Article Text
Abstract
Objective To quantify the relationship between imitative behaviour and poisonings in children.
Setting USA.
Methods This study is based on the evaluation of a large national database of unintentional oral ingestion poisonings involving children aged <5 years treated in US hospital emergency departments during 2004 and 2005. It begins with the premise that, among other factors, oral drug poisonings can result from children observing and imitating adult behaviour, but that non-oral drug and non-drug poisonings (to be referred to as non-drug poisonings) generally do not, because children do not see adults ingesting non-drug products. The study then compares and contrasts the child poisonings between the two poisoning categories. Differences in the poisoning rate between the oral drug and non-drug categories are estimated by the age and sex of the children. A binary logistic regression analysis is also conducted using non-drug poisonings as a control group to compare against oral drug poisonings.
Results There was a significant increase in the relative likelihood of oral drug poisonings beginning at age 20–23 months that is consistent with the expected onset of complicated imitative behaviours in children. Based upon our analysis, imitative behaviour may have contributed to about 17 300 child poisonings treated annually in the emergency department, possibly accounting for about 20% of poisonings involving children aged <5 years and 30% of the poisoning injuries involving children aged 20–59 months.
Conclusions Comprehensive efforts to prevent poisoning need to address the problem of imitative behaviour in children. Caregivers should never ingest medications in the presence of children.
- Imitation
- observational learning
- poisonings
- logistic regression
Statistics from Altmetric.com
The investigation of imitative behaviour in children has played an important role in the study of cognitive processes, sociocultural learning and language acquisition. In this body of enquiry, researchers have found that an infant's ability to imitate may begin the development of a sense of self that progresses through childhood.1 More generally, imitative behaviour—defined as learning by observing another person's behaviour2–4—seems to help children acquire language skills, expand their adaptive physical and social repertoires and develop an identity.5
Despite the many adaptive uses for imitative behaviour during childhood, imitation can also have a negative impact on the developing child: it can result in unintentional injury. One prominent example is the unintentional ingestion of oral drugs by children. A number of studies have suggested that imitation of adult behaviours is a factor in some medication poisonings,6–12 and at least two studies have explicitly addressed the role of imitation in their study design.13 14 Many poison prevention efforts also explicitly recognise imitation as an important factor and recommend that caregivers avoid taking medications in the presence of children.15–18 However, there has been little effort to quantify the relationship between poisonings and imitative behaviour. Neither the age at which these behaviours can be observed in the poisoning data nor estimates of the number of poisoning incidents associated with imitation has been reported in the literature.
Despite the lack of quantification, the role of imitation in child ingestion poisonings—behaviour in which children ingest oral drugs because they have observed adults do so—has strong face validity. The normal use of an oral drug entails an adult modelling a behaviour (swallowing a medication) that would seem, to a child, to be related to the ingestion of food or candy or liquids: it is often part of a regular daily routine and involves ingesting food-like products from small containers.
In contrast to oral drug ingestions, ingestions involving other potentially toxic household products such as household cleaners, ointments or others are not generally the object of observational learning because children do not see adults ingesting these types of products. It is possible that children may occasionally ingest non-oral drugs or non-drug products (to be referred to as non-drugs) in a mistaken attempt to imitate adult behaviour—for example, this might happen if a child saw an adult using mouthwash or believed the packaging for a non-drug was similar to that of an oral drug. Nevertheless, as a general proposition, children's imitative behaviours are likely to be triggered by the availability of oral drugs which they have seen adults ingest, but not by the availability of other non-drug household products.
This study is an exploratory analysis of the relationship between unintentional child ingestion poisonings and imitative behaviour. It attempts to identify the role of imitative behaviour in child poisonings by comparing and contrasting unintentional oral drug poisonings, those identified as having an imitative component, with other ingestion poisonings which do not. To do this, we used a large national database of child poisonings that provides information on the age and sex of the children poisoned as well as the products involved in the poisonings. In effect, we use non-drug poisonings as a control group against which to compare the oral drug poisonings.
Methods
Data were from unintentional ingestion poisonings treated in hospital emergency departments (EDs) in 2004 and 2005 and reported through the US Consumer Product Safety Commission's National Electronic Injury Surveillance System (NEISS). NEISS is a stratified national probability sample of hospital EDs that allows researchers to make national estimates of product-related injuries. The sample consists of about 100 of the approximately 5400 US hospitals that have at least six beds and provide 24 h emergency service. It includes four strata based on hospital size and a fifth consisting of children's hospitals. Hospitals within each stratum are ordered by state and zip code and selected systematically to ensure a wide geographical coverage.19 20
Participating hospitals provide information on all injuries treated in the ED involving consumer products. The information includes the age and sex of the victim, the injury diagnosis of record, a description of the product involved, the disposition of the case and a narrative field for further description of the injury circumstances. Because NEISS categorises the age of children under 2 years in months, these children were categorised in 4-month increments (eg, 0–3 months, 4–7 months, etc). Thereafter, ages are categorised in 12-month increments corresponding to years 2–4.
For this analysis, poisoning incidents were defined as oral ingestions of potentially toxic substances by children under age 5 years in the absence of caregiver supervision. Cases excluded from the analysis included poisonings resulting from: (1) ingestions supervised by caregivers; (2) inhaled vapours, fumes or gases; (3) insect bites or contact with living plants, shrubs or trees; and (4) medical procedures or parenteral drugs administered by healthcare professionals. We also excluded cases in which children ingested oral drugs intended for their own use because some of these poisonings could have resulted from behaviours learnt during earlier medication ingestions taken under adult supervision rather than from imitation.
Because NEISS is a national probability sample of hospitals, cases were assigned a sample weight based on the adjusted inverse of the known probability of selection of hospitals in each stratum. Adjustments to these weights were made for non-response, hospital mergers and changes in the sampling frame. National estimates of ED visits and corresponding 95% CIs were calculated using the SAS 9.2 ‘surveymeans’ procedure.21
Included poisonings were separated into two categories: (1) oral drug poisonings including oral prescription and oral non-prescription drugs and dietary supplements; and (2) non-drug poisonings including non-oral drugs and non-drug poisonings involving all other potentially toxic household substances. Population-based poisoning rates were estimated using midyear age-specific population estimates for 2004 and 2005.22
For the analysis, oral drug and non-drug poisonings were categorised by the age and sex of the victim and poisoning rate differences were estimated. A binary logistic regression analysis was also used to compare the oral drug and non-drug poisoning categories using the SAS 9.2 ‘surveylogistic’ procedure.21
Results
Demographic and injury characteristics
A total of 4997 ingestion poisonings involving children aged <5 years reported through NEISS in 2004 and 2005 were included in the study. These cases represent an estimated 136 893 (95% CI 114 312 to 159 474) poisonings treated in US hospital EDs.
Child poisonings presenting to EDs increased substantially beginning at age 8–11 months, rising from 45.5 per 100 000 for children aged 4–7 months to a high of 730.5 per 100 000 for children aged 20–23 months (table 1). Almost 56% of the poisonings were in boys.
Product involvement
As shown in table 2, an estimated 61.6% of the poisonings involved oral drugs, including oral prescription drugs (36.5%) and oral non-prescription drugs and supplements (20.4%). Another 4.7% were identified as unknown oral drugs. An estimated 38.4% of the poisonings involved non-drug products. These included household cleaning products such as laundry products and drain and oven cleaners (12.8%); ointments and preparations for external use (5.3%); personal care products such as soaps, perfumes and hair and nail products (4.8%); low-viscosity hydrocarbons such as lighter fluid or baby oil (3.3%); and miscellaneous other non-drug products (12.3%).
Poisoning rates and rate differences
Table 3 shows the distribution of poisonings by product category and age and sex of the victim. Estimates of population-based poisoning rate differences, measuring the excess rate of oral drug poisonings over the rate of non-drug poisonings,23 are also presented. For children aged <20 months, the rate differences were close to zero or negative—that is, oral drug poisonings were approximately equal to or somewhat less than non-drug poisonings. However, the rate differences became positive and statistically significant for children aged 20–23 months and older, increasing from 171.7 per 100 000 for children aged 20–23 months to 239.3 per 100 000 for children aged 24–35 months. Thus, beginning at age 20–23 months, oral drug poisonings substantially exceeded non-drug poisonings.
When children aged 20–59 months are grouped together, the rate difference amounted to about 129.1 (95% CI 96.1 to 162.2) poisonings per 100 000 children. This compares with a rate difference of −20.1 (95% CI −56.2 to 15.9) for children aged 0–19 months, a difference not significantly different from zero.
Regression results
Table 4 presents the regression results comparing the relative likelihood of oral drug poisonings with non-drug poisonings. The adjusted ORs for the age categories <16 months were not statistically different from that of the 16–19 month reference group. The OR for children aged 0–3 months was relatively large but was based on only 13 observations and probably reflected the types of products put in close proximity by caregivers.
In contrast, the ORs for all categories of children aged ≥20 months were significantly greater than the reference category (indicating an increase in the relative likelihood of oral drug poisonings), and rose monotonically with age. Additionally, the OR was significantly lower for boys than for girls.
Further evaluation showed that the OR for children aged 24–35 months was significantly higher than for those aged 20–23 months (p<0.01). While the ORs continued to rise for children older than 24–35 months, neither the increase from 24–35 months to 36–47 months nor the increase from 36–47 months to 48–59 months was statistically significant.
The regression results are illustrated in figure 1, showing the expected likelihood of an oral drug-related ingestion poisoning by age for girls aged ≥12 months. The step function shows the probability estimates derived from the categorical age variable, rising from roughly 50% for girls aged <20 months to 75% for girls aged 48–59 months. The broken line represents the derived probability estimates smoothed with a Hodrick–Prescott filter.
Discussion
The analysis shows a substantial increase in both the oral drug and non-drug poisoning rates beginning at age 8–11 months and extending through age 16–19 months (tables 1 and 3). This generalised increase in the poisoning rate probably reflects the increasing mobility and dexterity of growing children and their desire and ability to explore their environment through mouthing behaviour.
Beginning at age 20–23 months, the analysis shows a substantial and statistically significant increase in the likelihood that poisonings involved oral drugs as opposed to non-drugs. We hypothesise that this increased likelihood of oral drug ingestion reflects imitative behaviour—behaviour in which children ingest oral drugs because they have observed adults do so. Moreover, the onset of this imitative behaviour in poisonings is fully consistent with the observations of Piaget suggesting that complicated imitative behaviours commonly appear in children aged 18–24 months.2 24 This does not mean that no imitative behaviour exists before the age of 20 months, but rather that the increase in imitative behaviours—coupled with increased mobility, dexterity and cognitive capacities—becomes evident in unintentional ingestion poisonings at about the age of 20–23 months.
The likelihood of an oral drug-related ingestion poisoning was also somewhat lower for boys than for girls at each age level. The sex differences are not great but may suggest, for reasons that are unclear, that girls begin to engage in poisoning-related imitative behaviour at a slightly younger age than boys.
Quantifying the role of imitative behaviour in oral drug poisonings
Given our hypothesis that the increased likelihood of oral drug poisonings beginning at age 20–23 months reflects imitative behaviour, the poisoning rate associated with imitative behaviour can be approximated if the relative exposure to oral drug and non-drug poisoning risks can be quantified. The motivation for this conclusion is described in appendix 1. As a first approximation, the results in table 3 provide some evidence that oral drug and non-drug risk exposures are approximately equal. According to these results, the observed poisoning rate difference for children aged <20 months was −20.1 (95% CI −56.2 to 15.9) per 100 000 children. Thus, the rate difference for these children was negative, but not significantly different from zero. Consequently, if ED-treated injuries in children aged <20 months generally reflect the underlying risk exposure when little or no imitative behaviour is present, and oral drug and non-drug poisonings are approximately equal, it may not be unreasonable to conclude that children's exposure to oral drug and non-drug risks are approximately equal.
If the risk exposures are approximately equal, our model suggests that the oral drug poisoning rate attributable to imitative behaviour would be the difference between the oral drug and non-drug poisoning rates for children exhibiting such behaviour. This is simply the population-based poisoning rate difference which, as shown in table 3, amounted to 129.1 (95% CI 96.1 to 162.2) poisonings per 100 000 children aged 20–59 months. This estimate, sometimes referred to as an excess or attributable risk,23 amounted to roughly 17 300 child poisonings on an annual basis during 2004 and 2005. These poisonings would account for about 20% of all poisonings involving children aged <5 years treated in US EDs and about 30% of the poisonings involving children aged 20–59 months.25 26
This estimate must, of course, be considered a first approximation. We did not observe the poisonings directly, and neither the injured children nor their parents were questioned about the factors behind their child's ingestion behaviour. Other factors may also influence a child intentionally ingesting oral drugs. For instance, the candy-like appearance of many oral drugs may entice children to eat them whether or not they have witnessed an adult ingest them. However, food-like features may also be found in the non-drug comparison group. Some household lemon-scented cleaners have packaging that looks like lemonade; windshield wiper fluid can appear to be a blue sports drink; glitter glue may look like toothpaste. The presence of this confounding factor in both categories of poisonings examined within this study is difficult to quantify, but could have affected the results.
It is also possible that some features of oral drug containers may make them more appealing for children aged >20 months to put in their mouths, independent of imitative behaviour. Many are small and may be more manoeuvrable than containers for other potentially toxic substances. If this were the case, however, it would seem likely that these same characteristics would also make them appealing as mouthing objects for children aged <20 months, although younger children may have more limited strength and dexterity. Additionally, many common non-drug products such as nasal sprays, nail polish or eye drops are contained in similarly attractive small manoeuvrable packages.
Finally, it is possible that increased access to oral drugs rather than imitation may explain some of the relative increase in oral drug poisonings beginning at age 20–23 months. However, increased access (and hence increased risk exposure) due to improvements in children's mobility and dexterity is not limited to oral drugs: increased access also applies to non-drug products. Just as improvements in mobility and dexterity may increase access to storage areas for oral drugs, so too will they increase access to storage areas containing non-drug products.
Because of these possible confounding factors, further analysis based on alternative methodologies is needed to examine our assumptions and test our findings. Future empirical research in an experimental setting could evaluate children's imitative behaviour and its timing, as children interact with oral medication surrogates after observing adult modelling. It would also be useful to conduct detailed follow-up investigations with caregivers to try to evaluate the role of imitation in actual poisoning cases. Nevertheless, our study presents a methodology for estimating the potential role and magnitude of imitative behaviour in child poisonings, and suggests that comprehensive poison prevention efforts need to address the problem of imitation.
Limitations
The results of this study are subject to several limitations. As a surveillance system for US hospital EDs, NEISS does not provide information on child poisonings treated in other settings such as physicians' offices, clinics and ambulatory surgery centres.
It is possible that some cases were incorrectly included as oral drug ingestions because the narrative describing the circumstances surrounding the injury failed to note that the ingestion had been administered by an adult rather than taken autonomously by the child. It is also possible that some children mistakenly ingested non-drugs believing they were oral drugs they had seen adults ingesting. The first scenario would bias the estimate of imitative behaviour upwards; the second would bias it downwards.
We excluded from the analysis cases in which children ingested oral drugs intended for their own use because some could have resulted from learning that was not associated with imitation. However, some of these cases (an estimated 13 409 (95% CI 10 516 to 16 302) poisonings over 2 years) may, in fact, have resulted from imitative behaviour even though the drugs ingested were the children's own. This would bias downwards the estimated poisoning rate associated with imitation.
Finally, as mentioned earlier, we did not observe the poisonings directly and neither the injured children nor their parents were questioned about the motivations behind the child's ingestion behaviour.
Conclusions
This analysis attempted to quantify the role of imitative behaviour in oral drug ingestion poisonings in children. It suggests that the hazard may be substantial and could have amounted to more than 17 000 child poisonings treated annually in hospital EDs during 2004 and 2005. While further empirical research is needed, these results suggest that poison prevention efforts need to address the problem of imitative behaviour in oral drug poisonings. Parents and caregivers should be informed that, because children frequently imitate adult behaviours, they should never ingest medications in the presence of children nor should children ever be told that medications are like candy. Whenever available, oral drugs should be kept in child-resistant packaging, which has been shown to be effective in reducing unintentional medication poisonings.27–29 Finally, oral drugs should be stored out of the reach of children in secure cabinets, even when they are already in child-resistant packaging.30 31
What is already known on this subject
With more than 80 000 unintentional child poisonings treated in US hospital emergency departments annually, child poisonings remain an important public health concern.
A number of studies have suggested that imitation of adult behaviour is an important factor in child poisonings, but there has been little or no effort to quantify the relationship.
Neither the age at which these behaviours become observable in the poisoning data nor estimates of the number of poisoning incidents associated with imitation has been reported in the literature.
What this study adds
This study presents a methodology for quantifying the relationship between imitative behaviour and child poisonings involving oral drugs.
It suggests that the impact of imitative behaviour on the poisoning rate may be substantial, possibly accounting for more than 17 000 US hospital emergency department visits annually and for as many as 30% of poisonings among children aged 20–59 months.
Appendix
In this appendix we present a simple model for quantifying the rate of oral drug poisonings resulting from imitative behaviour. We begin with the premise that unintentional oral drug poisonings by children result from either imitative or general exploratory behaviour, such that
where OralDrugtotal represents the rate of oral drug poisonings (eg, oral drug poisonings per 100 000 children), OralDrugexplore represents the rate of oral drug poisoning associated with general exploratory behaviour and OralDrugimitate represents the oral drug poisoning rate associated with imitative behaviour.
Because the oral drug poisoning rate is related to the exposure of children to oral drug poisoning agents, the exposure-adjusted poisoning rate can be written as:
Children also have access to potentially toxic non-drug substances including household cleaning products, personal care products and others, such as those described in table 2. However, because children do not generally see adults ingesting these products, the unintentional non-drug poisoning rate is assumed to involve only general exploratory behaviour (NonDrugexplore) which, when adjusted for exposure to non-drug products (ExposureND), can be characterised as NonDrugexplore/ExposureND.
Because the mechanisms that underlie general exploratory behaviours leading to oral drug and non-drug poisonings are similar (ie, children see interesting objects in their environment and attempt to explore them by means of mouthing behaviour), it seems reasonable to assume that:
Thus, the exposure-adjusted oral drug poisoning rate associated with imitative behaviour should be approximately equal to the difference between the exposure-adjusted oral drug poisoning rate and the exposure-adjusted non-drug poisoning rate.
In the special case where the risk exposure to oral drugs is the same as the risk exposure to other non-drug products (ie, ExposureOD = ExposureND), the poisoning rate associated with imitative behaviour would equal the difference between the oral drug and non-drug poisoning rates (ie, OralDrugimitate = OralDrugtotal − NonDrugexplore). If, however, ExposureOD > ExposureND, then the poisoning rate attributable to imitative behaviour would have been somewhat lower—that is, if the risk exposure is relatively higher for oral drugs, a greater proportion of oral drug poisonings would have resulted from general exploratory behaviour, implying that fewer resulted from imitative behaviour. Conversely, if ExposureOD < ExposureND, then the poisoning rate attributable to imitative behaviour would have been somewhat higher.
References
Footnotes
Disclaimer The views expressed in this article are those of the authors. It has not been reviewed or approved by and may not necessarily reflect the views of the US Consumer Product Safety Commission. Because this article was written in the authors' official capacities, it is in the public domain and may be copied freely.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.