The Effect of Heat Waves on Mental Health in a Temperate Australian City

1Compared with the general population, persons with mental health problems often experience poorer overall health with higher rates of morbidity and mortality. Additionally, it has been well documented that, because of behavioral issues and medications that interfere with physiological homeostasis, those with mental illness are susceptible to the effects of extreme heat, as demonstrated by increases in hospital admissions and mortalities associated with heat waves. However, few studies have characterized specific mental and behavioral disorders (MBDs) that may be exacerbated by high temperatures.

2Mental illnesses may range from short-term bouts of depression and anxiety to long-term conditions such as developmental impairments, chronic depression, schizophrenia, or chronic anxiety disorders. Accounting for an estimated 13% of the national disease burden in Australia, mental illness is among the 10 leading causes of disease in 2003 and has been declared a National Health Priority Area. In the State of South Australia, studies have shown that as many as one in five adults has a mental problem.

3Our aim in this study was to identify mental disorders that contribute to heat-related morbidity and mortality in a temperate climatic region. This study is unique in that to our knowledge, no similar investigation has previously been undertaken. Moreover, we used both hospital admissions and mortality data spanning a 13-year period to establish a comprehensive overview of the temperature–health association. With mental disorders causing an already significant burden on the public health system, understanding the relationship between hot-weather extremes and psychiatric illness will assist in identifying populations at risk as global warming ensues, and provide valuable information for decision makers in the mental health and social service sectors.

4The city of Adelaide, with a population of 1.16 million, is the capital of the State of South Australia and is situated near the coast in the temperate region of southern Australia. The city’s climatic conditions include mild winters and long, hot, dry summers during which heat waves are common. Temperatures exceeding 40°C occur on average 3 days/year, making the climate ideal for the study of the health effects of heat waves.

5We obtained morbidity data for the Adelaide metropolitan area for the period 1 July 1993 to 30 June 2006 from the South Australian Department of Health (Adelaide, South Australia, Australia). We accessed principal hospital discharge diagnoses using the Integrated South Australian Activity Collection (ISAAC), an official collection of admitted patient activity in the state’s public and private hospitals. Data relating to individuals who resided outside of the Adelaide metropolitan area were excluded from the study. We obtained mortality data for the period 1 July 1993 to 22 December 2004 from the Australian Bureau of Statistics. Underlying causes of death were coded by the Australian Bureau of Statistics using computer-assisted coding.

6Daily counts of admissions and mortalities with principal discharge diagnoses or cause of death, respectively, attributed to MBDs were accessed. MBDs were defined in accordance with the International Classification of Diseases, 10th Revision [ICD-10; ]. Subclassifications of MBDs (ICD-10 codes F00–F99) included in the study are shown in Table 1. Dementia was identified separately, although it is already included within the classification of organic, including symptomatic, mental disorders. Additional to the aforementioned disorders, we included Alzheimer’s disease, senile degeneration of the brain, and senility to account for age-related cognitive disorders not captured within the F00–F99 subclassifications.

7We obtained climatic data from the Australian Bureau of Meteorology, Adelaide (Kent Town). Daily maximum ambient air temperatures for Adelaide were accessed from a central city weather station considered to be representative of conditions across the total metropolitan area, as advised by the Bureau of Meteorology.

8We analyzed hospital admissions and mortality data collectively and with stratification by age and sex. Seasons were defined as cool (1 April to 30 September) and warm (1 October to 31 March). Because the focus of the study was heat-related morbidity and mortality, we used only the warm season in the analysis. We defined heat waves as ≥3 consecutive days when the daily maximum temperature (Tmax) reached or exceeded 35°C, as in similar studies.

9We used conditional fixed-effects Poisson regression models to quantify the association with heat waves, with the referent period being all non–heat-wave days during the warm season. A goodness-of-fit test was applied to each model, and if significant overdispersion was detected, we used negative binomial maximum-likelihood regression. By performing within-year analyses, we adjusted for long-term trends. Results are expressed as incidence rate ratios (IRRs) with 95% confidence intervals (CIs).

10We explored the relationship between daily admissions or mortalities and Tmax across both seasons graphically using a lowess smoother, which performs locally weighted regressions of the y variable on the x variable to calculate smoothed data-point values. We determined the existence of a threshold by inspecting the plots; if a threshold was present, it was estimated visually and confirmed quantitatively using a nonlinear least-squares estimation incorporating the nl function in Stata, version 9.2. Based on the assumption of more than one segment in the relationship between two variables, this “hockey stick” method uses a function evaluator program to determine the cut point in the curve at which the change in slope occurs, providing this value is common to both functions. Parameters are determined using informative initial estimates of the slope and threshold values in iterative nonlinear regression models. Calculated threshold values should approximate visual estimates and have a statistical significance level of 0.05.

11The threshold model can be defined as follows:

12, where E(Y) is the expected value of the outcome variable, β0 is the baseline number of daily admissions, x0 is the threshold temperature value to be determined, β1 is the slope of the segment of the regression line prior to the changepoint; β0 is the slope after the change-point. If a nonzero slope is expected in Equation 1, the model becomes

13Hospital admission data spanned the period 1 July 1993 to 30 June 2006, and mortality data were obtained for 1 July 1993 to 22 December 2004. During this time, there were 31 and 29 heat waves, respectively, ranging in duration from 3 to 8 days, including an extreme heat event in January 2006 with temperatures over 4 continuous days exceeding 40°C. The highest recorded maximum temperature was 44.3°C during an 8-day heat wave in 2004.

14Regression results showed that 7 of the 15 categories of mental and cognitive disorders studied showed no significant association of hospital admissions with heat waves (Table 1, Figure 1); however, 2 of these 7 showed age- or sex-specific associations (Table 2). Therefore, 5 categories had no significant association of hospital admissions with heat waves: schizophrenia, schizotypal, and delusional disorders (ICD-10 codes F20–F29); disorders of adult personality and behavior (F60–F69); mental retardation (F70–F79); Alzheimer’s disease (G30–G30.9); and senile degeneration of the brain not elsewhere classified (G31.1). Behavioral and emotional disorders with onset usually occurring in childhood and adolescence (F90–F98) showed a statistically significant decrease in admissions during heat waves compared with non–heat-wave periods. Increases in hospital admissions for the remaining categories were found to occur during heat-wave periods, as discussed below.

15During the study period, a total of 171,614 admissions to Adelaide hospitals occurred with the principal discharge diagnoses of MBDs (F00–F99). Of these, 4,629 admissions occurred during heat waves, representing a daily mean of 38.6 admissions compared with 35.8 during the warm season. We observed the exposure–response association between admissions and Tmax graphically as a small rise and fall in admissions in cooler temperatures, and then a logarithmic trend as temperatures increased from moderate to extreme (Figure 2A). Using a nonlinear least-squares (hockey stick) estimation, we calculated a threshold temperature of 26.7°C (p = 0.001), above which MBD admissions increased markedly. Using iterative regressions, this threshold value remained robust to several parameter estimations.

16Overall, fixed-effects Poisson regression analysis accounting for overdispersion showed a 7.3% increase in admissions during heat-wave periods compared with non–heat-wave periods in the warm season (Table 1). Results were also significant for persons ≥ 75 years of age, with an IRR of 1.183 (95% CI, 1.088–1.286), and in male admissions in the 15- to 64-year age group (Table 2).

17In this category, encompassing dementia, cerebral disease, brain injury, and other trauma leading to cerebral dysfunction, dementia (F00–F03) accounted for more than half (56.2%) of the admissions. Compared with control periods, we found that hospitalizations for organic, including symptomatic, mental disorders increased by 21.3% during heat waves (Table 1) with increases in both male and female admissions (Table 2) and in the ≥ 75-year age group. Figure 2B shows a decline in admissions associated with moderate temperatures and a linear increase as temperatures rise from moderate to high.

18The risk of admission for dementia during heat waves increased overall by 17.4% (Table 1) and in females (Table 2). Because dementia accounted for a high proportion of organic, including symptomatic, mental disorder admissions, the shape of the relationship with Tmax (Figure 2C) is almost identical to that in Figure 2B, albeit on a different scale.

19In this category, the elderly (≥ 75 years of age) were the only age group to show an increase in hospitalizations during heat waves (IRR 1.567; 95% CI, 1.002–2.450), with female admissions of this age also showing a significant increase (Table 2).

20Mood (affective) disorders include depression, dysthymia, mania, and bipolar affective disorders. This category accounted for the largest proportion (33.9%) of MBD admissions. During heat waves, admissions increased 9.1% compared with referent periods (Table 1), with highest estimates in persons 15–64 years of age (IRR 1.102; 95% CI, 1.041–1.167).

21These disorders include anxiety disorders, panic disorder, agoraphobia, obsessive compulsive disorder, and posttraumatic stress disorder. Hospitalizations increased 9.7% compared with referent periods (Table 1), and age-specific regression showed that those most at risk during heat waves were the very young and the elderly. Figure 2D shows the relationship of admissions with temperature and demonstrates a gradual increase as Tmax rises above moderate.

22This group of syndromes includes eating and sleep disorders and accounted for only 1.6% of total MBD admissions over the study period. During heat waves, admissions increased for this classification in elderly females only (Table 2).

23Encompassing autism and developmental disorders of speech and language, this category accounted for 0.35% of MBD admissions. We observed a 64% increase in hospitalizations overall (Table 1), whereas the 15- to 64-year age group showed a > 3-fold increase (IRR 3.187; 95% CI, 1.185–8.571) during heat waves. Figure 2E shows admissions reaching a plateau at moderate temperatures and increasing sharply in hot weather.

24Senility includes senescence, asthenia, and debility, relating to mental infirmity and physical deterioration in the elderly. Although few admissions (n = 191) had the principal diagnosis of senility, our analysis showed a > 2-fold increase during heat waves (Table 1), with the strongest effect in the elderly (IRR 2.534; 95% CI, 1.232–5.211), and in elderly females (Table 2). Figure 2F reveals an almost U-shaped relationship, showing admissions at their lowest during mild temperatures and highest during the extremes of cold and hot weather.

25Mortalities attributed to the 15 categories of psychiatric disorders were relatively infrequent in Adelaide, and sample sizes during heat waves were not large. Table 3 displays regression results for the disorders showing significant increases in mortalities.

26Over the study period, 2,599 deaths were attributed to MBDs, representing < 1/day. During heat waves, there were 70 MBD deaths, with 81% (n = 57) of the decedents ≥65 years of age and regression results showing a 2.4-fold increase in deaths in the 65- to 74-year age group (Table 3).

27Most MBD deaths (76%, n = 1,964) were due to organic, including symptomatic, mental disorders, of which dementia comprised 99% (n = 1,953). Of the deaths occurring during heat waves, we classified 67% (n = 47) as organic, including symptomatic, mental disorders, and all were due to dementia. Regression results showed that the risk of death during heat waves compared with non–heat-wave periods was significant for those 15–64 years of age and in males of this age group (Table 3).

28During heat waves, deaths in this category increased only in females, specifically those in the 15- to 64-year age bracket (Table 3).

29Analysis of nonstratified data showed deaths due to schizophrenia and schizotypal and delusional disorders increased > 2-fold during heat-wave periods (Table 3). This was more pronounced in those ≥ 75 years of age and in males of this age group. Exploratory graphical analysis showed mortality to be lowest at moderate temperatures and highest when temperatures exceeded 35°C (data not shown), although low counts may have compromised the integrity of the association.

30This study builds on our previous investigation into the acute effects of heat waves, where among the range of diagnoses examined, mental disorders was the only group in which we observed increases in both hospital admissions and (age-specific) mortality in this temperate city. Other published literature has also identified persons with mental illness as being a heat-susceptible subgroup of the population. However, to our knowledge, this study is the first endeavor to comprehensively characterize specific disorders that contribute to increased psychiatric morbidity and mortality during heat waves. Our results suggest that several mental illnesses may be sensitive to exposure to high ambient temperatures. We have shown that hospital admissions for MBDs increased at temperatures above 26.7°C, and they also increased by 7.3% during heat waves compared with control periods. Specific nosologic subgroups for which an increase in admissions was evident include organic, including symptomatic, mental disorders; dementia; mood (affective) disorders; neurotic, stress-related, and somatoform disorders; and disorders of psychological development. Although classified separately to MBDs, senility was included in this study because a cognitive impairment of the elderly and admissions for senility increased > 2-fold during heat-wave periods. Hospitalizations for behavioral syndromes associated with physiological disturbances and physical factors, and disorders due to psychoactive substance use, increased for specific age and sex categories.

31Evidence has shown that having a pre-existing psychiatric illness can more than triple the risk of death during a heat wave. In the present study, we observed an increase in deaths classified as MBDs in the 65- to 74-year age group during heat waves. Additionally, deaths due to schizophrenia and schizotypal and delusional disorders increased > 2-fold, and those due to psychoactive substance use increased in all females and in females 15–64 years of age. Dementia deaths increased in the 15- to 64-year age group only, indicating the susceptibility of persons with early-onset dementia. Because our mortality data was limited by small sample sizes in most categories, the association between heat waves and mental health mortalities warrants cautious interpretation of some results, particularly where stratification of data compromised statistical power.

32Two issues concern the effects of heat on psychiatric illness: first, the nature of the psychiatric condition can be a risk factor for heat-related morbidity and mortality, and second, heat can be a risk factor for the exacerbation of the condition. Because heat-related illness is often underdiagnosed, individuals in our study with a case history of mental illness may have been suffering the effects of heat rather than (or as well as) the effects of their disorder, the latter taking likely precedence for classification purposes.

33That mental illness can be a risk factor for heat-related morbidity and mortality is plausible for physiological and behavioral reasons. Many medications used in psychiatry increase vulnerability to heat-related morbidity by altering the body’s ability to thermoregulate because of pharmacologic effects on the parasympathetic pathway. Drugs such as anti-psychotics, anticholinergics, antidepressants, sedatives, and mood stabilizers that impair sweating and/or increase heat production are used in the treatment of such conditions as dementia, Alzheimer’s disease, psychosis, mood disorders, personality disorders, and anxiety disorders. The inherent nature of these illnesses may also contribute to vulnerability during hot weather. Cognitive awareness of environmental conditions and the ability to undertake adaptive behaviors such as increased fluid intake or wearing appropriate clothing are important coping mechanisms that may be compromised in those with disabling mental illnesses such as Alzheimer’s disease, dementia, senility, psychosis, schizophrenia, and developmental disabilities. Additionally, severe or core activity limitations often presenting in the mentally ill can determine their degree of dependence and be a contributing factor to heat susceptibility. This notion is supported by our findings of increased admissions for individuals with senility, confirming the susceptibly of the frail elderly to heat-related illnesses or death.

34In contrast to the findings of others who reported a correlation of schizophrenia admissions with environmental temperature, we found no increase in admissions during heat waves for the nonstratified or age- and sex-stratified data. However, we observed a significant increase in mortalities attributed to schizophrenia and schizotypal and delusional disorders. A coroner’s study of 18 heat-related deaths during the 1999 Cincinnati heat wave also found schizophrenia to be a risk factor for mortality, with eight decedents having a mental illness, four of whom had schizophrenia. Assumption as to the nature of our varying morbidity and mortality rates in this category remains highly speculative. First, antipsychotic medications that impair thermoregulation may increase the risk of death in psychiatric patients during heat waves, although studies have shown patients were at risk in the 1950s before the introduction of antipsychotic drugs, suggesting that both schizophrenia and its medication increase the risk of heat-related death. Second, it is possible that, for these individuals, death may have occurred either rapidly before admission to hospital or under conditions of social isolation. Furthermore, consequential factors of living with such a debilitating mental illness include socioeconomic deprivation, solitary lifestyle, poor general health, and chronic illness, all of which are risk factors for heat-related mortality.

35In addition to schizophrenia, the ICD-10 category F20–F29 includes delusional disorders (codes F22.0–F22.9), a subcategory not investigated independently here. Altered mental status can be a symptom of heatstroke, heat stress, or hyperthermia, often presenting as delirium or disorientation. Heatstroke is common in the elderly, has a high fatality rate because of multiorgan dysfunction, and often goes either undiagnosed or not recorded as the underlying cause of death. Our results showed that eight of the nine patients whose deaths during heat waves were attributed to schizophrenia and schizotypal and delusional disorders were ≥ 75 years of age. That our results indicated more than twice the relative risk of death for these elderly persons during heat waves compared with control periods raises the possibility that they may have had delusional disorders rather than schizophrenia reported as the underlying cause of death. However, to avoid speculation, further investigation would be required at an individual rather than population level of data.

36A considerable body of literature has shown that dementia and cognitive impairment confer vulnerability to extreme heat. Psychotropic drugs with thermoregulatory side effects are commonly administered to dementia patients and may additionally modify vigilance levels and cognitive awareness of the need for adaptive behaviors to avoid thermal stress. Use of these medications has been shown to be associated with a significantly increased risk of heat-related hospitalization and death during heat waves. Our results showed an overall increase in hospital admissions for dementia during heat waves and an increase in dementia deaths in the 15- to 64-year age group. Although this latter finding was surprising, it is possible that older persons with dementia are more likely to reside in elderly care facilities, thereby lowering their risk of death during heat waves. With an estimated 231% rise in new dementia cases per year by 2050, these findings indicating that dementia may be a heat-sensitive illness suggest that the future burden on the health sector will be substantial.

37Evidence has shown that elevated temperatures may exacerbate psychiatric conditions. Environmental changes can affect mental health, with excessive heat and humidity reported to be major influences affecting mood and behavior. Fluctuations in weather have been noted to cause an increase in the incidence of mental stress, depression, and suicide. As temperatures rise to extreme, stresses of everyday home, social, or work life are likely to be compounded by lethargy, lack of sleep, and the inability to function normally during oppressively hot conditions. For those predisposed to acute or chronic mental problems, failure to gain relief from the heat for extended periods of time may trigger irritability and episodic psychological distress, accompanied by risk behaviors such as excess alcohol consumption, violence, and aggression. With heat waves predicted to occur more frequently, an increase in mental problems may be an indirect public health consequence of a warming climate. The effects of sustained heat and humidity, accompanied by drought, water restrictions, bushfires, and power outages (Essential Services Commission of South Australia 2006), are likely to have marked effects on the mental health of both rural and urban communities, with possible increases in the incidence of episodic or chronic stress, despair and depression, and health-damaging personal behaviors. This could have significance across the public health sector and implications for law enforcement and social cohesion.

38Statistics indicate an increase in reported mental illness in Australia, with 11% of persons reporting a long-term mental or behavioral problem in 2004–2005, up from 5.9% in 1995. Furthermore, as a consequence of the demographic shift, age-related mental impairments will be more common in years to come, placing financial strains on the economy. Recent figures show that expenditures on mental health services (excluding spending on Alzheimer’s and other nervous system disorders) totaled AU$3.0 billion in 2000–2001. To mitigate the effect of heat on the vulnerable, including the mentally ill, relevant health promotion and health intervention strategies should be considered, including the formulation of heat-wave response plans incorporating heat alerts to the public and health authorities when excessively hot conditions are expected. Similar heat-health action plans are now in place in most European cities after the health tragedy of the 2003 heat wave.

39Although population acclimatization to a warmer climate will undoubtedly occur and consequently lessen the impact of heat waves, adaptation strategies may, for behavioral and physiologic reasons, be less achievable in persons with mental illnesses. Our study design prevented us from examining possible adaptation over the study period or any potential differences in health impacts between early-and late-season heat waves. The assumption of nonvarying population response to heat over the study period may be one of several limitations of the study.

40Furthermore, our data assume correct ICD coding for MBDs. Errors in diagnostic classification would influence accuracy of effect estimates, particularly in cases of acute rather than chronic psychiatric illness. Substantial changes that occurred in the classification of diagnostic criteria for mental disorders in the transition from ICD-9 to ICD-10 may have increased the risk of misclassification by coders. For example, in the ICD-10, the classification of “dementia in Alzheimer’s disease” (F00–F00.9) is not listed in the same chapter as “Alzheimer’s disease” (G30–G30.9), highlighting the ease with which inaccuracies in coding can occur.

41Second, we acknowledge the small sample size in some categories, which limits statistical power. Further, inflation of type 1 error may arise when multiple comparisons are undertaken as in our study. Accordingly, we advise cautious interpretation of some results, especially mortality results.

42Finally, in an ecological study such as ours, it is not possible to determine whether those with a mental or behavioral disorder were admitted or died because of the effects of heat, or an increase in severity of their mental condition. Moreover, persons with psychiatric illnesses are more likely to have preexisting comorbidities and live in financial hardship and social isolation, risk factors we could not account for in this study. Further research using clinical records would reveal medication history and prognostic outcomes, thereby adding additional insights into the susceptibility of the mentally ill to the effects of heat.

43The present study has shown that heat waves pose a salient risk to the well-being of those with mental and cognitive disorders. The extent of the future risk will be largely determined by planning and mitigation. With heat extremes very likely to occur more frequently in countries around the globe, our findings have relevance on a local and international scale. Improvements in the management and care of the mentally ill will need to be addressed to avoid major economic and social costs to society of heat-related psychiatric morbidity and mortality exacerbated by climate change.