Hot of Not: Physiological versus Meteorological Heatwaves—Support for a Mean Temperature Threshold

The aim of this study was to determine whether a revised heat warning threshold provides an enhanced predictive tool for increases in Emergency Department heat-related presentations in Canberra, Australia. All Emergency Department triage records containing the word “heat”, as well as those diagnosing a heat related illness for the summer periods 2013/2014, 2014/2015, and 2015/2016 were searched. Then a medical record review was conducted to confirm that the patient’s presentation was related to environmental heat, which was defined by the final clinical diagnosis, presentation complaint and details of the patient’s treatment. Researchers then compared this presentation data, to a mean threshold formula. The mean threshold formula included the past three consecutive daily mean temperatures and the last measured temperature upon presentation. This formula was designed to take into account the variance of night-time lows, with concurrent daily ambient temperatures, and was used to determine whether there was a correlation between heat-related presentations and increasing mean temperatures. Heat-related presentations appeared to occur when the mean threshold temperature reached 25 °C (77 °F), with significant increases when the mean threshold reached 30 °C (86 °F). These results confirm that a mean temperature of 30 °C corresponds to a relevant local public health heat-related threat.


Introduction
There has been significant research worldwide pertaining to the effects of heatwaves on local populations and individual countries' associated warning systems, particularly post critical events, such as the Chicago and European heatwaves of 1995 and 2003 [1,2]. The Australian literature covers topics including planning, prediction of events, alert/awareness systems, population awareness of risk, stakeholder engagement, Emergency Department impact, spatial analysis and morbidity and mortality [1,[3][4][5][6][7][8][9][10][11][12][13][14][15]. These topics generally follow the standard emergency management cycle: Preparedness, Response, Recovery, and Mitigation [16]. The existing literature appears limited, focusing mainly on large city populations with a coastal influence. There has been minimal literature addressing landlocked populations [8]. Table 1. 2011 heat-related deaths and estimated deaths per annum. Data indicates that as the population increases, so will the heat-related deaths. This could place healthcare under increased pressure. Source: PwC Australia [21]. Many health warning systems do not adequately account for mean daily temperatures and do not adequately reflect how heatwaves affect the community. Mean threshold triggers have been identified as being better suited to adequately predict increased heat related presentations [1,13,15]. Nicholls et al. proposed an alert tool based on a daily mean temperature threshold of 30˝C (86˝F) [1]. This tool aids in the distinction between meteorological and physiological heatwaves through the recognition of the human body's tolerance for environmental heat, tempered by lower overnight temperatures. Variables such as economic, social, cultural and health status cannot be captured in such a simple heat alert tool, though may be accounted for in a validated deviation in the mean temperature threshold for a specific population cohort. For example, if the population is generally healthy, has easy access to air-conditioned environments and is normally exposed to sustained higher local temperatures, then a higher daily mean threshold than 30˝C (86˝F) may be indicated.
The primary intent of this research is to broaden the literature associated with changing the current warning system, based on a threshold of three consecutive days of a temperature >35˝C (95˝F), to a mean 3-day temperature threshold of 30˝C (86˝F), taking into consideration night-time lows. The reasoning for this argued change in threshold, is associated with better preparing the Emergency Department to expect an increase in heat related presentations. The activation of the current system, of three consecutive maximum day temperatures above >35˝C (95˝F), have not resulted in any Emergency Department presentations related to heat. A mean 3-day temperature threshold, would allow better allocation of resources, increasing effectiveness while minimising the impact of heat related events. As such, the researchers aimed to test the following hypothesis: does a mean temperature of 30˝C (86˝F) correspond to an increase in heat related presentations, in the context of a socioeconomically stable inland population.

Study Design and Context
A retrospective chart review was performed at a metropolitan Emergency Department seeing an average of 56,000 patients a year (2013/2014, 2014/2015, 2015/2016), in Canberra, Australia. At the last census (June 2013), the population of the Australian Capital Territory stood at 383,400 [22]. This study was approved by the Hospital Research Ethics Committee, approval code: 09-2016.

Participants and Data Collection
The local Emergency Department Information System (EDIS) was used to retrieve all Emergency Department triage information that contained the word "heat", as well as those diagnosed with a heat related illness, for the summer periods (December, January, February) of 2013/2014, 2014/2015, and 2015/16. In reviewing the EDIS data, researchers recorded: - The patient's medical record number -Patient's home demographics (suburb or town) -Arrival time and date The last measured environmental temperature and the past 3-day mean temperature related to the presentation date, and was gathered from meteorological data for the period. This data was gained from the nearest weather observation station to the hospital: Queanbeyan Bowling Club (site number 070072); latitude 35.36˝S, longitude 149.23˝E. This was due to the closure of the Canberra Airport observation station in 2010.
Researchers determined whether the patient's presentation was related to heat through analysis of the final Emergency Department clinical diagnosis, the presentation complaint and the patient's treatment details. Valid heat related presentations included those highlighted in Table 2 (binary yes/no response), with all other presentations discounted from the study. To confirm whether the presentation was related to heat, against the digital record, researchers audited the patient's notes, gained via the hospital's medical records department, to determine the following: -Did the presentation have signs and symptoms related and conducive to heat, and -Did the signs and symptoms result in treatment for a heat related illness, and -Was the final diagnosis upon discharge related to heat The researchers recorded whether the patient presenting with a heat related complaint was within a high risk population (the very young (<2 years old), the elderly (>64 years old) and those with significant comorbidities). Allocation was determined if the patient adhered to the criteria highlighted in Tables 3 and 4. Patients only needed to adhere to one of these characteristics, to be considered a high risk patient.
Researchers used a mean threshold formula. This formula takes into account the variance in night time lows with concurrent ambient temperatures. The current measure was believed to reflect the premise: -That as temperature means progressively build over three days (heatwaves), patients are less likely to be able to cope, leading to increased hospital presentations [1,15,18], and -Reflect studies [13,23,24] from major Australian cities that reported an increased mortality or morbidity, in relation to high ambient temperatures (heat events) As such, researchers used the following formula: where TDA = Last three day temperature average; LMT = Last measured temperature (09:00 or 15:00) before patient presentation. This formula was modified from Nicholls et al. [1] to include a 3 day averaged mean temperature, as recommended by Nairn and Fawcett (2015) [18,25]. This formula was then adapted to account for the last patient presentation temperature. This was believed to take into consideration current ambient and consecutive temperatures [25]. The formula was then used to determine whether there was a correlation between patient heat presentations and increasing mean temperatures.

Results
The results indicate that the last measured meteorological temperature before patient presentation to the Emergency Department averaged 32˝C (89.6˝F). Heat related presentations appeared to increase when the mean threshold temperature reached 25˝C (77˝F), with significant increases when the mean threshold reached 30˝C (86˝F) (Table 5; Figure 1). A Chi-square test of independence was performed (based on results highlighted in Table 5) to compare the mean threshold temperature formula (TDA + LMT)/2), to that of an expected 3 day temperature threshold above 35˝C. These results were statistically significant (p = 0.01), and confirm the hypothesis that a mean temperature of 30˝C (86˝F) corresponds to an increase in heat related presentations, representative of a potential public health emergency thus an indicative threshold.  Table 6.
Seventy-eight (n = 78) patient records were identified as containing the word 'heat' within the presenting complaint or diagnosis, as a result of 168,000 Emergency Department triage presentations being assessed electronically. Subsequently, 42 participants were excluded (Table 7) from the study due to not being a valid heat related presentation ( Table 2). Expansion of the health services impact concept, to a mean temperature of 30 °C (86 °F) found that at one of the territories emergency departments there were 36 patients with a heat related diagnosis (Table 8).
All 36 patients with a heat related presentation, were in the high risk category (Tables 3 and 4). The total number of patients with a single risk factor equalled 23, with the remaining 10 patients having one or more heat risk factors. Multiple risk factors most commonly related to age, conducting strenuous outdoor activity and having limited access to air-conditioning ( Table 9). The top 3 discharge diagnosis included, dehydration (n = 10), general heat related illnesses (n = 9) and heat stress (n = 7). The remaining diagnosis (n = 10) being a combination, detailed in Table 8. The primary symptom of a heat related illness was patient syncope (n = 17), with nausea and/or vomiting (n = 7), at 47.2% and 19.4% respectively (Table 10).
Results indicated that patient sociodemographic information varied. Patient age ranged from a 6 month old female to a 91 year old male. The mean age was 40 years (a female mean age of 36 years, and a male mean age of 43 years). The majority of patients resided in the ACT (n = 31), with a minority from interstate (n = 5). Most of the ACT residents were from the northern aspects of the ACT (n = 29), with the remaining (n = 2) from southern ACT ( Figure 2). Of note, southern ACT has a large public    Table 6.
Seventy-eight (n = 78) patient records were identified as containing the word 'heat' within the presenting complaint or diagnosis, as a result of 168,000 Emergency Department triage presentations being assessed electronically. Subsequently, 42 participants were excluded (Table 7) from the study due to not being a valid heat related presentation ( Table 2). Expansion of the health services impact concept, to a mean temperature of 30˝C (86˝F) found that at one of the territories emergency departments there were 36 patients with a heat related diagnosis (Table 8).
All 36 patients with a heat related presentation, were in the high risk category (Tables 3 and 4). The total number of patients with a single risk factor equalled 23, with the remaining 10 patients having one or more heat risk factors. Multiple risk factors most commonly related to age, conducting strenuous outdoor activity and having limited access to air-conditioning ( Table 9). The top 3 discharge diagnosis included, dehydration (n = 10), general heat related illnesses (n = 9) and heat stress (n = 7). The remaining diagnosis (n = 10) being a combination, detailed in Table 8. The primary symptom of a heat related illness was patient syncope (n = 17), with nausea and/or vomiting (n = 7), at 47.2% and 19.4% respectively (Table 10).
Results indicated that patient sociodemographic information varied. Patient age ranged from a 6 month old female to a 91 year old male. The mean age was 40 years (a female mean age of 36 years, and a male mean age of 43 years). The majority of patients resided in the ACT (n = 31), with a minority from interstate (n = 5). Most of the ACT residents were from the northern aspects of the ACT (n = 29), with the remaining (n = 2) from southern ACT ( Figure 2). Of note, southern ACT has a large public hospital (The Canberra Hospital and Health Services), which provides healthcare for south residences, most likely explaining the predominance of northern aspect presentation.
There were no periods, during the observed period, where the mean daily temperature met or exceeded 30˝C (86˝F). Whilst there were multiple meteorological events where the predicted/actual maximum daily temperatures reach or exceeded 35˝C (95˝F), these were tempered by significantly lower overnight temperatures <22˝C (71.68˝F).  Table 7. Excluded presentations during review of the medical records. The diagnosis/prognosis often had heat similar symptoms, although were discounted as not being directly associated with temperature.

Discussion
Mean daily temperatures seem to provide a predictive tool. Table 5 and Figure 1 illustrate that mean daily temperatures exceeding 25 °C (77 °F) led to an increased likelihood of heat related presentations to the local Emergency Department, with a significant likelihood of presentations the closer the mean temperature was to reaching or exceeding 30 °C (86 °F). A mean temperature of 30 °C (86 °F) would thus appear to provide a useful threshold for issuing heat alerts in Canberra. This is consistent with previous research [1] and corresponds to alternate recommendations setting a Canberra threshold of 25.9 °C (78.62 °F) [18]. The results described could validate a simple method for public heat alert systems elsewhere, to be implemented regionally or within individual hospitals/healthcare facilities. The Australian Bureau of Meteorology provides online access to forecast weather trends, thus allowing public health clinicians the ability to determine the past 3-day average. This is in contrast to a study by Barnett, Tony, and Clements (2010) [27] indicating that different temperature indicators, including mean, minimum, and maximum, have the same predictive ability in heat related mortality. Specifically, the current study found no heat-related presentations associated with a consecutive max temperature >35 °C (95 °F), whereas a mean threshold was significantly associated with increased heat-related presentations. Barnett, Tony, and Clements (2010) [27] found large differences in the best measurement measure between age groups, seasons, and cities, and that there was no one temperature measure that was superior to others overall (nationally). They recommended that new studies should chose temperature measures, based on practical concerns, as related to the area and data availability [27]. This current study, adheres to this recommendation, by applying a threshold measure that relates to the Canberra climate and patient demographic.
Changing Canberra's activation process to a mean threshold, is reflective of other heat warning systems within other Australian states. This includes the 2015 Victorian Heat Health Plan, which details that a heat health warning should be communicated when the minimum and maximum average temperature is >30 °C (86 °F). This plan changed from using maximum thresholds, to basing its activation process on the Bureau of Meteorology's Heatwave Service for Australia definition of heatwave as "three days or more of high maximum and minimum temperature that are unusual for that location" [28]. The South Australian Government has also moved away from an activation process of three consecutive days >35 °C (95 °F), and have specifically redefined their activation trigger to three or more consecutive days with an average daily temperature of >32 °C (90 °F). This

Discussion
Mean daily temperatures seem to provide a predictive tool. Table 5 and Figure 1 illustrate that mean daily temperatures exceeding 25˝C (77˝F) led to an increased likelihood of heat related presentations to the local Emergency Department, with a significant likelihood of presentations the closer the mean temperature was to reaching or exceeding 30˝C (86˝F). A mean temperature of 30˝C (86˝F) would thus appear to provide a useful threshold for issuing heat alerts in Canberra. This is consistent with previous research [1] and corresponds to alternate recommendations setting a Canberra threshold of 25.9˝C (78.62˝F) [18]. The results described could validate a simple method for public heat alert systems elsewhere, to be implemented regionally or within individual hospitals/healthcare facilities. The Australian Bureau of Meteorology provides online access to forecast weather trends, thus allowing public health clinicians the ability to determine the past 3-day average. This is in contrast to a study by Barnett, Tony, and Clements (2010) [27] indicating that different temperature indicators, including mean, minimum, and maximum, have the same predictive ability in heat related mortality. Specifically, the current study found no heat-related presentations associated with a consecutive max temperature >35˝C (95˝F), whereas a mean threshold was significantly associated with increased heat-related presentations. Barnett, Tony, and Clements (2010) [27] found large differences in the best measurement measure between age groups, seasons, and cities, and that there was no one temperature measure that was superior to others overall (nationally). They recommended that new studies should chose temperature measures, based on practical concerns, as related to the area and data availability [27]. This current study, adheres to this recommendation, by applying a threshold measure that relates to the Canberra climate and patient demographic.
Changing Canberra's activation process to a mean threshold, is reflective of other heat warning systems within other Australian states. This includes the 2015 Victorian Heat Health Plan, which details that a heat health warning should be communicated when the minimum and maximum average temperature is >30˝C (86˝F). This plan changed from using maximum thresholds, to basing its activation process on the Bureau of Meteorology's Heatwave Service for Australia definition of heatwave as "three days or more of high maximum and minimum temperature that are unusual for that location" [28]. The South Australian Government has also moved away from an activation process of three consecutive days >35˝C (95˝F), and have specifically redefined their activation trigger to three or more consecutive days with an average daily temperature of >32˝C (90˝F).
This figure was designed to consider cooler coastal winds. These changes provide further validation of a need to change the Canberra regions activation threshold to a mean temperature measure [29]. Furthermore, this is reflective of preceding research conducted in South Australia. One study found mortality was associated with heat events of three or more consecutive days with maximum temperatures >43˝C (109˝F), or an average temperature >34˝C (93˝F) [30], whereas another study [15] found that Emergency Department presentations became apparent above maximum and minimum temperatures of 34˝C (93˝F) and 22˝C (72˝F), with a mean presentation temperature of 28˝C (82˝F). These government changes and preceding studies, provide complying literature associated with the movement away from maximum temperature thresholds within Australia.
During the Canberra heatwave events of January and February 2014, the impact on the normal business of the Emergency Departments was distinguished as minimal, though it is noted that this view only highlights the requirement for a differentiation between physiological and meteorological heatwaves with such plans. As discussed, there are other indicators and variables associated with population resilience to heat based events such as the exacerbation of chronic conditions, an important consideration when creating/evaluating an extreme heat event plan or allocating a heat based threshold.
There is a risk associated with the setting of a "safe" (lower threshold) heat plan activation, in the form of alert fatigue both within the community and the health sector. The threshold must be set in an effort to minimise the public health risk presented to a community secondary to an extreme heat event, though due to the geographical breath and location of Australia, Australians are both accustomed to and aware of 'hot summers' and bushfire threats. Thus when considering public health heatwave/extreme heat notifications, if the threshold is too low, excluding those unaccustomed (visitors/tourists), there is a risk of alert fatigue where the recipients of the message experience a negative response (poor uptake of the message), when the message is delivered too frequently [31]. Further examination of heatwave events, including the experience of local medical officers/general practitioners and volunteer health care provider organisations, may help in eliciting a further understanding of this potential negative communication risk.
Complicating the physiological heatwave situation in Australian is the concurrent heightened bushfire risk that is also associated with sustained high meteorological temperatures. Not only do active bushfires have a direct impact on local meteorological conditions, they also complicate the public health situation with reduced air quality. These considerations should be taken into account, with other factors such as local population influences; economic, social, cultural and health status, in planning to minimise the impact of heat events [32].
Further factors such as wind speed and humidity play a defined role in physiological heatwaves events. The variance between coastal and inland threshold values/tools suggests such an influence, though further analysis is required to describe the variable. A more complex tool may be required to take these meteorological variables into account. It has been suggested that such variables may be addressed in a mean threshold range trigger, for example 27-30˝C (80.6-86˝F), with independent co-triggers around humidity and peak temperatures [18]. Emerging meteorological variables such as those noted secondary to climate change will also necessitate further analysis of heatwave threshold based plans.
Whilst a portion of the literature suggests that general/widespread adoption of a mean daily temperature threshold is not likely in regions such as Europe, due to their local experience around the public health impact of heatwaves and extreme high temperatures events, this should not discourage the application of a mean temperature type alert system within Australia. Such a mean temperature threshold system takes into account both heatwaves and extreme temperature events when the mean threshold is set at a point appropriate to the regional seasonal average temperatures [33]. A notification tool, as outlined in Figure 3, could be developed and used in the notification of heat related health risks (Figure 3). This tool could be displayed in prominent areas within the community and on local television. Modern media modalities such as social media should also be considered with all public health messaging, capturing large sectors of a populations, within a very short period. When deciding on or setting a trigger threshold for public health related heat events, health departments and agencies would be well advised to differentiate between physiological and meteorological heatwaves, as well as isolated extreme heat events. Such decisions need to take into account; the resilience of the local population, regional seasonal average temperatures, economic, social, cultural, and health status factors.
Within the defined local population/cohort, during the examined summer periods of sustained heat, there were significant outdoor mass gatherings, including the National Multicultural Festival and Fringe Festival. These events being held in Canberra on an annual basis. The outdoor nature and large participation in these public events presents the potential for a public health challenge/threat. Had there been an associated consequence, with significant heat related presentations, this may have presented a further example of physiological versus meteorological heatwaves. Fortunately, there were no significant heatwave related presentations to the local Emergency Departments, thus no there was no health impact noted in association with these large outdoor events.
A limitation of this study, could include that patients with low-moderate heat illnesses may not have been admitted or reported to the hospital. This could have therefore, resulted in an underreporting of the wider patient data, as this study only included those who presented to our Emergency Department. Further investigation into the health impact noted by general practitioners, prehospital healthcare providers and volunteer first aid providers, during a heatwave/extreme heat event, would expand on the complete public health picture. Furthermore this study does not directly address the aggravation of diseases related to heatwaves, or the possible deaths as a potential result of changes in temperature. For example, a future area of research could confirm the belief, that as the temperature climbs to >28 °C (82.4 °F), the risk of cardiac arrest increases. Future studies could also test the heat aggravations associated with chronic disease, such as acute coronary syndrome, renal disease, hypertension, and diabetes mellitus. It is believe that these sub-groups, and the associated commodities, are at greater risk for the ill effects from the heat.

Conclusions
Whilst there are many variables associated in setting a heatwave plan activation threshold, the minimal impact noted within this study, under the current Extreme Heat Management Plan, supports the application of a mean daily temperature based threshold, taking into account local factors such as socioeconomic variables. The population resilience experienced during the heatwaves of January and February 2014 is suggestive that the current plan trigger threshold of maximum daily temperatures reaching or exceeding 35 °C (95 °F) (actual or predicted) over three consecutive days, could be revised, considering the demonstrated tolerance. Results gained from this study validate existing mean threshold recommendations [1,15,18]. Further study into the influence of heatwaves and extreme heat events, outside of the established acute health care sector is indicated to further account for resilience and mass gathering impact.  When deciding on or setting a trigger threshold for public health related heat events, health departments and agencies would be well advised to differentiate between physiological and meteorological heatwaves, as well as isolated extreme heat events. Such decisions need to take into account; the resilience of the local population, regional seasonal average temperatures, economic, social, cultural, and health status factors.
Within the defined local population/cohort, during the examined summer periods of sustained heat, there were significant outdoor mass gatherings, including the National Multicultural Festival and Fringe Festival. These events being held in Canberra on an annual basis. The outdoor nature and large participation in these public events presents the potential for a public health challenge/threat. Had there been an associated consequence, with significant heat related presentations, this may have presented a further example of physiological versus meteorological heatwaves. Fortunately, there were no significant heatwave related presentations to the local Emergency Departments, thus no there was no health impact noted in association with these large outdoor events.
A limitation of this study, could include that patients with low-moderate heat illnesses may not have been admitted or reported to the hospital. This could have therefore, resulted in an underreporting of the wider patient data, as this study only included those who presented to our Emergency Department. Further investigation into the health impact noted by general practitioners, prehospital healthcare providers and volunteer first aid providers, during a heatwave/extreme heat event, would expand on the complete public health picture. Furthermore this study does not directly address the aggravation of diseases related to heatwaves, or the possible deaths as a potential result of changes in temperature. For example, a future area of research could confirm the belief, that as the temperature climbs to >28˝C (82.4˝F), the risk of cardiac arrest increases. Future studies could also test the heat aggravations associated with chronic disease, such as acute coronary syndrome, renal disease, hypertension, and diabetes mellitus. It is believe that these sub-groups, and the associated commodities, are at greater risk for the ill effects from the heat.

Conclusions
Whilst there are many variables associated in setting a heatwave plan activation threshold, the minimal impact noted within this study, under the current Extreme Heat Management Plan, supports the application of a mean daily temperature based threshold, taking into account local factors such as socioeconomic variables. The population resilience experienced during the heatwaves of January and February 2014 is suggestive that the current plan trigger threshold of maximum daily temperatures reaching or exceeding 35˝C (95˝F) (actual or predicted) over three consecutive days, could be revised, considering the demonstrated tolerance. Results gained from this study validate existing mean threshold recommendations [1,15,18]. Further study into the influence of heatwaves and extreme heat events, outside of the established acute health care sector is indicated to further account for resilience and mass gathering impact.