Extreme heat and paediatric emergency department visits in Southwestern Ontario

Abstract

Objective

The risk of adverse health events is expected to increase with hotter temperatures, particularly among the most vulnerable groups such as elderly persons and children. The objective of this study was to assess the association between extreme heat and daily emergency department visits among children (0 to 17 years) in Southwestern Ontario.

Methods

We examined the average maximum temperature, relative humidity, and daily paediatric emergency department visits in June through August of 2002 to 2019. We reviewed emergency department visits from two academic hospitals. Daily meteorological data from the local weather station were obtained from Environment and Climate Change Canada.

Results

Extreme heat, defined as the 99th percentile of the maximum temperature distribution, occurred at 33.1°C and was associated with an overall 22% increase in emergency department visits, compared to the reference temperature of 21°C. This association was mostly found between the second and fifth day after the exposure, suggesting a slightly delayed effect. The results of the sub-group analysis indicate that the risk of an emergency department visit due to infectious disease increases by 35% and the most pronounced association was noted in children aged 1 to 12 years.

Conclusions

Extreme heat is associated with an increased incidence of emergency department visits in children. As temperatures continue to increase, strategies to mitigate heat-related health risks among children should be developed.

Climate change is anticipated to lead to increased temperature variability and temperature extremes, with negative effects on health (1–4). Temperature ‘extremes’ result from increasing average global ambient temperatures, an increase in the frequency and severity of heat waves (a prolonged period of excessively hot weather relative to typical temperature in a particular location), and increased pollution and exposure to ultraviolet radiation (5). Extreme heat events usually occur during a heat wave; on days with extreme heat, temperature significantly exceeds the average temperature for the whole period of heat wave (6). Existing evidence indicates that extreme heat adversely affects health (7,8), particularly among elderly persons (8,9) and children (5,10–12). Children less than 5 years of age (9,10,13) may be particularly vulnerable to the adverse effects of high temperatures (13,14), due to their unique physiological, metabolic, and behavioral characteristics (15–17). High ambient temperatures have been linked with infectious diseases in children including gastrointestinal diseases, hand, foot and mouth disease, and respiratory diseases (14). Furthermore, the incidence of renal disease, fever, and electrolyte imbalance increases significantly during heat waves (14). Extreme temperatures are associated with increased emergency department (ED) visits among children (10,18,19). This increase has a lag of up to 3 days in infants, while toddlers and preschoolers are more likely to present to the ED on the day of the extreme temperature event (10).

Currently, there is limited evidence on how extreme heat can affect the health of children in Canada, including their use of ED services. Such evidence, however, is needed to monitor and understand national, provincial, and local across-time changes in the effects of extreme temperatures on paediatric health, as well as to inform local hospital and public health programs and policies that aim to mitigate adverse health effects. It is likely that ED preparedness and population-level prevention strategies aiming to address adverse environmental effects will produce more widespread and lasting improvements compared to individual prevention strategies (20). Thus, the objective of this study was to assess the association between extreme heat and daily ED visits among the paediatric population in Southwestern Ontario.

METHODS

Setting

Southwestern Ontario covers 36,797 km² and occupies most of the Ontario Peninsula, from the Bruce Peninsula in the north to Lake Erie in the south and from east to west, it stretches from Guelph to Windsor (21). In 2016, the total population was 2,583,544 (a population density of 70/km²) (21), with 383,822 residing in the City of London (22).

ED visits

We reviewed the medical charts of paediatric ED visits (0 to 17 years) from Children’s Hospital and University Hospital (both part of the London Health Sciences Centre), for years 2002 through 2019. Children residing outside of Southwestern Ontario were excluded. Daily numbers of ED visits due to all causes excluding injuries (ICD 10: S00-T98) were derived for and stratified by age (<1, 1 to 12, and 13 to 17 years); injuries (34.2% of all ED visits during the study period) were excluded because their association with heat may not be explained by physiological mechanisms, but rather, by indirect mechanisms related to an increase in outdoor activities. Based on previous literature (e.g., Iniguez et al. [23]) and biological plausibility, cause-specific daily ED visit counts were divided into two major groups of diseases, namely respiratory (ICD10: J00-J99) and infectious (ICD10: A00-B99) diseases.

Environmental data

Daily meteorological data from the weather station located at London International Airport (about 10 km from both hospitals) were obtained from Environment and Climate Change Canada (ECCC). We calculated daily maximum and average temperatures (°C) and relative humidity (%) from hourly measurements. We obtained daily Air Quality Health Index (AQHI) data from ECCC, available for the period of 2012 through 2019, which represents the relative risk of a mixture of common air pollutants (Ozone at ground level, Particulate Matter and Nitrogen Dioxide) known to have adverse effects on health. The AQHI is measured on a scale ranging from 1 to 10+, with values grouped into health risk categories (1 to 3: low, 4 to 6: moderate, 7 to 10: high, and 10+: very high) (24).

Data analysis

We measured the association between daily ED visits and daily maximum temperature with conditional quasi-Poisson regression and distributed lags non-linear models (DLNMs) (25,26). As we were interested in measuring the association with heat, we restricted the study period to June through August. DLNM resembles the bi-directional time-stratified case-crossover design which controls for long-term and seasonal trends by matching case and control days within a year, month, and day of the week. We modelled the non-linear and delayed association with daily maximum temperature through the so-called cross-basis function with a natural cubic spline function in the exposure-response dimension with internal knots at the 50th and 95th percentiles of the temperature distribution. Furthermore, a natural cubic spline with two internal knots placed in equally spaced values in the log scale was used for the lag dimension accounting for up to 7 days of lag. We controlled for national holidays with an indicator variable. Model choices were defined using quasi-Akaike Information Criterion. The analysis was repeated across two types of ED visits (respiratory and infectious diseases). We reported relative risks (RRs) and 95% confidence intervals (CIs) at the 99th percentile of the daily maximum temperature distribution, using the temperature at the minimum risk of ED visits as reference.

Additional sensitivity analyses were performed to test the robustness of results. We first repeated the main analysis using daily average temperature as the main exposure, instead of daily maximum temperature. We then controlled for relative humidity in the model with a linear term of the moving average across the same day and day before, and, in a separate analysis, for the warmer period defined as May to September. Finally, using 2012 to 2019 data, we controlled for air quality levels using daily AQHI of the same day of exposure, also as a linear function. All analyses were completed in R version 3.6.2 (27).

Ethical and institutional approvals for the study were granted by the Western University Health Science Research Ethics Board and Lawson Health Research Institute.

Results

There were 95,454 ED visits during the study period, an average of 57.6 per day. Children aged 1 to 12 years accounted for 63.5% of all visits (60,618 visits), followed by children aged 13 to 17 (20.9%; 19,946 visits), and children below 1 year of age (15.6%; 14,890 visits). ED visits due to infectious and respiratory diseases amounted for 13.5% and 16.6% of ED visits, respectively. Daily maximum and average temperatures ranged from 13.1 °C to 36.7°C and from 19.9 °C to 30.7°C, with an average value for the whole period of 25.7°C and 20.1°C, respectively. The daily AQHI ranged from 1.1 to 4.3, with an average value of 2.3.

Figure 1 (left panel) shows the overall cumulative association between maximum temperature and total ED visits (7 days of lag). Extreme heat, defined as the 99th percentile of the daily maximum temperature distribution (33.1°C) (vertical dashed line), was associated with a 22% increase in ED visits (RR: 1.22; 95% CI: 1.12 to 1.32), compared to the reference temperature of 21.0°C. Figure 1 (right panel) shows the distribution of the heat–ED visit association across 7 days after exposure. We observed that the association was mostly found between the second and fifth day after exposure, suggesting a slightly delayed effect of heat on ED visits.

Figure 1.

Overall cumulative association between maximum temperature and all-cause emergency department visits (excluding injuries) across 7 days of lag (left panel) and lag-specific heat–emergency department visits associations (right panel). Extreme heat is defined as daily maximum temperatures equal to the 99th percentile (33.1°C). RR Relative risk; CI Confidence interval.

Figure 2 shows the corresponding extreme heat–ED visit association estimates by disease group and by age. The overall risk of ED visits due to infectious disease increased by 35% (RR: 1.35; 95% CI: 1.07 to 1.72) during the 7 days after a heat day, while no evidence of association was found for respiratory diseases. For all ED visits, larger association estimates were found in children aged 1 to 12 years (RR: 1.33; 95% CI: 1.20 to 1.47) than in other age groups (see Supplementary Table 1). A smaller and more imprecise association was estimated in older children (RR: 1.12; 95% CI: 0.94 to 1.32) and the association was absent in the younger age group. Similar patterns across age categories were observed for the two disease groups. Only the children aged 1 to 12 years had an increased risk in ED visits due to infectious diseases (RR: 1.51; 95% CI: 1.13 to 2.00). The results of the sensitivity analysis indicate that (1) using daily average temperature as the main exposure, (2) controlling for relative humidity, (3) using a longer study period, and (4) controlling for AQHI did not substantially affect the relationship between heat and ED visits (see Table 1).

Figure 2.

Extreme heat–emergency department visit association by groups of causes and age categories. Relative risk (RR) and 95% confidence interval (CI) at the 99th percentile of the daily maximum temperatures (33.1°C) vs the temperature of minimum risk of emergency department visits.

Table 1.

Extreme heat-related emergency department visit (all-cause, excluding injuries) association estimates expressed as relative risk (RR) and 95% confidence interval (CI) at the 99th percentile of the maximum daily temperatures (33.1°C) vs the temperature of minimum risk of emergency department visits

Type of analysis	Heat, RR (95% CI)
Main analysis (daily maximum temperature)	1.22 (1.12, 1.32)
Sensitivity analysis
1) Using daily average temperature	1.17 (1.07, 1.28)
2) Controlling for relative humidity	1.22 (1.12, 1.32)
3) Warm period defined as May–September	1.15 (1.07, 1.25)
4a) Limiting study period to 2012–2019*	1.17 (0.96, 1.42)
4b) Controlling for daily AQHI*	1.13 (0.92, 1.37)

AQHI Air Quality Health Index.

* Sensitivity analysis 4a and 4b reported for the period 2012–2019 when daily AQHI data were available.

DISCUSSION

Extreme heat was associated with a 22% increase in ED visits among children in Southwestern Ontario. This association was mostly found between the second and fifth day after the exposure, which suggests a slightly delayed effect. Children between the ages of 1 and 12 experienced most of the ED visits (63.5%). Moreover, the effects of extreme heat were most evident in this age group for all causes (excluding injury), as well as infectious diseases.

This is the first study to examine the relationship between extreme heat and paediatric ED use in the Canadian context, and the results parallel previous studies (10,28). The observed lag effects of exposure for our population, assessed to up to a week post-exposure, are similar to results found for children in the USA, in Atlanta (5 to 18 years old) (28) and in New York City (NYC; <4 years old) (10). The number of very hot days in summer in the study region is expected to rise over the upcoming decades; in fact, the Climate Atlas of Canada predicts the number of very hot days (>30°C) per year in the City of London, Ontario will increase from an average of 12 (1976 to 2005) to an average of 33 (2021 to 2050) (29). Hospitals should develop appropriate surge-capacity and emergency preparedness protocols and procedures.

Children of different ages appear to experience heat-related illnesses differently and this is demonstrated by variability in ED presentation rates between age groups, and potentially explains differences in the day peak lag-effect after exposure between age groups and across studies (10,28). It is possible that behavioural modifications that alter heat exposure by age may modify the heat–illness association, as well as the different prevalence of infections and respiratory diseases by age. Unlike the association observed in the NYC paper (10), we did not see a significant association between extreme heat and ED visits for children <1 year old. These differences may be a result of our smaller sample size, the differences in geography, and/or other social factors not accounted for in our study.

Considerable attention has been given to the observed association of respiratory diseases, particularly asthma, and temperature (28,30–32), which is affected by other factors such as humidity (31), air pollution, and aerosolized pollen (32). Other disease and illness processes may be related to temperature, but larger studies would be required to identify associations.

There are limitations to this study. This was an ecological study, and we did not assess the individual factors that might mitigate or amplify heat effects, such as household socioeconomic factors. Also, the small sample size may have affected the ability to detect associations in the <1 year and 13 to 17 age groups, as the daily ED visit rates for these groups are substantially lower than for the 1 to 12 age group. The low number of ED visits that could translate into model non-convergence or low statistical power did not allow us to consider other groups of diseases (e.g., cardiovascular). Finally, in the sensitivity analysis involving removal of the potential confounding effect of air quality (AQHI), meteorological data came from a single weather station located at London International Airport which could produce some degree of exposure misclassification or Berkson error bias that could affect the precision of the estimates (33), but not the direction or magnitude of the association between heat and ED visits. We also did not control for the effects of each of the single air pollutants that are part of the AQHI; however, regardless of the confounding mechanisms of each air pollutant, the bias derived from not accounting for them would have the same direction (i.e., the confounding effect of each single pollutant would not be outbalanced between each other).

Overall, this study provides insight on the consequences of extreme heat among the paediatric population. Further investigation at the regional and national level are important to better understand the true impact of heat on child health, improve monitoring and surveillance strategies, and target interventions that are most effective to mitigate health-related illness in children.

Funding: This research was supported by funding from the Canadian Institutes of Health Research (Operating Grant ID: 410547), and the Department of Paediatrics and the Children’s Health Research Institute (Translational Research Grant Fund).

Potential Conflicts of Interest: Outside the submitted work, KKC has received a research award sponsored in part by Astra Zeneca. She has attend Merck sponsored conferences. She has also received honoraria for delivering CME talks from Sutherland Global Services Canada YLC and the Toronto Ontario Knowledge Translation Working Group. There are no other disclosures. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.