Heat Watch/Warning Systems Save Lives: Estimated Costs and Benefits for Philadelphia 1995-1998

Thomas J. Teisberg1, Kristie L. Ebi2*, Laurence S. Kalkstein3, Lawrence Robinson4, Rodney F. Weiher5

1 Teisberg Associates

1475 Ingleside Drive

Charlottesville, VA  22901

(434) 293-4003

tjteisberg@compuserve.com

 

2 EPRI

3412 Hillview Ave.

Palo Alto, CA  94304

(650) 855-2735

krisebi@epri.com

* Corresponding author

 

3 Center for Climatic Research

University of Delaware

Newark, DE  19702

(302) 831-8269

larryk@udel.edu

 

4 Deputy Health Commissioner

City of Philadelphia

1101 Market St

Philadelphia, PA  19107

(215) 685-5350

larry.robinson@phila.gov

 

5 NOAA

Office of Policy & Strategic Planning

HCHB, Rm 6117

US Department of Commerce

14th and Constitution Ave, NW

Washington, DC  20230

(202) 482-0636

rodney.f.weiher@hdg.noaa.gov

 

Dateline:____________________________________________________________


Heat Watch/Warning Systems Save Lives: Estimated Costs and Benefits for Philadelphia 1995-1998

 

Running title:  Heat Wave Warning Systems Save Lives

Keywords:

Heat

Mortality

Warning Systems

Value of Information

Methods

Abbreviations:

          PWWS = Philadelphia Hot Weather-Health Watch/Warning System

          NWS = National Weather Service

          VSL = Value of a Statistical Life


Abstract

The Philadelphia Hot Weather-Health Watch/Warning System was initiated in 1995 to alert the city’s population to take precautionary actions when hot weather posed risks to health.  The number of lives saved and the economic benefit of this system were estimated using data from 1995-1998.  Excess mortality in people 65 years of age and older was defined as reported mortality minus mortality predicted by a historical trend line developed over the period 1964-1988.  Excess mortality during heatwaves was explained using linear regression.  Two variables were convincingly associated with mortality: the Time of Season when a particular heatwave started, and a Warning variable indicating whether or not a heatwave warning had been issued.  The estimated coefficient of the Warning variable was about –2.6, suggesting that when a warning was issued, 2.6 lives were saved, on average, for each warning day and for three days after the warning ended.  Given the number of warnings issued over the three-year period, the system saved an estimated 117 lives.  Estimated dollar costs for running the system were small compared with estimates of the value of a life. 


1.          Introduction

Severe and sustained episodes of summer heat are associated with increased morbidity and mortality, particularly in temperate regions.  The death toll in an unprepared region can be substantial.  For example, during July 1-14, 1993, the eastern United States experienced a severe heatwave with high temperatures (33.9 – 38.3 ºC) and high humidity (36 – 58%) (US CDC 1994).  During July 6-14, the Philadelphia Medical Examiners determined 118 deaths were heat related (either a core body temperature of 40.6 ºC or higher, or a body found in a hot, unventilated environment).  This is certainly an underestimate, as heat is associated with increased mortality from a number of causes other than heat stroke (Shen et al. 1998).  Historically, cardiovascular diseases have accounted for 13% to 90% of the increase in overall mortality during and following a heatwave, cerebrovascular disease 6% to 52%, and respiratory diseases 0% to 14% (Kilbourne 1997).  Heatwaves also increase the rate of nonfatal illnesses. 

Partly in response to heatwaves in 1993 and 1994, the Philadelphia Hot Weather-Health Watch/Warning System (PWWS) was initiated in 1995 to alert the city’s population when weather conditions posed risks to health (Kalkstein et al. 1996, Mirchandani et al. 1996, Sheridan and Kalkstein 1998).  Appendix 1 includes a description of the PWWS, including the measures taken during a heatwave to reduce morbidity and mortality.  The National Weather Service (NWS) advises the Philadelphia Department of Health of potentially dangerous weather based on the forecasts generated by the PWWS and other information.  The Philadelphia Department of Health then implements emergency precautions and mitigation procedures to reduce mortality risk (Kalkstein et al. 1996).  This system is the basis for more than a dozen other heat-health watch warning systems being instituted in cities worldwide (Kalkstein 2003; Sheridan and Kalkstein 1998).  This is the first attempt to calculate the number of lives saved and the economic benefit of these systems in reducing heat related mortality. 

2.          Methods

Kalkstein et al. identified two types of hot weather patterns, maritime tropical and dry tropical, which have been historically associated with elevated mortality in Philadelphia (Kalkstein et al. 1996).  We refer to these weather patterns as "heatwaves."  Our analyses included daily data for all such heatwaves during 1995-1998, plus up to three additional days after the end of each heatwave.  The three days following a heatwave were included because mortality effects can lag heatwaves by up to three days (Curriero et al. 2002, Laschewski and Jendritzky 2002). 

For the days in our sample, mortality data were obtained from the National Center for Health Statistics for the 65 and older age group in the Philadelphia SMSA for 1995 through 1998, the last year for which mortality data were available (National Center for Health Statistics 2000).  We chose to limit our analyses to this age group because they are vulnerable to excessive heat; therefore the statistical evidence regarding the effectiveness of the warning system should be strongest for this age group (Semenza et al. 1996, Smoyer 1998, US CDC 1995).  The analyses were based on excess mortality, where excess mortality was calculated as the difference between reported mortality and the underlying mortality trend estimated from years prior to 1995, as described in Appendix 1 (Kalkstein et al. 1996). 

We developed a statistical explanation of excess mortality using the following explanatory variables: daily weather variables, the duration in days of each heatwave, the daily sequence number indicating the time of the 139-day summer season on which each heatwave commenced (Time of Season), the classification of each heatwave day as maritime tropical or dry tropical, and an indicator variable to show whether a heatwave warning was issued (Warning Indicator).  Weather data for the time period 1995-1998 included temperature at 5AM, and temperature and dew point at 5PM (US Department of Commerce 2002). 

A heatwave warning was issued only if the local NWS office concurred with a PWWS recommendation that a warning be issued.  Before the development of the PWWS, the NWS issued most heat/health alerts based on guidelines that relied heavily on the computation of the “heat index” (NOAA 1994).  An excessive heat warning was issued when daytime heat index values were expected to exceed 40.5 °C for more than three hours a day on two consecutive days, or when the daytime heat index was expected to exceed 46 °C for any length of time (Kalkstein et al. 1996). 

During the time period of the study, the local NWS office agreed to use the PWWS for guidance in the issuance of heat advisories and excessive heat warnings.  However, forecasters did not completely rely on the system and were often conservative in issuing advisories and warnings.  In some cases, forecasters would not call an advisory or warning if the heat index was below 38 °C, even if the system called for such an issuance.  Some forecasters felt that the system called too many advisories and warnings and were concerned that the public might become less responsive to subsequent warnings (Kalkstein, personal communication).  An early evaluation of the system indicated that, because of this policy, warnings were not called frequently enough and that heat-related deaths occurred on days when the system called for a warning and the forecasters did not issue one (Kalkstein et al. 1996).  More recently, forecasters are using the PWWS as primary guidance, although they are advised to use discretion and not call advisories and warnings based exclusively on the system.

The NWS concurred with a PWWS recommendation that a warning be issued less than one-quarter of the time; during the period 1995-1998, there were 70 days the PWWS suggested a heat warning, but only 16 days when the NWS concurred and warnings were issued.  In addition, there were 5 days on which the NWS issued warnings, even though the PWWS did not recommend one, making a total of 21 warning days.  Counting the days immediately following the end of each of these warnings, our sample included a total of 45 days for which a warning could potentially have affected excess mortality during 1995-1998. 

Excess mortality was linearly regressed against these 45 heatwave days using the linear regression capabilities provided in MicroSoft Excel.

3.          Results

a.          Estimation of Lives Saved

Among the explanatory variables considered for inclusion in the regression model, only Time of Season and Warning Indicator were convincingly associated with excess mortality.  The omission of the daily weather variables might seem surprising, but is less so when one recognizes that all the days in our statistical sample were identified as potentially life-threatening based on weather variables.  Thus, within this subset of days, weather differences might be relatively small in terms of their effects on mortality.  Other variables that did not significantly contribute to the explanation of excess mortality were the duration of the heatwave and airmass type (maritime tropical or dry tropical).  The latter means that there were no significant differences between the airmasses; obviously, both explain excess mortality, as they were the basis for warnings. 

Excluding variables not significantly associated with mortality left the following regression:

Excess Mortality = 3.27 - 0.049 x Time of Season – 2.58 x Warning Indicator

The t-statistics were (-2.55) for Time of Season and (-1.43) for the Warning Indicator.  A low R-squared value of 0.04 suggests that mortality depends on weather to only a minor extent, which is not too surprising because omitted disease processes and accidents would be expected to be the primary determinants of morality.  The coefficient for Time of Season means that excess mortality (assuming no warning) could range from +3.22 to –3.54 lives depending on whether a heatwave occurred on the first or last day of the season.

The estimated coefficient of the Warnings Indicator meant that when a warning was issued, assuming no mortality displacement, 2.6 lives were saved, on average, for each day a warning was issued as well as for the three days immediately following the warning.  Mortality displacement results when deaths are brought forward in time; e.g., some individuals who died during a heatwave would have died anyway within a short period of time after the heatwave.  Because there were 45 days during 1995-1998 when warnings were issued (or the effect of warnings persisted over the three-day lag after the warning ended), the estimated total number of lives saved was 117.

Because serial correlation of errors is a potential statistical problem in time series analysis, we also estimated the association using heatwaves as the sampling unit.  For this purpose, excess mortality was defined as average daily excess mortality for the days of the heatwave (and three days after), and the Warning Indicator was set to one whenever there was a warning issued on at least one of the heatwave days.  The results obtained were very similar to those presented above; in particular the coefficient of the Warnings Indicator was the same, i.e. a savings of 2.6 lives per day.

b.          Estimation of the Benefits of the PWWS

For some public policy decisions, lives are assigned monetary value.  The USEPA, for example, places values on lives (value of a statistical life or VSL) to assess the benefits of policies that reduce pollution (Smith et al. 2001).  VSL are in units of dollars per life saved.  The key piece of data underlying the VSL is the value of mortality risk reduction.  It has been estimated by analyzing wage rates in occupations with differing mortality risks, to identify the wage differential due to mortality risk, and by contingent valuation, i.e. asking a sample of people to assess their own willingness to pay to reduce risks.  The results of both methods are usually similar.

Two recent studies addressed the question of whether the value of a statistical life is lower for older people or people with health problems (Krupnick et al. 2000, Smith et al. 2001).  This is of particular interest in connection with the PWWS, because the population most at risk from heat is older and frequently suffers from health problems.  Smith et al. estimated VSLs for older Americans who were still working, and reported a range of $5.3 - $6.6 million (Smith et al. 2001).  While the people in the Smith study were younger (generally 50-65 years old) and probably healthier than the people most at risk from heatwaves, Smith et al. found no indication that the VSL fell with age over this range, nor that the VSL was lower for those with serious illness.  Krupnick et al. estimated the VSL for Canadians 40 to 75 years old, and reported a range of $0.82 to $2.6 million (US dollars) (Krupnick et al. 2000).  Krupnick et al. also found that poor health did not lower the VSL.  However, they concluded that Canadians 70-75 years old were willing to pay about a third less for mortality risk reductions. 

We decided to use the USEPA estimate of $6.12 million for the value of a statistical life (Smith et al. 2001) and accepted the Krupnick et. al. finding that VSLs eventually fall off with age, being a third lower for the 70-75 age group.  Thus, we assumed $4 million for the VSL among people 65 years of age or older in Philadelphia.  This implies that the gross benefits of the Philadelphia heatwave warning system could be on the order of $468 million (117 lives saved times $4 million).

c.          Estimation of the Costs of the PWWS

As discussed in detail in Appendix 1, when a heatwave warning is declared in Philadelphia, the city takes a number of actions to reduce the risks of heat related mortality.  Most of these actions do not have direct monetary costs, including actions taken by city employees as a normal part of their jobs, actions taken by volunteers, and delayed actions (i.e. halting service suspensions).  A few of the actions taken do have direct costs, including the Heatline and additional EMS crews.  Costs for these actions are primarily additional wages.  We estimated these costs to be $1000/day on weekdays and $3000/day on weekends for the Heatline, and $4000/day for EMS crews.  Because we don’t have full information on all direct costs, we assumed they would be on the order of $10,000 per day when a heatwave warning is issued.  During 1995-1998, there were 21 such days.  This implies that the total cost of the Philadelphia system was on the order of $210,000 over the three-year period.

The Philadelphia system depends on weather forecasts generated by the NWS.  One might want to allocate some portion of the costs of those forecasts, and count it as part of the costs of the system.  Obviously, making such an allocation would be very difficult and we did not attempt to do so.  Instead, the net benefit estimate reported below may be interpreted as one of many societal benefits derived from weather forecasting systems.

d.          Estimation of the Net Benefits of the PWWS

Obviously, if the Philadelphia system saves any lives at all, it will have large estimated dollar benefits.  This is because the VSL, for even one life, is bigger than the costs of running the system.  In fact, system costs are so far into the “noise” of the estimate that they are essentially irrelevant.  Thus, based on this analysis, we would conclude that the net benefits of the Philadelphia heatwave warning system were around $468 million over the three-year period, 1995-1998.

4.          Conclusions

The results suggest that issuing a warning lowered daily mortality by about 2.6 lives on average, based on a number of assumptions.  One is that the baseline used to generate the estimates of excess mortality is valid.  Excess mortality was calculated as the difference between reported mortality and the underlying mortality trend estimated from years prior to 1995.  The mortality trend line directly reflects changes in population, and indirectly reflects secular time trends in the housing stock, peoples’ lifestyles, access to health care, quality of health care, and any other underlying factors that would tend to change mortality rates over time.  We believe that variations around such a mortality trend line are a better indicator of the effects of daily weather, and of efforts to counter those effects, than are medical examiners’ determinations that deaths were caused by extreme temperatures. 

Another assumption is that there was no mortality displacement.  The assumption of no mortality displacement is likely to have inflated the estimate of the number of lives saved.  It was not possible to estimate mortality displacement with the data available.  There is a wide range of estimates of mortality displacement following heatwaves, from very little to explaining much of the net excess mortality (Huynen et al. 2001, Kalkstein 1998, Kunst et al. 1993, Laschewski and Jendritzky 2002, Rooney et al. 1998).  On the other hand, because the NWS declared warnings only about one-quarter of the time that the PWWS forecast suggested declaring a warning, our estimate of the number of lives saved by the PWWS is likely to be an underestimate. 

As implied above, it would seem that more lives might have been saved if more warnings had been declared.  However, there is a question about whether the daily number of lives saved might be different between days when only the PWWS recommended a warning and those days when both the PWWS and the NWS concurred on declaring a warning.  To address this question, we looked at temperatures on different subsets of days.  On the 70 days the PWWS suggested a warning, the average temperature was 32.68 ºC.  On 16 of these days, the NWS concurred and a warning was issued; on these days plus the additional five days the NWS issued a warning independent of the PWWS, the average temperature was 32.77 ºC.  On the remaining 54 days when the PWWS suggested a warning and the NWS did not concur, the average temperature was 32.34 ºC.

While these data suggest that the NWS was marginally more conservative about declaring warnings, the temperature differences were very small.  This supports the view that the number of lives saved per day might have been about the same if warnings had been declared on additional days.  Moreover, the low estimated cost of warnings argues for being more liberal about issuing warnings.  In support of this, recent analyses in London and Toronto suggest that excess mortality begins at relatively low temperatures (Hajat et al. 2002, Semenza et al. 1999, Smoyer-Tomic and Rainham 2001). 

However, there remains some risk that if too many warning are declared, they may become less effective, especially with regard to the voluntary actions taken when a warning is issued.

The value of the t-statistic suggests that there is about an 8% probability of our estimate of a reduction in excess mortality of 2.6 lives per day occurring by chance.  We interpret this as reasonably strong statistical confirmation of our prior expectation that warnings would save lives.

The R-squared for the linear regression is low, which implies that most of the variation in excess mortality is attributable to variables not included in our model.  This was expected because weather is only one of a number of factors related to mortality.  Indeed, it is reasonable to expect that disease processes and accidents are the dominant causes of mortality in the 65 and older age group; we do not have variables representing these phenomena in our model.  However, as long as these omitted variables are uncorrelated with our included variables (most importantly, uncorrelated with the warnings indicator), the estimated coefficients of our included variables will be unbiased.  Particularly in the case of the warnings indicator, it’s unlikely that this variable is statistically correlated with whatever myriad of other variables is driving the unexplained variation in daily excess mortality.

Therefore, over the period 1995-1998, we conservatively estimate that the Philadelphia Hot Weather-Health Watch Warning System saved approximately 117 lives.  This estimate of total benefits is likely to be low.  First, it only considered mortality reductions in the 65 and older population – if there were mortality reductions in younger populations too, this would add to the estimated benefits.  Second, we did not estimate any increased morbidity associated with excess heat exposure.  It is reasonable to assume that morbidity also was reduced by the PWWS.  Hospitalizations were shown to increase in London and Chicago during the 1995 heatwaves (Rooney et al. 1998, Semenza et al. 1999).  Morbidity reduction is an additional benefit not included in the analyses. 

Evaluations are underway for heat watch warning systems established in other cities, including Toronto, Dayton and Phoenix, although there currently are no direct estimates of the number of lives saved.  An observation support the value of such systems.  However, an observation based on the first year of operation of the Philadelphia system during the hot summer of 1995 supports our results.  It was noted that, as the summer progressed, the mortality algorithm within the system overestimated heat-related mortality with greater regularity and by a greater amount.  This supports the suggestion that people respond to the heat advisories and warnings in ways that save lives (Kalkstein et al. 1996).  Further, in a detailed evaluation of heat watch warning system operation in a number of cities, it was clear that cities utilizing systems like the one in Philadelphia were developing and instituting more sophisticated heat intervention plans whenever warnings were called (Kalkstein 2003).  These improved plans may be a direct result of system development.

The methods described here can be used to evaluate the costs and benefits of implementing heatwave early warning systems.  Our results suggest that the benefits in terms of lives saved will far outweigh the operational costs of such systems, at least for cities located in temperate regions.


Appendix 1.          Philadelphia hot weather-health watch/warning system (PWWS)

Dr. Laurence S. Kalkstein designed the Philadelphia Hot Weather-Health Watch/Warning System to predict periods when there is high risk of heat-related mortality (Kalkstein et al. 1996).  Past weather was classified into more or less homogenous categories (synoptic categories), referred to as airmass types, based on air temperature, dewpoint temperature, cloud cover, sea level pressure, wind speed, and wind direction.  Associations were determined between airmass types and mortality.  Cause of death, place and date of death, age and race were extracted from National Center for Health Statistics files for the Philadelphia Standard Metropolitan Statistical Area for the years 1964-1966, 1973-1976, 1978 and 1980-1988; these were the years for which the date of death was available (Kalkstein et al. 1996).  The analysis used the total number of deaths per day.  Direct standardization was used to adjust all mortality data for changes in the total population over time.  A mortality trend line was constructed for each time period and daily mortality expressed as deviations around this baseline.  The mean daily mortality for each synoptic air mass category was determined, along with the standard deviation, to ascertain whether particular categories were associated with high or low mortality.  Potential lag times of one to three days were accounted for in the analysis; the highest mortality was found with no lag time. 

Two airmass types were associated with elevated mortality in Philadelphia: maritime tropical and dry tropical.  As not all days within these air masses resulted in elevated mortality, a stepwise multiple regression analysis was used to identify which days within the air mass were associated with increased mortality.  The regression identified several variables that were predictive of elevated mortality.  These were the number of consecutive days the air mass was present, the maximum temperature, and the time of season (e.g., whether the oppressive air mass occurs early or late within the summer season) (Kalkstein et al. 1996). 

The accuracy of the forecast data and the performance of the categorizations were verified by backcasting, with archived model forecasts used as predictors.  For the summer of 1988, the 24-hour forecast identified 89% of the oppressive days and the 48-hour forecast identified 71%.  The system was designed to operate from May 15 through September 30 (139 days).

Each day the PWWS forecasts the air mass category for the current day and the subsequent two (Kalkstein et al. 1996).  The beginning and end of a heatwave are determined by air mass type and by the mortality predictive equations associated with each air mass.  The system was designed to generate a health watch, a health alert or a health warning based on air mass type.  The criteria for a health warning are that a maritime tropical or dry tropical air mass type is forecast for either that afternoon or the following morning, and the predictive model associated with that particular air mass forecasts four or more heat related deaths.  The local office of the National Weather Service determines whether or not to issue a warning based on the PWWS forecasts, the heat index and other information.  For example, in 1995 the NWS issued a heat warning on 9 of the 15 days recommended by the PWWS. 

The City of Philadelphia and other agencies and organizations institute a series of intervention activities when the NWS issues a warning.  Television and radio stations and newspapers are asked to publicize the oppressive weather conditions, along with information on how to avoid heat-related illnesses.  In addition, these media announcements encourage friends, relatives, neighbors, and other volunteers (“buddies”) to make daily visits to elderly persons during the hot weather.  These buddies are asked to ensure that the most susceptible individuals have sufficient fluids, proper ventilation and other amenities to cope with the weather.  A “Heatline” is operated in conjunction with the Philadelphia Corporation for the Aging to provide information and counseling to the general public on avoidance of heat stress.  The Heatline telephone number is publicized by the media and by a large display seen over much of the center of Philadelphia.  When a warning is issued, the Department of Public Health contacts nursing homes and other facilities boarding persons requiring extra care to inform them of the high-risk heat situation and to offer advice on the protection of residents.  The local utility company and water department halt service suspensions during warning periods.  The Fire Department Emergency Medical Service increases staffing during warnings in anticipation of increased service demand.  The agency for homeless services activates increased daytime outreach activities to assist those on the streets.  Senior centers extend their hours of operation of air-conditioned facilities during warning periods. 


Acknowledgements:

This work was funded by EPRI under contract EP-P5431/C2688 and by NOAA under contract 40AANA1A1110.  The authors would like to thank Dr. Andrew Solow for his useful suggestions on the statistical analysis.


References:

Curriero F.C., K.S. Heiner, J.M. Samet, S.L. Zeger, L. Strug, and J.A. Patz, 2002: Temperature and mortality in 11 Cities of the Eastern United States.  Am. J. Epidemiol., 155, 80-87.

Hajat S., R.S. Kovats, R.W. Atkinson, and A. Haines, 2002: Impact of hot temperatures on death in London: a time series approach.  J. Epidemiol. Community Health, 56, 367-372.

Huynen M., P. Martens, D. Schram, et al., 2001: The impact of cold spells and heatwaves on mortality rates in the Dutch population.  Environ. Health Perspect. 109, 463-70.

Kalkstein, L.S., 1998: Climate and Human Mortality: Relationships and Mitigating Measures.  Advances in Bioclimatology: Human Bioclimatology, A. Auliciems, Ed., Springer, Berlin, 161-177.

Kalkstein L.S., 2003: Description of our heat/health watch-warning systems: their nature and extent, and required resources.  Final report to Stratus Consulting Company, Boulder, Colorado, 31pp.

Kalkstein L.S., P.F. Jamason, J.S. Greene, J. Libby, and L. Robinson, 1996: The Philadephia Hot Weather-Health Watch/Warning System: development and application, Summer 1995.  BAMS 177, 1519-1528.

Kilbourne E.M., 1997: Heat waves and hot environments.  The Public Health Consequences of Disasters, E.K. Noji, Ed., Oxford University Press, New York, 245-269.

Krupnick A., A. Alberini, M. Cropper, N. Simon, B. O’Brien, R. Goerce, and M. Heintzelman, 2000: Age, health, and the willingness to pay for mortality risk reductions: a contingent valuation survey of Ontario residents.  Discussion Paper 00-37, Resources for the Future, Washington D.C.

Kunst A.E., C.W.N. Looman, and J.P. Mackenbach, 1993: Outdoor air temperature and mortality in the Netherlands: a time-series analysis.  Am. J. Epidemiol., 137, 331-341.

Laschewski G. and G. Jendritzky, 2002: Effects of the thermal environment on human health: an investigation of 30 years of daily mortality data from SW Germany.  Clim. Res., 21, 91-103.

Mirchandani H.G., G. McDonald, I.C. Hood, and C. Fonseca, 1996: Heat-related deaths in Philadelphia – 1993.  Am. J. Forensic Med. Pathol., 17, 106-108.

National Center for Health Statistics, 2000: Standardized micro-data mortality transcripts.  U.S. Department of Health, Education, and Welfare, Washington, DC.

NOAA, 1994: Regional Operations Manual Letter.  Eastern Region National

Weather Service, Bohemia, NY.

Rooney C., A.J. McMichael, R.S. Kovats, and M.P. Coleman, 1998: Excess mortality in England and Wales and Greater London during the 1995 heatwave.  Epidemiol. Community Health, 52, 482-486.

Semenza J.C., J.E. McCullough, W.D. Flanders, M.A. McGeehin, and J.R. Lumpkin, 1999: Excess hospital admissions during the July 1995 heat wave in Chicago.  Am. J. Prev. Med., 16, 269-277.

Semenza J.C., C.H. Rubin, K.H. Falter, et al., 1996: Heat-related deaths during the July 1995 heat wave in Chicago.  N. Engl. J. Med., 335, 84-90.

Shen T., H.L. Howe, C. Alo, and R.L. Moolenaar.  1998: Toward a broader definition of heat-related death: comparison of mortality estimates from medical examiners’ classification with those from total death differentials during the July 1995 heat wave in Chicago, Illinois.  Am. J. Forensic Med. Pathol., 19, 113-118.

Sheridan S.C., and L.S. Kalkstein, 1998: Heat watch-warning systems in urban areas.  World Resource Rev., 10, 375-383.

Smith V.K., H. Kim, and D.H. Taylor Jr., 2001: Do the ‘near’ elderly value mortality risks differently?  Paper presented at USEPA Workshop, Economic Valuation of Mortality Risk Reduction: Assessing the State of the Art for Policy Applications, Silver Spring MD.

Smoyer K.E., 1998: A comparative analysis of heat waves and associated mortality in St. Louis, Missouri – 1980 and 1995.  Int. J. Biometeorol., 42, 73-86.

Smoyer-Tomic K.E, and D.G.C. Rainham, 2001: Beating the heat: development and evaluation of a Canadian hot weather health-response plan.  Environ. Health Perspect., 109, 1241-1248.

U.S. Centers for Disease Control and Prevention, 1994: Heat-related deaths – Philadelphia and United States 1993-1994.  MMWR 43, 453-455.

U.S. Centers for Disease Control and Prevention, 1995: Heat- related illnesses and deaths – United States 1994-1995.  MMWR 44, 465-468.

U.S. Department of Commerce, 2002: Climatological Data: United States.  Asheville, NC:NOAA/NCDC.


Copyright © University of Delaware, 2003 December.
Synoptic Climatology Lab
Comments, suggestions, or questions may be sent here.