Acute impacts of extreme temperature exposure on emergency room admissions related to mental and behavior disorders in Toronto, Canada

Abstract

Background

The purpose of this study was to assess the effects of extreme ambient temperature on hospital emergency room visits (ER) related to mental and behavioral illnesses in Toronto, Canada.

Methods

A time series study was conducted using health and climatic data from 2002 to 2010 in Toronto, Canada. Relative risks (RRs) for increases in emergency room (ER) visits were estimated for specific mental and behavioral diseases (MBD) after exposure to hot and cold temperatures while using the 50th percentile of the daily mean temperature as reference. Poisson regression models using a distributed lag non-linear model (DLNM) were used. We adjusted for the effects of seasonality, humidity, day-of-the-week and outdoor air pollutants.

Results

We found a strong association between MBD ER visits and mean daily temperature at 28 °C. The association was strongest within a period of 0–4 days for exposure to hot temperatures. A 29% (RR=1.29, 95% CI 1.09–1.53) increase in MBD ER vists was observed over a cumulative period of 7 days after exposure to high ambient temperature (99th percentile vs. 50th percentile). Similar associations were reported for schizophrenia, mood, and neurotic disorers. No significant associations with cold temperatures were reported.

Limitations

The ecological nature and the fact that only one city was investigated.

Conclusions

Our findings suggest that extreme temperature poses a risk to the health and wellbeing for individuals with mental and behavior illnesses. Patient management and education may need to be improved as extreme temperatures may become more prevalent with climate change.

Introduction

Ambient temperature has long been suspected to play a role in the psychotic exacerbation of core symptoms for many specific mental and behavior disorders (McMichael et al., 2006). Individuals with mental illness are often susceptible to the effects of extreme temperature due the disruption of normal thermoregulation from psychotropic medication usage, psychiatric illness, and various behavior symptoms (Hansen et al., 2008, Shiloh et al., 2005). Several studies have demonstrated increases in hospital admissions and emergency room (ER) visits during periods of elevated temperature for individuals with mental illness in Australia (Nitschke et al., 2007, Nitschke et al., 2011, Wang et al., 2012, Khalaj et al., 2010), United States (Semenza et al., 1999, Semenza et al., 1996, Kaiser et al., 2001), Taiwan (Sung et al., 2011), Canada (Vida et al., 2012) and England and Wales (Page et al., 2012).

However, there is limited evidence regarding specific mental and behavior disorders that contribute to increases in ER visits (e.g. mood disorders, schizophrenia, substance abuse, etc.). In addition, given that the effects of ambient temperature on human health is distributed over a period of several days to over 1 week (Ye et al., 2012), the lagged structure of the effects of ambient temperature on mental and behavior illnesses remains unknown.

The aim of our study was to investigate the relationship between ambient temperature and mental and behavior illness related emergency room admissions in the City of Toronto, Canada. Toronto is located in the southern part of the province of Ontario. It is the largest city in Canada and the provincial capital of Ontario with approximately 2.7 million residents (Statistics Canada 2011). The city experiences a wide range of hot and cold temperature throughout the year with daily mean temperature ranging from −20.3 to 31.3 °C as reported by Environment Canada. How extreme temperature (hot and cold) can exacerbate specific mental illnesses in Toronto and the lagged effects of such associations are of interest.

The study population was comprised of the residents of Toronto from April 1st 2002 to March 31st 2010. The study population was identified using relevant postal codes from a provincial computerized database for tracking emergency room admissions: NACRS (National Ambulatory Care Reporting System). NACRS captures over 97% of the ER visits in the province of Ontario and has good reabstraction accuracy (Canadian Institute of Health Information, 2012). Access to the morbidity data was granted through data sharing agreement between CIHI and the Public Health agency of Canada (PHAC).

The daily number of emergency room (ER) visits was selected as the outcome metric to represent the acute effect of ambient temperature on mental illness morbidity. This was done after reviewing the current literature between mental illness and exposure to ambient temperature. Daily ER visit counts were obtained from the NACRS for the city of Toronto from April 1st 2002 to March 31st 2010. ER visit records were included in the counts if the primary reason of the visit was classified under the ICD-10 (International Classification of Diseases, 10th Revision) related to mental and behavior disorders (F00–F99). We also investigated ambient temperature effects for specific diseases: Schizophrenia (F20–F29), Mood Disorders (F30–F39), Neurotic Disorders (F40–F48), and Mental and Behavior disorders due to Psychoactive Substance use (F10–F19).

Hourly measurements of temperature and humidity for the city of Toronto from April 1st 2002 to March 31st 2010 were obtained from Environment Canada using the monitoring station at Toronto Pearson International Airport (latitude:43°40′36″N; longitude: 79°37′50″W) located approximately 26 km west of downtown Toronto. Daily mean, minimum and maximum temperatures and mean relative humidity were computed. Air pollution data were obtained from the National Air Pollution System (NAPS). It comprised of hourly measurements of gaseous pollutants: sulfur dioxide (SO₂), nitrogen dioxide (NO₂), carbon monoxide (CO), and ozone (O₃). The air pollutants were measured using UV Fluorescence (SO₂), chemiluminescence (NO₂), infrared gas correlation (CO) and UV absorption (O₃) from several fixed monitoring stations within the city of Toronto. Mean daily concentrations of each pollutant were computed by taking daily averages and then averaging across multiple stations for each day. Daily means of PM_2.5 (particulate matter less than 2.5 μm in aerodynamic diameter) were computed in the same method and were measured using a combination of tapered element oscillating microbalances, gravimetric filters, virtual impactors, and beta radiation attenuation (Allen., 2010).

Daily mean temperature was used as the exposure metric after model selection with several other indices of temperature. Its association with the daily number of emergency room (ER) visits related to mental and behavior disorders was assessed using a time series study design. Given that the relationship between adverse health outcomes and ambient temperature is lagged across specific timeframes and best described as a non-linear curve (Ye et al., 2012; Turner et al., 2012; Basu and Samet 2002), we employed the use of distributed lag non-linear models (DLNM). The DLNM is built on the foundation of combining two functions (temperature and lag) into a “cross-basis”, a bi-dimensional matrix which allows the estimation of possible non-linear temperature and morbidity effects across specific lag periods (Gasparrini and Armstrong, 2010, Gasparrini, 2011). In the context of our study, the DLNM was used to estimate the relative increase in ER visits from exposure to daily mean temperature on mental and behavior disorders and specific diseases: schizophrenia, substance abuse related mental and behavior disorders, mood disorders, and neurotic disorders for both hot and cold temperatures. For extreme hot temperature, the 99th percentile of the daily mean temperature distribution throughout the study period was used as the exposure metric and the 1st percentile of the daily mean temperature was used to represent exposure to extreme cold temperature. The 50th percentile of the temperature distribution was used as the reference point. A quasi Poisson regression was fitted using generalized linear modeling (GLM) to model the daily counts of emergency room visits based on the predictor variables: daily mean temperature and lag. The degrees of freedom for mean temperature and lag were set to 5. The lag period for the DLNM was set to 30 days. To account for the effects of season, a smoother for time of 10 degrees of freedom per annum was included. Also, a categorical variable was used for each day of the week. In addition, we performed sensitivity analyses by using varying degrees of freedom for the time smoother (6, 7, and 8) in order to compare different scenarios of smoothing. The quasi Akaike's information criterion (QAIC) was used to measure the relative goodness of fit for our statistical models. A lower QAIC value indicates an improvement in model fit (Gasparrini, 2011).

We adjusted for the confounding effects of air pollution by including mean daily concentrations of NO₂, CO and O₃ as covariates in our analysis. Other pollutants such as SO₂ and PM_2.5 were excluded during the process of model selection. Based on previous work, the effects of these pollutants were adjusted for lag periods of 0–2 days (Goldberg et al., 2011, Ye et al., 2012, Gosling et al., 2008). In addition, we accounted for humidity by using the daily mean concentration as a covariate. Finally, we adjusted for the effects of age by examining the association between ambient temperature and mental illness related ER visits in specific age groups (0–14, 15–39, 40–59, 60+). Gender was evaluated using the same approach.