Objective. We describe the frequency and patterns of injury affecting 96 359 children between 0 and 10 years old and living in Alberta, Canada.
Design. This population-based, longitudinal study involved children born in the 3 fiscal years of April 1, 1985 to March 31, 1988, recruited before age 1, and who remained in the study until at least age 5. We used the International Classification of Diseases, Ninth Revision, Clinical Modification chapter-17 diagnostic codes provided by physicians. Codes were grouped into 17 categories; injury episodes were calculated, and age- and gender-specific incidence rates for each category were calculated. The age, pattern, times of greatest risk, and the effect of gender on the type and incidence of injury were determined.
Setting. Health care administrative data were obtained from all fee-for-service health care venues in Alberta between April 1, 1985 and March 31, 1998 providing services to children registered with the Alberta Health Care Insurance Plan and otherwise meeting entrance criteria.
Results. Nearly 84% of children received care for an injury during the study period, and in any given year ∼21% of the population studied had at least 1 injury. Repeat injury was common (73%), and boys were more likely than girls to be injured and to have repeat injury. The most common injuries were dislocations and sprains, open wounds, and superficial injuries and contusions. Burns, poisoning, intracranial injury, and foreign bodies were the next most common, and fractures were least common. Approximately 10% of injuries were multiple-category injuries. Rates varied greatly by injury category, age, and gender. Hospitalization rates varied in a similar manner and commonly accounted for ∼10% of all services. Males were most likely to have an injury, and aboriginal children or children who had received welfare at some time were at greatest risk.
Conclusions. Administrative data can be used to estimate the incidence of injury in a pediatric population. Distinct patterns of injury occur at different ages. Recurrent injury is common. Almost identical proportions of injury (46%) are treated in emergency departments and physicians’ offices.
Nearly 1 in 4 children in the United States are injured each year, costing the American society $347 billion per year.1 Most injuries are preventable, and preventable injury is the major cause of death and morbidity among children >1 year old. In Canada, although absolute rates dropped significantly over the past 20 years, in 1997 they accounted for 32% of deaths in children aged 1 to 4 years, for 41% of all deaths among children aged 5 to 9, and for 52% of deaths for children aged 10 to 14.2 Death due to injury is a poor indicator of the extent of the problem,3,4 because it does not reflect the high ratio of injuries to deaths. Guyer and Gallagher3,4 estimated that for every child killed in a motor vehicle accident, 48 were injured. Although a physician’s office is the usual place for injuries to be dealt with professionally, injuries still account for 9% of all preschool hospital admissions2; they are the second leading cause of hospitalization among children aged 5 to 9 and the leading cause among children 10 to 14.2 Injury is a key cause of days away from school, or from the normal developmental tasks of childhood, and can lead to permanent disability.5
Few studies of injury epidemiology provide a clear description of the frequency and types of injury occurring over the first 10 years of life. Most descriptions are based on sources such as mortality files, death certificates, medical examiner reports, hospital discharge data, emergency department (ED) data, emergency medical services data, trauma registries, health maintenance organization records, and poison-control centers.6 Each of these can skew the data in various ways. Various researchers have used samples of claims data7,8 to explore aspects of injury epidemiology but have not used them to describe patterns of injury occurring in childhood. Survey data have also been used.9 These permit data to be obtained describing the familial, socioeconomic, and physical environment and the context of the injury event, but surveys do not cover all individuals and they depend on personal recall.10 A key problem with much of the data from the sources listed above is determining the appropriate denominator with which to calculate rates.6 Cases should come only from the population in the denominator, but with many of the sources described above, accurate determination of a denominator is impossible.
This study attempts to address some of the above-mentioned problems and presents an estimate of the incidence of injury in children in Alberta, Canada. It uses diagnoses provided by physicians when submitting billing data for reimbursement in a universal provincial health care insurance program. We use these data to describe the prevalence of injury in children recruited in the first year of life and followed through 9 years old and living in Alberta from April 1, 1985 through March 3, 1999. These data include all physician services attributed to injury and seen in physicians’ offices, EDs, and hospitals. With these data we can calculate rates of injury specific to injury type, age, gender, time of year, and geographic region. Although physician diagnostic data have been used before, data from such a large and diverse population are rare, particularly when an accurate denominator is available reflecting the total population at risk. This research asks: What are the rates of injury from infancy through age 9 among Alberta children, and how do these rates vary by age and gender? No other published Canadian study exists that uses health care administrative data in this manner to describe patterns of injury, and no published study anywhere has captured virtually the entire population of children in a large, defined, geographical and administrative area.
With a population of ∼2.8 million people, Alberta is Canada’s western-most prairie province. Most residents are of European descent, but there is also a significant number of First Nations (Aboriginal) Canadians and individuals of Asian descent. Approximately 75% of children and adolescents live in an urban setting; the rest live in rural areas, mainly small towns, farms, and ranches. Extreme poverty is uncommon. Health care in Alberta is universal, and all medically necessary procedures and investigations are provided through the Alberta Health Care Insurance Plan administered by Alberta Health and Wellness (AHW).
The study used a closed-cohort design, and the data are based on the health care records of a cohort of children alive at 5 years old and who were registered with Alberta Health Care Insurance Plan before age 1 year (usually in the neonatal period) between April 1, 1985 and March 31, 1987. Excluded children are those dying before age 5 years, those born in Alberta but left Alberta before age 5, those who first moved to Alberta after age 1, or those who were never registered. Children were followed from age 5 until they left the plan due to emigration or death.
Data were available regarding the service date, the site of service, and the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM).11 Only records for which the ICD-9-CM code reflected chapter 17 (Injury and Poisoning) were used. Within this chapter, we did not consider those “injuries” having ICD-9-CM codes 905–909.9 (late effects), 958–958.8 (traumatic complications), and 995–999.9 (complications of medical care). Each diagnosis could be reported with up to 4 digits (eg, 900.1), but commonly only 3 digits were used. Only diagnostic data were provided and external cause-of-injury codes were not available. For analytic purposes, care was grouped as being provided in a doctor’s office or ED or as hospital inpatient care.
Demographic data were obtained from a registration file maintained by AHW for all children enrolled. The data included the child’s birth date, gender, health region, insurance premium status, and the end date for insurance coverage, if applicable. Health care is universal in Alberta but is administered through a publicly funded insurance plan, which covers the cost of premiums if the individual cannot do so or if the individual is covered by an outside agency. The latter is the case for First Nations children who also had treaty status; these children had their premium paid for by the federal government. Not all First Nations children had treaty status, but all treaty-status children were First Nations. Treaty-status children form a recognizable subset of the population whose health care premium is paid for in full by the federal government. The insurance premium status was used to create a crude index of economic status. Children with no premium subsidy were considered to be at the highest socioeconomic status level and constitute the reference group for comparison purposes. Children receiving some subsidy were considered as next highest. Although some of the treaty-status children could have come from well-off families, the majority lived near the poverty line. Children receiving welfare were the least well off. This results in 4 categories of socioeconomic status: no subsidy, subsidy, treaty, and welfare. The child’s “study age” was the integer age at the time of an event. The age denominator was the integer age on March 31 of the fiscal year of the event; this was the date used by AHW for administrative purposes. Identification numbers unique to each child and present in both data sets linked demographic data and injury data. It is estimated that >99% of eligible children are registered with AHW.12
These are administrative data, and their contents cannot be verified practically. The data cannot be linked, for example, to the corresponding written record in a physician’s office, nor can they be readily linked to a hospital record. Hence, the injury data are unverified, and coding or data-entry errors may not have been detected. This could result in diagnostic error. In an attempt to reduce this possibility of error, we grouped the ICD-9-CM codes into 17 categories that were sufficiently broad such that the error is not likely to be significant or systematic. Thus, for example, all ICD-9-CM codes 820–829 are considered as a fracture of the lower limb. By “rolling up” the codes, the accuracy is improved13 but at the expense of precision.
The data reflect every patient service for which a physician was reimbursed by fee for service. Except for a few small, subspecialty programs such as intensive care physicians, some of whom provided their services by contract, all patient services were provided this way; thus, virtually all contacts with the health care system were captured. Contacts with chiropractors and some forms of dental care are included, but other nonphysician contacts such as with physiotherapists are not included, and we cannot estimate their use. Services provided to children registered with AHW but who received medical or surgical services out of province were included in the data set.
Considering each record as an injury event would be misleading, because some injuries clearly need multiple services before resolution; therefore, we created “episodes” of injury based on the assumption that a child would be unlikely to have a similar injury within a given time period from the onset of an injury. We chose 180 days to be consistent with Lestina et al.8 Thus, an episode of injury included all visits for the same category for which the time interval between visits for the same injury type was <180 days. Thus, a child could be seen by a physician for the first time at day “0” (t0) and be seen again at t32, t128, t256, and t400, and, because the time interval between visits was <180 days, all these visits would constitute 1 episode. Another child could have visits at t0 and t181, and these would be considered 2 episodes. The use of episodes reduces the number of false injury counts because of the severity of some injuries. It also permits an estimate to be made of the number of services per episode.
It is also possible that a single injury episode might consist of several injury categories. For example a child could fall and have a fractured skull and an intracranial injury. The presence or absence of multiple-category injuries (MCIs) occurring at the same time was determined by examining the data for other injuries reported within 7 days of the onset of a new injury. Because several categories are involved for 1 episode of MCI injury, this type of episode is “reported” twice: first as part of an MCI and second, to count and generate rates for individual injuries, the individual injuries sustained in the MCI are counted within their respective categories.
The data available covered the injury history of all children for 9 years, and thus the likelihood of a child having >1 injury recorded was high. We determined the presence of repeat injury and its frequency by determining the presence of a different category of injury with an interval of at least 8 days from the previous category. Events that were part of an MCI were not counted.
For each episode of injury, we determined the site (office, ED, or hospital) providing the highest level of care. Data from those episodes for which the highest level of care was hospital were extracted, and rates of hospitalization were created for each category.
Rates for each category of disease were created by using the total number of registered children for any given age and gender as the denominator. The numerator was the number of appropriately stratified episodes of a specific injury. Significant (P < .05) differences between genders for individual categories or for differences between services and episodes were determined by calculation of age-specific confidence intervals (CI). Factors associated with injury were explored by using logistic regression, as were factors accounting for repeat injury. All statistical analysis was performed by using Stata.14 Logistic regression was used to determine factors associated with the event of ever having an injury, having a repeat injury, and leaving the study before the end. Variables included were gender (reference: female), an index of the child’s economic status (reference: no subsidy), and whether the child moved residence within the province during the study period (reference: no movement). For repeat injury, the variable age at first injury was used as well as those mentioned above.
The University of Alberta Faculty of Medicine Ethics Review Committee approved this research. Confidentiality of subjects was maintained by using anonymized personal identifiers. Only 1 author (L.W.S.), an employee of AHW and under a bond of confidentiality, had access to the code that encrypted the identifiers.
There were 96 359 children enrolled; 51.3% were male. The last year with data for the entire cohort was 1997, when data for all children at age 9 were available. At this time, 94 354 children remained; thus 2005 children (2.1%) did not complete a full 9 years of follow-up; of these, 79 were lost due to death. Logistic regression describing those who did not complete showed that children who left were slightly more likely to have been on welfare and to have moved sometime during the study period.
The categories of injury and their associated ICD-9-CM codes are shown in Table 1, which shows the number of patient services (sorted in descending order), the number of episodes, and the number of individual patients affected for each ICD-9-CM code. Of the 96 359 children, 80 943 (84%) had a total of 345 887 injury service events within 242 456 episodes of injury. These episodes are summarized in Table 2, which shows the rates of episodes of injury for each injury category.
Boys were more likely to be injured (odds ratio [OR]: 1.82; 95% CI: 1.76, 1.88). Also, children who needed partial premium subsidy sometime during the study span were slightly less likely to be injured (OR: 0.82; 95% CI: 0.79, 0.86) than children needing no subsidy; First Nations children (OR: 1.35; 95% CI: 1.25, 1.46) and children who had been on welfare (OR: 1.57; 95% CI: 1.50, 1.66) were more likely to have been injured. Finally, children who moved residence were more likely to be injured (OR: 1.17; 95% CI: 1.13, 1.22).
The differences in rates between services and episodes for the top 9 categories, accounting for 97% of the service events, are presented in Fig 1, which also shows the rates for all injury combined and for MCIs. The solid diamonds reflect the presence of a significant (P < .05) difference between groups at the given age. Several different patterns between services and episodes are evident. In all instances, the higher the rate, the gap between services and episodes increases. With fractures and dislocations, the gap increases with age, whereas the gap is fairly constant with respect to superficial injury/contusion and open wounds; the gap is also constant for intracranial injury. Burns and poisoning show the greatest gap in early childhood.
Figure 2 shows the patterns of episodes of injury by age and gender for the top 9 categories, plus the pattern for all injuries combined and for MCIs. Again, the solid diamonds show significant differences, but the clinical significance of the differences is less certain. The most common ages for injury are ages 1 to 2, with a gradual drop in the preschool years and then a steady rise with age. Within this general pattern, however, there is significant age and variation by gender in the incidence of injury. Thus, fractures increase in incidence with age, as do dislocations, strains, and sprains; whereas poisoning is most common at age 2 and drops off significantly by age 4. Boys predominate in all categories.
The rates for “all injury” for each year varied between 143 per 1000 for infant girls and 374 per 1000 for 1-year-old boys. By age 9, 84% of children had a recorded injury. The most common injuries were open wounds (accounting for 28.5% of all injury episodes), superficial injuries and contusions (27.2%), and dislocations, strains, and sprains (16.5%). Fractures accounted for 8.5% of episodes, with nearly 18% of children having a diagnosis of fracture sometime in the first 9 years of life. Boys and girls generally had similar patterns of injury but often differed significantly in rates of specific injury, with boys in most instances more likely to be injured than girls. The patterns observed reflect to some degree the developmental status of the child and are consistent with the findings of others.15 Examination of the male/female rates for hospitalization for each of the injury types shows patterns very similar to those seen in Fig 2 (data not shown); the absolute values are considerably lower, however.
There were 7457 episodes of MCIs in 7355 children (9.4% of those injured), and 25 272 services were associated with these episodes. MCIs were most common in 1- and 2-year-old children. There were many combinations of injury, but most were a 2-way (97.8%) or 3-way combination of fractures, open wounds, superficial injury, dislocations, strains, and sprains, burns, and intracranial injury. A common combination at all ages was open wounds and superficial injury, accounting for 21% of all MCIs. There were distinct age-associated patterns of injury. In infants, the most common combination was open wound and intracranial injury, accounting for 22.4% of MCIs. The combination of intracranial injury and skull fracture accounted for an additional 13.7% of MCIs in infants. Among 1- to 4-year-old children, open wounds in association with dislocations, strains, and sprains, intracranial injury, or fractures of the upper or lower limbs accounted for 26% of MCIs. Dislocations, strains, and sprains and fractures of upper and lower limbs accounted for an additional 9.8% of MCIs, and intracranial injury in association with open wounds or superficial injury accounted for 14% of MCIs. In children aged 5 to 9, dislocations, strains, and sprains in association with fractures of the upper and lower limb, open wounds, or superficial injury accounted for 36.5% of MCIs.
The rates of hospitalization are shown in Fig 3 for each category of injury. Most commonly, rates were between 10% and 20% of the total rate for a category. The proportion of children with fractures requiring hospitalization was close to 80% at some ages. With few exceptions males were more likely to be hospitalized, and much of the differences between genders is statistically significant.
In early life, nearly 60% of episodes of injury are seen at least once in an ED. As the child ages, the highest level of care for an injury episode is most likely to be in a doctor’s office. Hospital care is provided in 10% to 12% of injuries in young children; this figure drops to ∼6% by age 9. Overall, ∼46.1% of injuries are treated in the physician’s office, and 46.3% are treated in the ED; 7.6% of injuries required hospitalization. Because there were 79 deaths and ∼41% of deaths in the 5- to 9-year-old range are due to injury,2 we estimate that ∼32 children died due to injury. Thus, there was ∼1 death for every 73 hospitalizations or 1 death for every 1616 ED visits among children aged 5 to 9.
Because these data are longitudinal, we could determine the frequency of a new injury occurring in the same child. Of the 80 943 children with an injury, 60 737 (73%) had >1 injury in the 9-year span of observation. Overall, the number of injuries per child averaged 3.0 ± 2.0 (mean ± standard deviation). The average age at first repeat was 4.7 ± 2.6 years. Eighty-nine percent of children had ≤5 injuries; 59 children had ≥15 separate episodes of injury. The age at first injury was the single-most important determinant of a repeat injury. Children with initial injuries before age 2 were 3.99 (95% CI: 3.84, 4.13) times as likely to have a repeat injury than were children whose first injury occurred at ≥2 years old. The next most important factor was gender, with boys being 1.42 (95% CI: 3.7, 1.47) times more likely than girls to have a repeat injury. Treaty-status children and children who had been on welfare were 1.41 (95% CI: 1.31,1.52) and 1.31 (95% CI: 1.25, 1.37) times, respectively, more likely than the reference group to have repeat injury. Mobility was only marginally associated with repeat injury (OR: 1.04; 95% CI: 1.00, 1.08). As might be expected, there was a steady downward progression of likelihood of repeat injury as age at first injury rose. We explored factors surrounding children with extremes of injury, comparing children with 1 repeat injury to those with ≥4 repeat injuries. Males are more likely to have ≥4 injuries (OR: 1.74; 95% CI: 1.68,1.81) compared with children who had 1 injury during the study period. Treaty-status children (OR: 1.71; 95% CI: 1.58,1.86) and children who received welfare (1.59; 95% CI: 1.51, 1.68) were more likely than the reference-group children to have ≥4 injuries. Children who moved between health regions were marginally (OR: 1.07; 95% CI: 1.03,1.1) more likely to have multiple injuries.
These results present a comprehensive picture of the patterns of injury of young children. Unequal access to health care due to economic constraint was not a problem, because health care is universal, and physicians serve all parts of the province. The study is population based and reflects virtually every stratum of society in Alberta; thus, the data are readily generalizable to Canadian populations but possibly less so to countries in which health insurance is not universal. The entire fee-for-service injury history of each child was available for analysis, and virtually every injury requiring medical attention and accruing to the children in this cohort is recorded. Any existing selection bias lies in the fact that the study children had to be registered by age 1 and maintain continued residence in Alberta until age 5. Thus, children who had multiple, between-province residences may have been undercounted.
As used elsewhere,4,7,9,15 we categorized the data into injury types. Although this eases data analysis, it also reduces diagnostic precision; thus, a fracture of the leg could be a compound fracture of the tibia, but it also could be a broken toe. Each type of injury would carry the same weight for analysis.
Various types of injury evolve apparently in concert with the age and developmental status of the child and their exposure to risk. Rivara et al15 commented on this developmental pattern, and it makes intuitive sense. The gender differential so apparent in Fig 2 is consistent with the results of others.4, 15–17 Except for foreign body, for which rates for girls predominated at ages 1 and 2, we found that males were generally more likely than females to be injured, a finding reflecting a more risk-taking behavior of male children.
The rates of injury are consistent with others,1,4,9 and, even allowing for multiple injuries in a child in a given year, these figures suggest that significant proportions of children are injured every year. More importantly, they demonstrate that nearly 84% of children are medically treated for injury some time in the first 10 years of life. Several authors1,9,10,16 have suggested that ∼20% to 25% of children are injured in a 1-year period. We determined that ∼21.0% of children had at least 1 injury each year.
We cannot comment as to the causes of injury, nor can we comment on context, because these data were not collected. The lack of external cause-of-injury codes is particularly regrettable, because much of the literature uses them to form the basis for categories for examination and for discussing injury epidemiology from the perspective of cause rather than type. Our study of necessity must examine types of injury and patterns. Using episodes to describe the data are necessary. This approach has been used by Lestina et al,8 who used a 180-day interval to describe injury episodes. Episodes reduce the chance of overcounting the number of injuries incurred, they help assess injury severity, and they allow some analysis of health care utilization and cost estimates. Using a 180-day interval may underreport a child’s injury experience. The effect of this is a conservative estimate of injury among children. A child could have a similar injury within a 180-day period, and this would not be counted as a new episode. However, it is certain that we would double count many episodes if we did not use some form of window of time during which all similar diagnoses would be ignored. The use of the concept of an episode also allows for an estimate of injury severity; thus, the increasing gap between all services and 180-day episodes seen with increasing age for dislocations suggests that more medical services are required and the overall acuity of dislocation rises with age (Fig 1). Similar arguments might be made for fracture of the arm and leg. Burns, on the other hand, show a large gap at ages 1 to 3, suggesting that these burns are more severe than later in childhood and that they pose a greater risk at this age.
MCIs were common, affecting nearly 10% of those injured. These combinations may reflect diagnostic differences; thus, the common combination of open wounds and superficial injury reflects differing diagnoses of the same injury. MCIs may reflect diagnostic differences or a changed diagnosis after later assessment. Finally, it may reflect 2 concurrent diagnoses.
Because the data are complete, it is possible to determine with some confidence where the child received medical care. In the early years, children were more likely to go to an ED than in later years, and hospitalization was also more frequent. The doctor’s office was a significant source of medical care at all ages; this permits a more-complete picture of injury than seen in most studies, and it may reduce the bias in the type of injuries described in these studies. The rates for hospitalization showed varying patterns. Boys generally were more likely to be hospitalized, but the pattern remained similar in most cases for both boys and girls; it was the magnitude that changed. Although hospitalization generally infers increased severity of injury, it is difficult to assign severity within the hospitalization data; many of these admissions may have been overnight stays for observation.
Localizing the site of highest level of care permits an estimate of an “injury pyramid” relating death to other injuries. Our ratio of death/hospitalization/ED visits for children aged 5 to 9 of 1:73:1612 is in fair agreement to the findings of Gallagher et al,4 who had a ratio of 1:45:1300 in Massachusetts, and 1:43:1732 reported by Gofin et al17 among Israeli children. Both Gallagher et al and Gofin et al used the age range of 0 to 17 inclusive, whereas our data reflect only ages 0 to 9. The ratios vary by age group. Using data provided by Health Canada,18 the death/hospitalization ratio for 0 to 17 years is 1:50, but for children aged 5 to 9, it is 1:86. With our data we can extend the pyramid for 5- to 9-year old children to 1:73:1612:1606 by adding in the death/physician visits ratio.
The data are longitudinal, and thus we can detect repeat injury in the same child; in fact, 76% of children saw a physician for a repeat injury. Boys and children having their first injury at an early age are more likely to have a repeat injury. Neither of these facts is surprising, and an early age for the first injury just means there is a longer exposure time for later injuries. The fact that First Nations children and children who had received welfare at some time during the study period were also at greater risk for repeat injury may point to factors within the environment in which they live. A very small proportion of children had ≥15 separate episodes of injury. We cannot determine whether these children were “accident prone,” lived in a dangerous environment, or had some other underlying pathology or chronic illness that predisposed them to this very high figure. This figure seems very high and is a cause for concern for those children.
This study used health care administrative data to describe the patterns of injury in 96 359 children aged 0 to 9 years. Nearly 86% of children sustained at least 1 injury during the 10-year span of the study; 76% of those with an injury sustained a repeat injury. Boys were more likely to have an injury, as were First Nations children and children needing welfare sometime during the study period. These statistics underline the importance of injuries in the health of children.
This research was supported by a grant from the Children’s Health Foundation of Northern Alberta.
Alberta Health and Wellness provided the data used.
- Received July 25, 2002.
- Accepted July 7, 2003.
- Address correspondence to Donald William Spady, MD, Department of Pediatrics, 2C3.92 WMC, University of Alberta, Edmonton, Alberta, Canada T6G 2R7. E-mail:
- ↵Danseco ER, Miller TR, Spicer RS. Incidence and costs of 1987–1994 childhood injuries: demographic breakdowns. Pediatrics.2000;105(2) . Available at: www.pediatrics.org/cgi/content/full/105/2/e27
- ↵Canadian Institute of Child Health. The Health of Canada’s Children, a CICH Profile. 3rd ed. Ottawa, Canada: Canadian Institute of Child Health; 2000
- ↵Collins JG. Impairments due to injuries: United States, 1985–87. Vital Health Stat 10.1991;(177) :1– 55
- ↵Robertson LS. Injury Epidemiology. New York, NY: Oxford University Press; 1992
- ↵Health Care Financing Administration. International Classification of Diseases, Ninth Revision, Clinical Modification. 3d ed. Vol 1, 2, and 3. Official authorized addendum effective October 1, 1990. Los Angeles, CA: Health Care Financing Administration; 1990
- ↵Svenson LW, Woodhead SE, Platt GH. Estimating the prevalence of asthma in Alberta: a study using provincial health care records. Chronic Dis Can.1993;14 :28– 33
- ↵Stata Statistical Software [computer program]. Release 7.0. College Station, TX: Stata Corporation; 2001
- ↵Health Canada. For the Safety of Canadian Children and Youth. Ottawa, Canada: Minister of Public Works and Government Services; 1997
- Copyright © 2004 by the American Academy of Pediatrics