- Email: [email protected]
The impact of warm weather on mass event medical need: a review of the literature
American Journal of Emergency Medicine (2010) 28, 224–229
www.elsevier.com/locate/ajem
Brief Report
The impact of warm weather on mass event medical need: a review of the literature Matthew B. Baird MD a , Robert E. O'Connor MD a , Allen L. Williamson RN b , Benjamin Sojka BA, EMT-P c , Kostas Alibertis BA, EMT-P d , William J. Brady MD a,b,⁎ a
Department of Emergency Medicine, University of Virginia, Charlottesville, VA 22908, USA Special Event Medical Management, University of Virginia Medical Center, Charlottesville, VA 22908, USA c Charlottesville-Albemarle Rescue Squad, Charlottesville, VA 22901, USA d Western Albemarle Rescue Squad, Crozet, VA 22906, USA b
Received 26 August 2008; revised 3 October 2008; accepted 26 October 2008
Abstract Over the last 20 years, interest in medical need at mass events has increased. Many studies have been published identifying the characteristics of such events that significantly impact the number of patients who seek care. Investigators agree that weather is one of the most important variables. We performed a literature search using several biomedical databases (MEDLINE via PubMed, the Cochrane database, BMJ's Clinical Evidence compendium, and Google Scholar) for articles addressing the effect of weather on medical need at mass events. This search resulted in 8 focused articles and several other resources from the reference sections of these publications. We found that the early literature is composed of case reports and predominantly subjective observations concerning the impact of weather on medical need. Most investigators agree upon a positive relationship between heat/humidity and the frequency of patient presentation. More recent authors make attempts at quantifying the relationship and propose prediction models for patient volume and medical personnel requirements. We present an ancestral review of these studies, discuss their results collectively, and propose a simplified algorithm for predicting patient volume at mass events. This review is intended for event planners and mass event emergency medical personnel for planning future events. We also hope to stimulate further study to develop and verify prediction models. © 2010 Elsevier Inc. All rights reserved.
1. Introduction Mass events are defined as groups of greater than 1000 people [1-4]. These gatherings take place in a variety of settings and under a variety of conditions. Consequently, providing medical care at such gatherings is a challenging ⁎ Corresponding author. Department of Emergency Medicine, University of Virginia, Charlottesville, VA 22908, USA. E-mail address: [email protected] (W.J. Brady). 0735-6757/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.ajem.2008.10.034
task for event planners and emergency medical personnel. Although some events may require only a single BLS team, others require multiple basic life support (BLS) and advanced life support (ALS) teams, a fleet of transport vehicles, and physician presence to provide acceptable care [2]. Predicting the extent of medical need required at mass events is a topic that has gained more interest over the last 20 years. The earliest literature includes an analysis of event characteristics that significantly affect patient presentation rates [6-10]. These characteristics include event type (rock concert, sporting event, papal mass, family reunion, etc),
The impact of warm weather on mass event medical need Table 1 Reference
225
Studies addressing the effect of weather on medical need at mass gatherings Subjects
Study design
Regression analysis was performed, investigating the effect of weather (heat and humidity) on the number of patients seen at the Indianapolis 500 auto race from 1983 to 1989. A prediction model was developed. Data from 1990 were sed to test its validity. Michael Unspecified number of Retrospective literature review of 35 case and Barbera [7] patients from 35 case reports concerning the effect of certain mass reports. event characteristics on number of patients seeking medical care. Mann-Whitney U test was used as a correlation tool. Arbon et al [8] 11 956 patients presenting Survey-based observational study involving patients from 201 mass events in 12-month for medical care at a period in Australia. Data included the mass event incidence and type of patients presenting for treatment and on the environmental factors that influenced these presentations. Zeitz et al [9] N7000 patients N7 years Retrospective review of casualty reports from an annual mass event in south Australia over 7 years. Determinants of workload were reviewed. Milsten et al [10] 5899 patients from Retrospective review of all patient visits 216 events during 3-year period in Baltimore area. The effect of type, sex, age, weather, and attendance on medical usage rates (MURs) was evaluated. Perron et al [11] 716 patients from 20 Retrospective review of patients seen at a division I football games single football stadium seating 61 625 in the southeast United States from 1999 to 2003. Heat index was calculated for each game using a standard chart, and the Pearson product-moment correlation (PPMC) was calculated to determine positive correlation between heat index and PPTT. Retrospective review of medical records Kman et al [12] Unspecified number of subjects from 47 division I rom football games from 2 outdoor stadiums from 2001 to 2005, and weather football games data from the National Weather Service. A binomial model was used to develop a formula to correlate temperature with number of patients seen. Retrospective observational study Hartman et al [13] Unspecified number of involving review of records from events subjects from 55 mass of various sizes under various conditions. gathering events Each event was analyzed based on heat index, attendance, presence of alcohol, age demographics, and crowd intentions. Each condition was assigned a score (0-2) based on impact on medical need. Based on total score, each event was labeled as either “major,” “intermediate,” or “minor.” To verify the classification system, the number of patients seen during each event was analyzed. In their classification system, heat index N90°F was assigned 2 points; b90, 1 point; and climate-controlled events, 0 points. Bowdish et al [6] Unspecified number of spectators seeking medical care at annual auto race (1983-1990)
Relevant findings There was a significant correlation between dew point (temperature and humidity) and patient load, but the model served as a poor predictor for the 1990 race. Regression analysis is an inadequate prediction tool “Hot climate conditions,” with unspecified parameters, had a significant positive relationship with frequency of patient visits. Weather, particularly relative humidity, was positively related to increases in patient presentation rates. Regression modeling and attention to historical data can help improve planning. The number and type of “casualty presentations” correlated significantly with maximum daily temperature and humidity. The MURs at events held when the temperature was N80°F were significantly lower than at events held at b80°F (4.9 vs 8.10 PPTT) PPMC was calculated as 0.607, indicating a strong positive correlation between heat index and patient volume. Linear modeling predicted that for every 10° increase in heat index, 3 more PPTT will seek medical care. A formula to estimate number of patients based on temperature and number of attendants was derived. The formula shows an increase of 3.64-4.05 PPTT, with a temperature increase from 20°C-21°C (11% increase). The mean number of patient encounters followed the classification system in order, with major events at 71 patients, intermediate events at 6.3, and minor events at 2.3. The number of full evaluations and transfers also correlated with the classification system.
226
M.B. Baird et al.
crowd size, crowd intent, age demographic, venue (outdoor vs indoor), the presence of physical boundaries, time of day, and availability of alcohol [7,8,10,13]. Heat and humidity are the 2 significant weather variables identified by these studies [6,8,10]. In recent investigations, more authors have attempted to develop quantitative prediction models to use for planning of future events [11-13]. Humidity [8], temperature [12], and heat index [11] (a combination of heat and humidity) have been included in these algorithms. The purpose of this review is to summarize the available literature on the effect of warm weather on mass event medical need and develop simple algorithms for the estimation of patient volume. We hope that event planners and mass event medical professionals will consider these algorithms when allocating medical resources for future events.
2. Methods A search of available literature was performed using multiple biomedical databases. Our search of MEDLINE database via PubMed server yielded the most results. The following keywords were searched: weather, climate, heatindex, mass event, event medicine, mass event medicine, sporting event medicine, mass event planning, and mass gatherings. Additional searches were performed using a combination of the first 3 terms and the following terms joined by the “AND” command. Identical searches were performed using the Cochrane Library, BMJ's Clinical Evidence compendium, and Google Scholar. Once all articles were compiled, we chose to review only those articles involving analysis of the effect of weather on mass event medical need. This sorting process produced 8 peer review articles. Citations within these articles were searched to identify additional references pertinent to this review.
3. Results The effect of environmental factors on medical need at mass events was first observed in 1980 by Pons et al [3] as a subjective observation: “There appeared to be a direct relationship between the recorded temperature during the game and the number of patients evaluated.” No quantitative analysis of this relationship was explored at that time. Several articles in the 1980s briefly discussed the effects of weather,
Fig. 1
but the topic was not the primary object of these investigations, and thus, no quantitative analysis was attempted [4,5]. Subsequent investigations studied the effect of weather in more detail. These studies are summarized in Table 1 and discussed below. In 1992, Bowdish et al [6] was the first group to specifically study the correlation between weather and medical need at mass events using regression analysis. They found a significant correlation between dew point, a calculation involving temperature and humidity, and the number of patients requiring medical care at a large annual automobile race. From their data, they developed a mathematical prediction tool. However, in the year after their initial analysis (1990), their model was unable to accurately predict the number of spectators requiring medical attention at their event. They concluded, therefore, that regression analysis was not sophisticated enough to predict medical need. From 1992 to 2005, several retrospective observational studies were performed to correlate the need for medical care at mass events with various environmental factors [7-10]. These factors included crowd size, age, sex, event type (sporting events vs papal masses vs concerts), the availability of alcohol, temperature, and humidity. Arbon et al [8] were the first to devise a predictive model based on their observations. Regarding weather, they were not able to find a correlation between temperature and patient presentation rate (number of presentations per 1000 patrons) but did observe a linear relationship between humidity and medical need. This study, despite this linear relationship between humidity and medical need, was hampered by 2 issues: First, their mathematical model is complex and involves 9 different variables. And unfortunately, this model is most appropriately used in larger events—this particular model is designed for use at gatherings with more than 25,000 people. Furthermore, there are no published studies evaluating the validity of the model. Interestingly, Milsten et al [10] found a negative correlation between temperature and medical usage rates at Baltimore area events in 2003. This conclusion is in stark contrast to all other studies used for this review and perhaps indicates that temperature alone is not an adequate predictive factor for medical need at mass events. In 2005, Perron et al [11] quantified the relationship between medical need, defined as the number of patients per 10 000 spectators (PPTT), and the heat index, a combination
Hartman et al [13] scoring system (reproduced from reference text).
The impact of warm weather on mass event medical need Table 2
227
Heat index calculation chart
of temperature and humidity, azt football games in the southeast United States. They concluded that with every 10° rise in the heat index, the PPTT increased by 3. This was the first clinically significant prediction model specifically correlating heat and presentation rates. In 2007, Kman et al [12] derived a formula to predict patient volumes based on temperature and event attendance. The model was derived from data collected from college football games in the southeast United States. The resulting formula is displayed below: eð7:43830:24439Tð1Þ + 0:0156032Tð2Þ0:000229196Tð3ÞÞ Tattendance This formula predicts that an increase in temperature from 20°C to 21°C will increase the PPTT from 3.64 to 4.05 (an 11% increase). No additional studies have been published at this time to validate or disprove this model. Finally, in 2008, Hartman et al [13] developed a simple prediction model. They tested their model on 55 mass events of various sizes and conditions. The scoring system, illustrated in Fig. 1, accurately separated the events based on number of patients presenting for care, indicating the potential utility of this model as a prediction tool. This model considered 5 different variables—weather (defined as heat index), number of participants, presence of ethanol, average age of crowd, and crowd “intentions” (ie, extent of rowdiness). The authors developed a predictive tool; they went on to recommend the extent of health care provider coverage for different events, based on this scoring system. For “minor” events, they recommend a single transport vehicle with 1 ALS and 1 BLS provider. For “intermediate” events, 2 transport units and several ALS (1-3) and BLS (1-6) providers are recommended. For “major” events, they recommend “multiple ALS and BLS personnel, specialized equipment and in some cases, physician presence.”
4. Discussion The available literature clearly indicates that weather impacts the need for medical care at mass events. All studies reviewed, with the exception of one [10], demonstrate a positive correlation between temperature and/or humidity and patient presentation rates [3-9,11-13]. Unfortunately, most data are subjective. In recent years, more effort has been made to quantify the relationship and develop prediction models. Some of the first models developed were quite complex, involving multiple variables. In recent years, more simplified tools have been developed. None of the proposed models have been verified. Consequently, there is not a universally accepted algorithm to predict medical need at mass gatherings. One of the reasons for our inability to accept a universal model has been disagreement on the most influential weather variable. Arbon et al [8] used humidity in their prediction model yet found no correlation between temperature and medical need. Kman et al [12] did find a significant correlation between temperature and presentation rates and use temperature as a variable in their model. Milsten et al [10] actually found that the number of patients dropped when the temperature rose above 80°F. These conflicting data likely suggest that temperature alone is not predictive. Perhaps, the heat index, a measure of temperature and humidity, is a more appropriate predictor of mass event medical need? Along this line of query, Perron et al [11] studied heat index, noting an association between increasing number of patients seen and rising heat index. Hartman et al [13] also used heat index in their prediction model (Fig. 1), which they used to successfully categorize mass events based on medical need. Thus, heat index seems to be the most promising weather variable for use in a simple universal prediction model. As a combination of heat and humidity, the value is determined
228
M.B. Baird et al.
using a standard 2-axis chart (Table 2). It can be reliably used at any venue regardless of climate differences. Although the heat index may allow us to control for climate, multiple other variables must be accounted for when predicting medical need for future events. Type of event (children's entertainment, papal mass, sporting event, rock concert, etc), crowd size, participant age, presence of alcohol, the presence of physical boundaries, and time of day have all been shown to significantly affect the presentation rate [7-10,12,13]. Arbon et al [8] included all of these variables into their model, yet no subsequent investigators have used the model, perhaps because of its complexity. The goal is a simple, user-friendly version of the model that does not exclude important variables. This model, although noting an association of humidity and medical need, was complex in terms of input data requirement and cumbersome in application, ultimately not demonstrating significant predictive ability. The key to developing such a model involves close attention to historical data [8,9]. When planning for a specific event, data from similar events will provide more accurate information than any universal model can predict. Most events are repeated on an annual or semiannual—or even more frequent—basis at the same venue, with the same policies regarding alcohol. The use of historical data here is very important in that the recurring nature of the particular event “controls” for many variables which are consistent. Thus, by including historical data in a predictive model, many variables can be accounted for, including event type, crowd size, alcohol availability, and boundaries. Weather is the most inconsistent variable in that it changes from week to week, month to month, season to season, and year to year. Therefore, a simple mathematical model would be helpful to estimate the impact of changes in weather. This model would use historical data, which, as noted, includes consideration of participant number, presence of alcohol, crowd intentions, event boundaries, and others. The model, using specific event historical information and weather, is based on past event characteristics and heat index. The model was developed from a combination of the findings from the reviewed studies. Specifically, the model incorporates the relationship between heat index and patient volume determined by Perron et al [11] (with every 10° increase in the heat index, the number of patients presenting for care per 10 000 patrons increased by 3). Y = NH +
HI XH 3 10
Y = estimated number of PPTT NH = average number of PPTT from prior events HI = current heat index XH = average heat index from prior events
To illustrate the usefulness of the proposed model, we will present 3 hypothetical examples involving a large, an intermediate, and a small mass event. The large event is a college football game with an attendance of 61 250, a temperature of 95°F, and a humidity of 70%. Historical data from 2004 indicate that the patient volume was 17.5 PPTT (NH) and the heat index was 81 (XH). First, the current heat index must be determined by using the heat index calculation chart (Table 2) and the current temperature and humidity. For this event, the heat index is 124 (HI). Now, the algorithm can be applied: Y = 17:5 +
124 81 3 10
Y = 30:4 PPTT The large increase in the heat index results in a significant increase in patient volume. To find out the actual number of patients predicted, simply multiply Y by the attendance, remembering that Y is the number of PPTT patrons. The result is shown below: Number of patients = 61; 250 patrons
30:4 patients 10; 000 patrons
Y = 186:2 patients The intermediate event is a horse race with 12 320 expected attendants. It is held in the spring, with a current heat index of 77. Last year, the heat index was 67 and the number of patients seen was 12.5 PPTT. Again, apply the formula: Y = 12:5 +
77 67 3 10
Y = 15:5 PPTT Number of patients = 12; 320 patrons
15:5 patients 10; 000 patrons
Number of patients = 19:1 patients The small event is wedding reception involving 1250 people. It is held in the spring, with a heat index of 69. The available historical data are for a wedding involving 1210 people held in the summer, with a heat index of 93. Records indicate that a total of 3 people presented for medical care, which correlates to 24.8 PPTT (NH). Y = 16:5 +
69 93 3 10
Y = 9:3 PPTT Number of patients = 1; 250 patrons
9:3 patients 10; 000 patrons
Number of patients = 1:2 patients Again, the use of historical data allows for multiple variables to be accounted for. The result is a simple model that can be applied to events in any location.
This example illustrates the utility of the model when the heat index drops from the historical level.
The impact of warm weather on mass event medical need These examples not only demonstrate the use of the proposed prediction model but also emphasize the importance of attendance on expected changes in patient volume. Identical changes in the heat index will result in greater changes in the number of patients for larger event. This may have staffing implications for these events, which would not be necessary for smaller gatherings. To further test the model, specific small, intermediate, and large events were reviewed. In a general sense, the model was highly accurate in larger events, whereas at smaller venues, it performed less well. Currently, the model is being prospectively evaluated at the testing institution. One problem with the above model is that it requires available historical data. If such data are not available, an alternative method for predicting medical need is necessary. The prediction model presented by Hartman et al [13] may be useful in this circumstance. It is easy to use, gives a rough estimate of the number of patients to expect, and suggests the amount and type of medical personnel required. It has not been widely tested, however, so collection of data from such an event is critical for the planning of future events. Unlike the model proposed by Hartman et al [13], the algorithm above does not predict the extent of necessary medical coverage. Historical information should include the number and type of providers used at previous events, but it is difficult to predict when additional providers are needed. Collaboration between care providers and administrators associated with a given event will be important when making this determination. The available data have multiple limitations requiring further study. Although algorithms have been developed over the past decade, in this review, none have been verified. Use of these algorithms and documentation of their accuracy are necessary for the development of a universal prediction tool for medical need at mass events. The capabilities of medical care providers are also a topic worthy of future investigation. Such data would be helpful in determining when additional providers are needed at an event based on the predicted patient load. In addition, the effect of cold weather has not been well documented. Intuition suggests that an increase in cold-related illnesses would occur when the temperature drops below a certain level. The definition of “cold weather” and its effect on presentation rates will require significant study in appropriate locations.
229
5. Conclusions This literature review of the impact of warm weather on medical need at mass events included 8 original studies and other references directed by those publications. Overall, the literature suggests a positive relationship between temperature/humidity and presentation rates. The heat index, a combination of temperature and humidity, is the most promising weather variable for prediction models. The importance of historical data is also paramount in predicting medical need at future events. Several prediction models have been proposed, and this study presents a possible algorithm based on these models. Further study is necessary to verify these algorithms and establish a universal prediction model.
References [1] Jaslow D, Yancey A, et al. Mass gathering medical care. Prehosp Emerg Care 2000;4(4):359-60. [2] Martin-Gill C, Brady WJ, Barlotta K, et al. Hospital-based healthcare provider (nurse and physician) integration into an EMS-managed mass gathering event. Am J Emerg Med 2007;25(1):15-22. [3] Pons PT, Holland B, Alfrey E, et al. An advanced emergency medical care system at National Football League games. Ann Emerg Med 1980; 9:203-6. [4] Spaite DW, Criss EA, Valenzuela TD, et al. A new model for providing prehospital medical care in large stadiums. Ann Emerg Med 1988;17: 825-8. [5] Sanders AB, Criss E, Steck P, et al. An analysis of medical care at mass gatherings. Ann Emerg Med 1986;15:17-21. [6] Bowdish GE, Cordell WH, Bock HC, Vukov LF. Using regression analysis to predict emergency patient volume at the Indianapolis 500 mile race. Ann Emerg Med 1992;21(10):1200-3. [7] Michael JA, Barbera JA. Mass gathering medical care: a twenty-five year review. Prehosp Disaster Med 1997;12(4):305-12. [8] Arbon P, Bridgewater MB, Smith C. Mass gathering medicine: a predictive model for patient presentation and transport rates. Prehosp Disaster Med 2001;16(3):109-16. [9] Zeitz KM, Schneider DP, Jarrett D, et al. Mass gathering event: retrospective analysis of patient presentation over seven years. Prehosp Disaster Med 2002;17(3):147-50. [10] Milsten AM, Seaman KG, Liu P, et al. Variables influencing medical usage rates, injury patterns, and levels of care for mass gatherings. Prehosp Disaster Med 2003;18(4):334-46. [11] Perron AD, Brady WJ, Custalow CB, et al. Association of heat index and patient volume at a mass gathering event. Prehosp Emerg Care 2005;9:49-52. [12] Kman NE, Russell GB, Bozeman WP, et al. Derivation of a formula to predict patient volume based on temperature at college football games. Prehosp Emerg Care 2007;11(4):453-7. [13] Hartman N, Brady WJ, O'Connor R, et al. Predicting resource utilization at mass gatherings using a simplified stratification scoring model. Am J Emerg Med 2008 [accepted, but not yet published].
www.elsevier.com/locate/ajem
Brief Report
The impact of warm weather on mass event medical need: a review of the literature Matthew B. Baird MD a , Robert E. O'Connor MD a , Allen L. Williamson RN b , Benjamin Sojka BA, EMT-P c , Kostas Alibertis BA, EMT-P d , William J. Brady MD a,b,⁎ a
Department of Emergency Medicine, University of Virginia, Charlottesville, VA 22908, USA Special Event Medical Management, University of Virginia Medical Center, Charlottesville, VA 22908, USA c Charlottesville-Albemarle Rescue Squad, Charlottesville, VA 22901, USA d Western Albemarle Rescue Squad, Crozet, VA 22906, USA b
Received 26 August 2008; revised 3 October 2008; accepted 26 October 2008
Abstract Over the last 20 years, interest in medical need at mass events has increased. Many studies have been published identifying the characteristics of such events that significantly impact the number of patients who seek care. Investigators agree that weather is one of the most important variables. We performed a literature search using several biomedical databases (MEDLINE via PubMed, the Cochrane database, BMJ's Clinical Evidence compendium, and Google Scholar) for articles addressing the effect of weather on medical need at mass events. This search resulted in 8 focused articles and several other resources from the reference sections of these publications. We found that the early literature is composed of case reports and predominantly subjective observations concerning the impact of weather on medical need. Most investigators agree upon a positive relationship between heat/humidity and the frequency of patient presentation. More recent authors make attempts at quantifying the relationship and propose prediction models for patient volume and medical personnel requirements. We present an ancestral review of these studies, discuss their results collectively, and propose a simplified algorithm for predicting patient volume at mass events. This review is intended for event planners and mass event emergency medical personnel for planning future events. We also hope to stimulate further study to develop and verify prediction models. © 2010 Elsevier Inc. All rights reserved.
1. Introduction Mass events are defined as groups of greater than 1000 people [1-4]. These gatherings take place in a variety of settings and under a variety of conditions. Consequently, providing medical care at such gatherings is a challenging ⁎ Corresponding author. Department of Emergency Medicine, University of Virginia, Charlottesville, VA 22908, USA. E-mail address: [email protected] (W.J. Brady). 0735-6757/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.ajem.2008.10.034
task for event planners and emergency medical personnel. Although some events may require only a single BLS team, others require multiple basic life support (BLS) and advanced life support (ALS) teams, a fleet of transport vehicles, and physician presence to provide acceptable care [2]. Predicting the extent of medical need required at mass events is a topic that has gained more interest over the last 20 years. The earliest literature includes an analysis of event characteristics that significantly affect patient presentation rates [6-10]. These characteristics include event type (rock concert, sporting event, papal mass, family reunion, etc),
The impact of warm weather on mass event medical need Table 1 Reference
225
Studies addressing the effect of weather on medical need at mass gatherings Subjects
Study design
Regression analysis was performed, investigating the effect of weather (heat and humidity) on the number of patients seen at the Indianapolis 500 auto race from 1983 to 1989. A prediction model was developed. Data from 1990 were sed to test its validity. Michael Unspecified number of Retrospective literature review of 35 case and Barbera [7] patients from 35 case reports concerning the effect of certain mass reports. event characteristics on number of patients seeking medical care. Mann-Whitney U test was used as a correlation tool. Arbon et al [8] 11 956 patients presenting Survey-based observational study involving patients from 201 mass events in 12-month for medical care at a period in Australia. Data included the mass event incidence and type of patients presenting for treatment and on the environmental factors that influenced these presentations. Zeitz et al [9] N7000 patients N7 years Retrospective review of casualty reports from an annual mass event in south Australia over 7 years. Determinants of workload were reviewed. Milsten et al [10] 5899 patients from Retrospective review of all patient visits 216 events during 3-year period in Baltimore area. The effect of type, sex, age, weather, and attendance on medical usage rates (MURs) was evaluated. Perron et al [11] 716 patients from 20 Retrospective review of patients seen at a division I football games single football stadium seating 61 625 in the southeast United States from 1999 to 2003. Heat index was calculated for each game using a standard chart, and the Pearson product-moment correlation (PPMC) was calculated to determine positive correlation between heat index and PPTT. Retrospective review of medical records Kman et al [12] Unspecified number of subjects from 47 division I rom football games from 2 outdoor stadiums from 2001 to 2005, and weather football games data from the National Weather Service. A binomial model was used to develop a formula to correlate temperature with number of patients seen. Retrospective observational study Hartman et al [13] Unspecified number of involving review of records from events subjects from 55 mass of various sizes under various conditions. gathering events Each event was analyzed based on heat index, attendance, presence of alcohol, age demographics, and crowd intentions. Each condition was assigned a score (0-2) based on impact on medical need. Based on total score, each event was labeled as either “major,” “intermediate,” or “minor.” To verify the classification system, the number of patients seen during each event was analyzed. In their classification system, heat index N90°F was assigned 2 points; b90, 1 point; and climate-controlled events, 0 points. Bowdish et al [6] Unspecified number of spectators seeking medical care at annual auto race (1983-1990)
Relevant findings There was a significant correlation between dew point (temperature and humidity) and patient load, but the model served as a poor predictor for the 1990 race. Regression analysis is an inadequate prediction tool “Hot climate conditions,” with unspecified parameters, had a significant positive relationship with frequency of patient visits. Weather, particularly relative humidity, was positively related to increases in patient presentation rates. Regression modeling and attention to historical data can help improve planning. The number and type of “casualty presentations” correlated significantly with maximum daily temperature and humidity. The MURs at events held when the temperature was N80°F were significantly lower than at events held at b80°F (4.9 vs 8.10 PPTT) PPMC was calculated as 0.607, indicating a strong positive correlation between heat index and patient volume. Linear modeling predicted that for every 10° increase in heat index, 3 more PPTT will seek medical care. A formula to estimate number of patients based on temperature and number of attendants was derived. The formula shows an increase of 3.64-4.05 PPTT, with a temperature increase from 20°C-21°C (11% increase). The mean number of patient encounters followed the classification system in order, with major events at 71 patients, intermediate events at 6.3, and minor events at 2.3. The number of full evaluations and transfers also correlated with the classification system.
226
M.B. Baird et al.
crowd size, crowd intent, age demographic, venue (outdoor vs indoor), the presence of physical boundaries, time of day, and availability of alcohol [7,8,10,13]. Heat and humidity are the 2 significant weather variables identified by these studies [6,8,10]. In recent investigations, more authors have attempted to develop quantitative prediction models to use for planning of future events [11-13]. Humidity [8], temperature [12], and heat index [11] (a combination of heat and humidity) have been included in these algorithms. The purpose of this review is to summarize the available literature on the effect of warm weather on mass event medical need and develop simple algorithms for the estimation of patient volume. We hope that event planners and mass event medical professionals will consider these algorithms when allocating medical resources for future events.
2. Methods A search of available literature was performed using multiple biomedical databases. Our search of MEDLINE database via PubMed server yielded the most results. The following keywords were searched: weather, climate, heatindex, mass event, event medicine, mass event medicine, sporting event medicine, mass event planning, and mass gatherings. Additional searches were performed using a combination of the first 3 terms and the following terms joined by the “AND” command. Identical searches were performed using the Cochrane Library, BMJ's Clinical Evidence compendium, and Google Scholar. Once all articles were compiled, we chose to review only those articles involving analysis of the effect of weather on mass event medical need. This sorting process produced 8 peer review articles. Citations within these articles were searched to identify additional references pertinent to this review.
3. Results The effect of environmental factors on medical need at mass events was first observed in 1980 by Pons et al [3] as a subjective observation: “There appeared to be a direct relationship between the recorded temperature during the game and the number of patients evaluated.” No quantitative analysis of this relationship was explored at that time. Several articles in the 1980s briefly discussed the effects of weather,
Fig. 1
but the topic was not the primary object of these investigations, and thus, no quantitative analysis was attempted [4,5]. Subsequent investigations studied the effect of weather in more detail. These studies are summarized in Table 1 and discussed below. In 1992, Bowdish et al [6] was the first group to specifically study the correlation between weather and medical need at mass events using regression analysis. They found a significant correlation between dew point, a calculation involving temperature and humidity, and the number of patients requiring medical care at a large annual automobile race. From their data, they developed a mathematical prediction tool. However, in the year after their initial analysis (1990), their model was unable to accurately predict the number of spectators requiring medical attention at their event. They concluded, therefore, that regression analysis was not sophisticated enough to predict medical need. From 1992 to 2005, several retrospective observational studies were performed to correlate the need for medical care at mass events with various environmental factors [7-10]. These factors included crowd size, age, sex, event type (sporting events vs papal masses vs concerts), the availability of alcohol, temperature, and humidity. Arbon et al [8] were the first to devise a predictive model based on their observations. Regarding weather, they were not able to find a correlation between temperature and patient presentation rate (number of presentations per 1000 patrons) but did observe a linear relationship between humidity and medical need. This study, despite this linear relationship between humidity and medical need, was hampered by 2 issues: First, their mathematical model is complex and involves 9 different variables. And unfortunately, this model is most appropriately used in larger events—this particular model is designed for use at gatherings with more than 25,000 people. Furthermore, there are no published studies evaluating the validity of the model. Interestingly, Milsten et al [10] found a negative correlation between temperature and medical usage rates at Baltimore area events in 2003. This conclusion is in stark contrast to all other studies used for this review and perhaps indicates that temperature alone is not an adequate predictive factor for medical need at mass events. In 2005, Perron et al [11] quantified the relationship between medical need, defined as the number of patients per 10 000 spectators (PPTT), and the heat index, a combination
Hartman et al [13] scoring system (reproduced from reference text).
The impact of warm weather on mass event medical need Table 2
227
Heat index calculation chart
of temperature and humidity, azt football games in the southeast United States. They concluded that with every 10° rise in the heat index, the PPTT increased by 3. This was the first clinically significant prediction model specifically correlating heat and presentation rates. In 2007, Kman et al [12] derived a formula to predict patient volumes based on temperature and event attendance. The model was derived from data collected from college football games in the southeast United States. The resulting formula is displayed below: eð7:43830:24439Tð1Þ + 0:0156032Tð2Þ0:000229196Tð3ÞÞ Tattendance This formula predicts that an increase in temperature from 20°C to 21°C will increase the PPTT from 3.64 to 4.05 (an 11% increase). No additional studies have been published at this time to validate or disprove this model. Finally, in 2008, Hartman et al [13] developed a simple prediction model. They tested their model on 55 mass events of various sizes and conditions. The scoring system, illustrated in Fig. 1, accurately separated the events based on number of patients presenting for care, indicating the potential utility of this model as a prediction tool. This model considered 5 different variables—weather (defined as heat index), number of participants, presence of ethanol, average age of crowd, and crowd “intentions” (ie, extent of rowdiness). The authors developed a predictive tool; they went on to recommend the extent of health care provider coverage for different events, based on this scoring system. For “minor” events, they recommend a single transport vehicle with 1 ALS and 1 BLS provider. For “intermediate” events, 2 transport units and several ALS (1-3) and BLS (1-6) providers are recommended. For “major” events, they recommend “multiple ALS and BLS personnel, specialized equipment and in some cases, physician presence.”
4. Discussion The available literature clearly indicates that weather impacts the need for medical care at mass events. All studies reviewed, with the exception of one [10], demonstrate a positive correlation between temperature and/or humidity and patient presentation rates [3-9,11-13]. Unfortunately, most data are subjective. In recent years, more effort has been made to quantify the relationship and develop prediction models. Some of the first models developed were quite complex, involving multiple variables. In recent years, more simplified tools have been developed. None of the proposed models have been verified. Consequently, there is not a universally accepted algorithm to predict medical need at mass gatherings. One of the reasons for our inability to accept a universal model has been disagreement on the most influential weather variable. Arbon et al [8] used humidity in their prediction model yet found no correlation between temperature and medical need. Kman et al [12] did find a significant correlation between temperature and presentation rates and use temperature as a variable in their model. Milsten et al [10] actually found that the number of patients dropped when the temperature rose above 80°F. These conflicting data likely suggest that temperature alone is not predictive. Perhaps, the heat index, a measure of temperature and humidity, is a more appropriate predictor of mass event medical need? Along this line of query, Perron et al [11] studied heat index, noting an association between increasing number of patients seen and rising heat index. Hartman et al [13] also used heat index in their prediction model (Fig. 1), which they used to successfully categorize mass events based on medical need. Thus, heat index seems to be the most promising weather variable for use in a simple universal prediction model. As a combination of heat and humidity, the value is determined
228
M.B. Baird et al.
using a standard 2-axis chart (Table 2). It can be reliably used at any venue regardless of climate differences. Although the heat index may allow us to control for climate, multiple other variables must be accounted for when predicting medical need for future events. Type of event (children's entertainment, papal mass, sporting event, rock concert, etc), crowd size, participant age, presence of alcohol, the presence of physical boundaries, and time of day have all been shown to significantly affect the presentation rate [7-10,12,13]. Arbon et al [8] included all of these variables into their model, yet no subsequent investigators have used the model, perhaps because of its complexity. The goal is a simple, user-friendly version of the model that does not exclude important variables. This model, although noting an association of humidity and medical need, was complex in terms of input data requirement and cumbersome in application, ultimately not demonstrating significant predictive ability. The key to developing such a model involves close attention to historical data [8,9]. When planning for a specific event, data from similar events will provide more accurate information than any universal model can predict. Most events are repeated on an annual or semiannual—or even more frequent—basis at the same venue, with the same policies regarding alcohol. The use of historical data here is very important in that the recurring nature of the particular event “controls” for many variables which are consistent. Thus, by including historical data in a predictive model, many variables can be accounted for, including event type, crowd size, alcohol availability, and boundaries. Weather is the most inconsistent variable in that it changes from week to week, month to month, season to season, and year to year. Therefore, a simple mathematical model would be helpful to estimate the impact of changes in weather. This model would use historical data, which, as noted, includes consideration of participant number, presence of alcohol, crowd intentions, event boundaries, and others. The model, using specific event historical information and weather, is based on past event characteristics and heat index. The model was developed from a combination of the findings from the reviewed studies. Specifically, the model incorporates the relationship between heat index and patient volume determined by Perron et al [11] (with every 10° increase in the heat index, the number of patients presenting for care per 10 000 patrons increased by 3). Y = NH +
HI XH 3 10
Y = estimated number of PPTT NH = average number of PPTT from prior events HI = current heat index XH = average heat index from prior events
To illustrate the usefulness of the proposed model, we will present 3 hypothetical examples involving a large, an intermediate, and a small mass event. The large event is a college football game with an attendance of 61 250, a temperature of 95°F, and a humidity of 70%. Historical data from 2004 indicate that the patient volume was 17.5 PPTT (NH) and the heat index was 81 (XH). First, the current heat index must be determined by using the heat index calculation chart (Table 2) and the current temperature and humidity. For this event, the heat index is 124 (HI). Now, the algorithm can be applied: Y = 17:5 +
124 81 3 10
Y = 30:4 PPTT The large increase in the heat index results in a significant increase in patient volume. To find out the actual number of patients predicted, simply multiply Y by the attendance, remembering that Y is the number of PPTT patrons. The result is shown below: Number of patients = 61; 250 patrons
30:4 patients 10; 000 patrons
Y = 186:2 patients The intermediate event is a horse race with 12 320 expected attendants. It is held in the spring, with a current heat index of 77. Last year, the heat index was 67 and the number of patients seen was 12.5 PPTT. Again, apply the formula: Y = 12:5 +
77 67 3 10
Y = 15:5 PPTT Number of patients = 12; 320 patrons
15:5 patients 10; 000 patrons
Number of patients = 19:1 patients The small event is wedding reception involving 1250 people. It is held in the spring, with a heat index of 69. The available historical data are for a wedding involving 1210 people held in the summer, with a heat index of 93. Records indicate that a total of 3 people presented for medical care, which correlates to 24.8 PPTT (NH). Y = 16:5 +
69 93 3 10
Y = 9:3 PPTT Number of patients = 1; 250 patrons
9:3 patients 10; 000 patrons
Number of patients = 1:2 patients Again, the use of historical data allows for multiple variables to be accounted for. The result is a simple model that can be applied to events in any location.
This example illustrates the utility of the model when the heat index drops from the historical level.
The impact of warm weather on mass event medical need These examples not only demonstrate the use of the proposed prediction model but also emphasize the importance of attendance on expected changes in patient volume. Identical changes in the heat index will result in greater changes in the number of patients for larger event. This may have staffing implications for these events, which would not be necessary for smaller gatherings. To further test the model, specific small, intermediate, and large events were reviewed. In a general sense, the model was highly accurate in larger events, whereas at smaller venues, it performed less well. Currently, the model is being prospectively evaluated at the testing institution. One problem with the above model is that it requires available historical data. If such data are not available, an alternative method for predicting medical need is necessary. The prediction model presented by Hartman et al [13] may be useful in this circumstance. It is easy to use, gives a rough estimate of the number of patients to expect, and suggests the amount and type of medical personnel required. It has not been widely tested, however, so collection of data from such an event is critical for the planning of future events. Unlike the model proposed by Hartman et al [13], the algorithm above does not predict the extent of necessary medical coverage. Historical information should include the number and type of providers used at previous events, but it is difficult to predict when additional providers are needed. Collaboration between care providers and administrators associated with a given event will be important when making this determination. The available data have multiple limitations requiring further study. Although algorithms have been developed over the past decade, in this review, none have been verified. Use of these algorithms and documentation of their accuracy are necessary for the development of a universal prediction tool for medical need at mass events. The capabilities of medical care providers are also a topic worthy of future investigation. Such data would be helpful in determining when additional providers are needed at an event based on the predicted patient load. In addition, the effect of cold weather has not been well documented. Intuition suggests that an increase in cold-related illnesses would occur when the temperature drops below a certain level. The definition of “cold weather” and its effect on presentation rates will require significant study in appropriate locations.
229
5. Conclusions This literature review of the impact of warm weather on medical need at mass events included 8 original studies and other references directed by those publications. Overall, the literature suggests a positive relationship between temperature/humidity and presentation rates. The heat index, a combination of temperature and humidity, is the most promising weather variable for prediction models. The importance of historical data is also paramount in predicting medical need at future events. Several prediction models have been proposed, and this study presents a possible algorithm based on these models. Further study is necessary to verify these algorithms and establish a universal prediction model.
References [1] Jaslow D, Yancey A, et al. Mass gathering medical care. Prehosp Emerg Care 2000;4(4):359-60. [2] Martin-Gill C, Brady WJ, Barlotta K, et al. Hospital-based healthcare provider (nurse and physician) integration into an EMS-managed mass gathering event. Am J Emerg Med 2007;25(1):15-22. [3] Pons PT, Holland B, Alfrey E, et al. An advanced emergency medical care system at National Football League games. Ann Emerg Med 1980; 9:203-6. [4] Spaite DW, Criss EA, Valenzuela TD, et al. A new model for providing prehospital medical care in large stadiums. Ann Emerg Med 1988;17: 825-8. [5] Sanders AB, Criss E, Steck P, et al. An analysis of medical care at mass gatherings. Ann Emerg Med 1986;15:17-21. [6] Bowdish GE, Cordell WH, Bock HC, Vukov LF. Using regression analysis to predict emergency patient volume at the Indianapolis 500 mile race. Ann Emerg Med 1992;21(10):1200-3. [7] Michael JA, Barbera JA. Mass gathering medical care: a twenty-five year review. Prehosp Disaster Med 1997;12(4):305-12. [8] Arbon P, Bridgewater MB, Smith C. Mass gathering medicine: a predictive model for patient presentation and transport rates. Prehosp Disaster Med 2001;16(3):109-16. [9] Zeitz KM, Schneider DP, Jarrett D, et al. Mass gathering event: retrospective analysis of patient presentation over seven years. Prehosp Disaster Med 2002;17(3):147-50. [10] Milsten AM, Seaman KG, Liu P, et al. Variables influencing medical usage rates, injury patterns, and levels of care for mass gatherings. Prehosp Disaster Med 2003;18(4):334-46. [11] Perron AD, Brady WJ, Custalow CB, et al. Association of heat index and patient volume at a mass gathering event. Prehosp Emerg Care 2005;9:49-52. [12] Kman NE, Russell GB, Bozeman WP, et al. Derivation of a formula to predict patient volume based on temperature at college football games. Prehosp Emerg Care 2007;11(4):453-7. [13] Hartman N, Brady WJ, O'Connor R, et al. Predicting resource utilization at mass gatherings using a simplified stratification scoring model. Am J Emerg Med 2008 [accepted, but not yet published].