0.05. This significance-driven approach was adopted because of the exploratory nature of these analyses and because little is known about risk factors for childhood injuries within these contexts. In addition, data were examined for the presence of consistent patterns.


Characteristics of the children and their households are shown in Table 1, illustrating some striking between-country differences (e.g. urban residency) as well as expected similarities (e.g. age and sex).

Patterns of injury occurrence

Patterns of reported injury occurrence are shown in Table 2. Overall injury prevalence ranged from 9.0% in Vietnam to 19.5% in India. The most frequent type of injury was serious falls (5.6–12.3%). Few children in any country had broken a bone (0.4–2.1%), while prevalence of burns was between 2.1% and 8.3%. Between 0.6% and 2.5% of children were reported to have sustained a near-fatal injury. In each country, males and females had very similar levels of each injury type.

Table 2.   Injury prevalence
 Ethiopia n = 1999, Total(%)India n = 2011, Total(%)Peru n = 2052, Total(%)Vietnam n = 1999, Total(%)
Burn that left a scar 87 (4.4)166 (8.3) 42 (2.1) 66 (3.3)
Broken bone 41 (2.1) 19 (0.94) 17 (0.83)  8 (0.40)
Serious fall141 (7.1)213 (10.6)253 (12.3)111 (5.6)
 Fall resulted in vomiting or loss of consciousness 70 (51.1) 31 (14.6) 76 (30.0) 17 (15.3)
Near-fatal injury 15 (0.75) 41 (2.0) 52 (2.5) 13 (0.65)
 Taken for treatment for this near- fatal injury 11 (73.3) 25 (61.0)  8 (15.4)  9 (69.2)
Any injury242 (12.3)393 (19.5)330 (16.1)180 (9.0)

Risk factors for injury

Results from multivariable models are shown in Table 3. Some risk factors and injury types could not be explored because of low prevalence. There was a considerable degree of agreement in terms of direction and magnitude of effect of risk factors between countries and injury types.

Table 3.   Risk factors for injury: multivariable models
Risk factorOdds ratio P-value (95% confidence interval)
  1. *As defined in Table 1.

  2. †Not statistically significant but presented in order to show pattern. Other variables in the model are not adjusted for this factor.

  3. ‡Odds ratio across increasing categories as defined in Table 1; quintiles for WI.

 Caregiver common mental disorder1.77 (1.15–2.73), P = 0.0101.71 (1.20–2.45), P = 0.0032.16 (1.29–3.62), P = 0.0041.45†(0.80–2.62), P = 0.217
  11, P = 0.1101, P = 0.0161, P = 0.011 
  20.83 (0.37–1.85)3.01 (1.04–8.73)0.53 (0.33–0.86) 
  30.76 (0.37–1.56)1.71 (1.13–5.29)0.34 (0.11–1.04) 
  40.64 (0.28–1.46)   
  50.32 (0.13–0.80)   
 Father does not live in household0.58 (0.35–0.96), P = 0.034   
 Single-parent household  3.04 (1.29–3.62), P = 0.001 
 Age‡  1.90 (1.16–3.09), P = 0.0101.57 (1.10–2.23), P = 0.012
 Long-term health problem   2.85 (1.46–5.57), P = 0.002
 Rural vs. urban residence   1.99 (1.06–3.73), P = 0.032
Broken bones
 Caregiver common mental disorder1.98 (1.13–2.47), P = 0.0183.67 (1.83–7.34), P < 0.0012.56 (1.16–5.68), P = 0.020 
  1 1, P = 0.0051, P < 0.001 
  2 2.75 (0.77–9.75)3.63 (1.11–11.89) 
  3 0.57 (0.14–2.24)8.92 (2.89–27.55) 
 Father does not live in household2.22 (1.13–3.47), P = 0.005 2.88 (1.04–7.96), P = 0.042 
 Long-term health problem 3.93 (1.25–12.33), P = 0.019  
Serious falls
 Caregiver common mental disorder2.44 (1.49–4.01), P < 0.0011.74 (1.22–2.47), P = 0.0022.33 (1.84–2.95), P < 0.0014.20 (2.98–5.93), P < 0.001
  1 1, P < 0.0011, P = 0.0131, P < 0.001
  2 0.26 (0.093–0.73)1.39 (0.99–1.96)1.64 (0.99–2.72)
  3 1.62 (0.66–3.97)1.77 (1.20–2.60)3.60 (2.41–5.39)
  4   2.97 (2.00–4.41)
 Regularly cared for by non-household member   1.83 (1.29–2.60), P = 0.001
 Regularly left alone or with under-5s  2.31 (1.26–4.24), P = 0.0074.77 (1.96–11.60), P = 0.001
 Age‡1.42 (1.16–1.75), P < 0.001 1.30 (1.09–1.54), P = 0.0031.30 (1.01–1.66), P = 0.040
 Caregiver education (none vs. some)  0.70 (0.49–1.00), P = 0.049 
 Long-term health problem1.94 (1.28–2.92), P = 0.002 1.57 (1.08–2.29), P = 0.017 
 Wealth index‡   0.83 (0.71–0.97), P = 0.016
Near-fatal injury
 Caregiver common mental disorder 1.51†(0.67–3.37), P = 0.3191.57 (1.02–2.41), P = 0.041 
  1 1, P = 0.0061, P = 0.021 
  2 0.28 (0.11–0.72)2.09 (1.11–3.91) 
  3 1.31 (0.66–2.61)2.49 (1.26–4.93) 
 Father does not live in household  2.13 (1.21–3.73), P = 0.008 
 Regularly cared for by non-household member 2.53 (1.05–6.08), P = 0.038  
 Wealth index‡ 0.69 (0.59–0.81), P < 0.001  
 Long-term health problem 4.82 (2.00–11.64), P < 0.001  
Any injury
 Caregiver common mental disorder2.02 (1.34–3.04), P < 0.0011.68 (1.28–2.20), P < 0.0012.26 (1.86–2.75), P < 0.0012.68 (1.94–3.70), P < 0.001
  1  1, P = 0.002 
  2  1.45 (1.09–1.93) 
  3  1.84 (1.27–2.65) 
 Mother not main caregiver0.29 (0.10–0.84), P = 0.023   
 Father does not live in household  1.46 (1.10–1.95), P = 0.010 
 Regularly cared for by non-household member 1.88 (1.05–3.36), P = 0.0351.83 (1.05–3.21), P = 0.0341.65 (1.23–2.21), P < 0.001
 Regularly left alone or with under-5s   4.23 (1.67–10.7), P = 0.002
 Age‡1.45 (1.24–1.69), P < 0.001 1.26 (1.08–1.47), P = 0.0031.53 (1.20–1.95), P < 0.001
 Long-term health problem1.42 (1.30–1.94), P = 0.0311.76 (1.07–2.92), P = 0.0271.36 (1.00–1.85), P = 0.005 

Children of caregivers classified as probable cases of CMD consistently appeared to be at greater risk of injury, generally experiencing an approximate doubling of odds of injury. Region of residence was also a common risk factor. Within each country, the association with region varied between injury types, which may explain the lack of association between region and ‘any injury’ in three of the four countries.

Older children sustained more burns in Peru and Vietnam, and more falls in Ethiopia, Peru and Vietnam. Long-term health problems of the child were associated with increased odds of burns in Vietnam; broken bones in India; falls in Ethiopia; Peru and Vietnam; near-fatal injury in India; and any injury in Ethiopia, India and Peru.

The child being regularly cared for by a non-household member was associated with an approximate doubling of odds of serious falls in Vietnam; near-fatal injury in India; and any injury in India, Peru and Vietnam. Regularly leaving the child alone or with other children under-5 was associated with increased odds of falls in Peru and Vietnam, and any injury in Vietnam. This was not a statistically significant risk factor in Ethiopia or India, where the proportion of children regularly left alone or with under-5s is greater than in Peru or Vietnam.

Wealth index was only statistically significant for falls in Vietnam and near-fatal injury in India; in both of these cases increasing wealth index was associated with decreasing odds of injury. Variables that were statistically significant in only one situation were single-parent households (increased odds of burns in Peru), the mother not being the main caregiver (decreased risk of any injury in Ethiopia) and rural residence (increased odds of burns in Vietnam). Gender, single-room homes and number of children in the household did not show significant associations with any injury type in any of the countries.


This study supports the growing body of evidence about injuries in children in developing countries. Despite the young age of the study children, an age when they might be less exposed to a hazardous environment, a relatively high proportion of them had experienced a non-trivial injury. Within the limited child injury research and intervention agenda, infants are often neglected, and the levels of injury in the YL cohorts add to the evidence that this omission is unjustified.

The four YL countries represent a range of environmental and social contexts in which the risk of injury might be expected to vary substantially. Some of the variation in injury occurrence may be attributable to reporting differences whereby respondents in different societies differentially interpret questions relating to ‘serious’ falls or ‘near-fatal’ injuries. Some of the excess fall morbidity in India and Peru may thus be attributable to caregivers reporting less serious injuries. However, there is no clear correlation between relative prevalence of falls and proportion of falls leading to head injury, suggesting that although reporting may have differed between countries, this is not the only factor responsible for the differences in prevalence of falls. Nevertheless, a number of consistent patterns were seen. For example, among the injury types covered by YL, falls were the most common type of injury in each country, which is consistent with the findings of many studies in the developing countries (Bangdiwala et al. 1990; Kibel et al. 1990; Adejuyigbe et al. 1992; Sharma et al. 1993; Tandon et al. 1993; Doumi et al. 1994; Semple et al. 1998; Linnan et al. 2003; UNICEF/TASC 2004).

The proportion of children receiving treatment after a reported near-fatal injury varied between countries, potentially reflecting both access to health care and reporting tendencies. In Peru, 50.0% of children were taken for treatment for what the caregiver considered was a near-fatal injury, compared with 73.3% in Ethiopia, despite health care access being the greatest in Peru and the lowest in Ethiopia.

A striking result of this study is the consistent association shown between injury occurrence and the caregiver's likelihood of having a CMD. Despite much evidence (predominantly from the developed world) showing that poor caregiver mental health has adverse consequences for other aspects of child health and well-being, there is very limited literature on its effects on child injury risk. One study showed an association between impaired caregiver mental health and an increased risk of burns/scalds and two or more injuries (O'Connor et al. 2000), but no association with long-term injury, although the definition of long-term injury is unclear. Another study found no association between caregiver mental health and child injury risk (Ramsay et al. 2003). The only study identified in a developing country setting is a community-based study in Brazil, which found a 1.5-fold increase in odds of any child injury in the fortnight preceding interview for children whose mothers were classified as stressed or depressed (Reichenheim & Harpham 1989). It is important to recognize the limitations of a screening tool, such as the SRQ-20, for classifying individuals. Further, the causal mechanisms, including the direction of association, cannot be ascertained from the data presented here. It could be hypothesized that the effect of caregiver CMD on child injury risk may be mediated through decreased supervision level, or that difficulties caused by a child's injury result in poorer caregiver mental health. Nevertheless, the fact that this indicator of caregiver mental health shows such a consistent association with child injury risk across various injury types and in the four countries, draws together two issues in public health, which are largely neglected in terms of resource allocation.

Long-term health problems of the child were associated with several injury types. Other studies have shown poor child health to be associated with increased risk of injury (Forjuoh et al. 1995b; Ramsay et al. 2003; Mattila et al. 2004). The specific problems suffered by the children are wide ranging, from epilepsy to eczema, but small numbers preclude disaggregating the effects of separate health problems on injury risk. Some of the injuries may have been a consequence of the long-term health problem, for instance, epileptic seizures, while some injuries may result in lasting disability.

Other risk factors that showed varying degrees of consistency across countries and injury types were region of residence, father not living in the household, single-parent families, age, wealth index, regular care of child by a non-household member, leaving the child alone or with other under-5s, caregiver had no education and main caregiver not the biological mother. The increased injury risk associated with regular care of a child by a non-household member may be attributable to supervision levels, particularly if the child is in a crèche with a high child-to-staff ratio; however the timing of injury is not known. In Brazil, injury rates were lower in children cared for by non-relatives (Bangdiwala & Anzola-Perez 1990). Although wealth index was in itself only statistically significant in two models, all of the consistent risk factors were themselves associated with wealth index and are thus likely to be mediating factors for its impact on injury risk.

Gender, number of children in the household and single-room households were not statistically significant risk factors in any of the final models, despite being identified by other studies as important risk factors for child injury (Bijur et al. 1988; Janson et al. 1994; Bang et al. 1997; Ahmed et al. 1999; Scholer et al. 1999; UNICEF 2001; Delgado et al. 2002; Blakely et al. 2003; Celis et al. 2003; Ramsay et al. 2003; UNICEF/TASC 2004; Villalba-Cota et al. 2004; Petridou et al. 2005; Rahman et al. 2005). Reasons for this may be related to the settings and age group being quite different from many of these studies. For example, the importance of gender, and perhaps other risk factors, may only emerge in older children. The lack of association with urban/rural residence in all but one model may be partly explained by having included region in the analysis. Many other studies examining risk factors did not include multiple potential risk factors, and as such the apparent effects of some of these factors could be due to confounding.

Methodological considerations

As mentioned earlier, the data are cross-sectional and thus directionality of associations cannot be confirmed. Furthermore, analysis is limited to non-fatal injuries. The questionnaires did not incorporate questions on severity or external cause. Although some instances of violence were reported, it is likely that intentional injury has been under-reported in this study. As such, the findings relate primarily to unintentional injury. Furthermore, community-level influences on injury risk, such as environmental hazards and availability of safe play spaces, were not explored here.

In addition to the possibility of different interpretations of the injury questions referred to earlier, questions relating to potential risk factors may have been interpreted differentially between countries; for instance, the Ethiopian team reported that many respondents had difficulty understanding the caregiver mental health questions (Alemu et al. 2003). Any misclassification of caregiver mental health is likely to be non-differential and thus may have underestimated the association with injury.

Another issue that could affect the reliability of prevalence estimates is recall. Caregivers were asked about injuries sustained over the child's life so far, an average recall period of 12 months but ranging between 6 and 17 months. A household survey of non-fatal injury in Ghana concluded that 1- to 3-month recall periods should be used when calculating overall non-fatal injury prevalence, but recall periods of 12 months could be used when assessing severe injury (Mock et al. 1999). An American study showed a 40% decline in the reported injury prevalence using 12-month compared with 1-month recall period, with larger declines for younger children and less severe injuries (Harel et al. 1994). However, the study had no measure of ‘actual’ prevalence; so it is not possible to tell whether the longer period underestimates or the short period overestimates. If recall is an issue, it could be expected that injury prevalence in the 1-year-olds may be higher than is estimated here.

Analyses of ‘any injury’ involved grouping of a diverse set of injuries with distinct causes. It is unlikely that all injuries were captured, thus these analyses may not depict the entire injury burden.

Young Lives has broad aims; data on specific issues, therefore, tend to reflect breadth across topic areas rather than depth. It was designed to study relationships between aspects of childhood poverty, rather than as a tool for collecting national statistics, and thus nationally representative samples were decided against. The generalizability of the study's findings is subject to some limitations, but this is not a feature unique to YL. The strong consistencies observed in the risk factors here suggest that their importance crosses social, cultural and political boundaries. A particular strength of these analyses is that the breadth of the YL data allowed examination of a wide range of potential risk factors.

In conclusion, the overall prevalence of non-trivial injuries among these young children was high, confirming the need to consider childhood injury as a public health priority in developing countries. The risk factor results indicate several areas worthy of follow-up through more in-depth studies and through longitudinal data. In particular, the associations seen with caregiver mental health add evidence to the growing interest in understanding the effects of this factor on a range of child well-being indicators.


The authors are grateful to the numerous members of the YL team who contributed to the design and conduct of the project, to the cohort participants and their families in each country and to the UK Department for International Development, which funded the first phase of the YL project. Laura Howe was funded by a UK Medical Research Council Advanced Course Studentship.