Results

During the two summers, 930 injuries from falls were reported. Table 1 presents the nature and the body part of the injury. It also contains a severity score derived from the AIS. Nearly 12% of all injuries were head injuries, and 55% were fractures. Other injuries included sprains/strains (11%), contusions (11%) and lacerations (8%). More than half of the children sustained an injury to their upper limbs (54%). Overall, 42% of all injuries were rated as AIS 1, 43% as AIS 2, and 15% as AIS 3. Table 2 presents descriptive statistics. The mean age was 5.8 years, and 54% were boys. One child out of four was injured in a residential playground. A module (composite playstructure) was most often involved, followed by climbers, swings, and slides. Approximately 60% of the parents reported being present when the incident occurred.

Almost 27% of injuries were sustained on grass, mostly in residential settings. Among these, 75% were fractures or head injuries, compared with 61% on sand (69% of AIS2–3 were on grass versus 54% on sand, table 3). Correspondingly, residential playgrounds were associated with a greater proportion of fractures and head injuries (75%) and of AIS2–3 (64%). The adjusted risk of fractures and head injuries and of AIS2–3 injuries was 1.7 times greater on grass than on sand. These injuries were also more likely to happen at home than at public playgrounds although the association was not statistically significant for AIS (ORa= 1.29; CI: 0.91, 1.84) .

Discussion

This is one of the first epidemiological studies to confirm that grass is not a good protective surface beneath play equipment. A child is more likely to have a head injury or fracture (or an AIS 2–3 injury) when falling on grass compared with sand.

One limitation of this study concerns the choice of paediatric hospitals because children visiting them may be different than those who consult elsewhere. Hence, the proportion of severe injuries and of fractures and head injuries observed in this study may overestimate the proportion that would be found if all of the emergency rooms in Montreal hospitals were included. Furthermore, children from outside of the Island of Montreal tend to go to paediatric hospitals for more severe injuries. Combined with the fact that these children also have more playground equipment at home than children residing elsewhere on the Island, we thought that this could inflate the effects. To verify this, a sub-analysis was performed limited to children from the Island of Montreal. The results were less striking, but in the same direction. In the final model, we controlled for this possible bias by including the accident location.

Other potential biases in this study would have decreased the relation between grass, nature, and severity of the injury. Because misclassification of surfaces by the parent is not associated with injury severity, it would bias the results toward the null given that parents tend to judge unsafe surfaces as safe. It was not possible to adjust for the height of equipment because this information from the parents was not reliable. Nevertheless, elsewhere we have shown that adjustment for height strengthens the association between surface and severity because equipment on sand tends to be taller.5

The multivariate analyses performed lent strength to the conclusion by allowing us to control for different personal and environmental covariates, and thus verify that the association between surfaces and nature of injury is independent of other potential confounders. Unfortunately, the number of grass and sand surfaces at home and at public playgrounds did not allow us to verify that the relation between grass, nature, and severity of the injury was the same in each of these two settings, independently. Nor could we control for the way children play at home or at a public playground, although we did control for variables that could be related to this (age, sex, number of children in the family, frequency of use of equipment). Our conclusions are further reinforced by the fact that the results were the same when the models were run separately for 1991 and 1995 and with AIS as the outcome variable. These results contrast with those of Sacks and his collaborators,3 but we believe that if they had controlled for height, they would not have considered grass as a safe surface.

A difference was observed between the two models with regard to the risk of fractures and head injuries, and to the risk of AIS2–3 at residential playgrounds when compared with public playgrounds. These results may be partly explained by the fact that it is mainly the head injuries sustained by young children that were rated as AIS1 and that this was more pronounced at home than in public playgrounds. Nevertheless, our results suggest that the risk of having a fracture or head injury is higher at home than at a public playground (ORa= 1.69) and that the risk of having an injury rated as AIS2–3 tends to be larger, although the results were not significant (ORa= 1.29). The use of residential playgrounds has increased since 1991, especially among children aged 1 to 4.4 As we have shown that children were exposed to higher risks of fractures and head injuries at home than in public playgrounds, more attention should be given to playground surfacing at home.