Preparing for Extreme Heat Events: Practices in Identifying Mortality

Abstract

Climate change is increasing the frequency and severity of extreme heat events. These events affect cities in increasingly abrupt and catastrophic ways; yet, many of the deaths caused by exposure to heat have gone unnoticed or are inaccurately identified, resulting in a lack of urgency in addressing this issue. We aim to address this under-identification of deaths from heat waves in order to better assess heat risk. We investigated death records in New York City from 2010 to 2012 to identify characteristics that vary between deaths officially categorized as caused by heat wave exposure (oHDs) and those possibly caused by heat (pHDs). We found that oHDs were more often black and of a younger age than would typically be expected. We also found that there was a lack of evidence to substantiate that an oHD had occurred, using the NYC official criteria. We conclude that deaths from heat waves are not being accurately recorded, leading to a mis-estimation. Training regarding the collection and interpretation of evidence may improve preparedness for heat events.

The authors investigated death records in New York City from 2010 to 2012 to identify characteristics that vary between deaths officially categorized as caused by heat wave exposure and those possibly caused by heat. They found that people who die in the heat were more often black and of a younger age than would typically be expected. Training in the collection and interpretation of evidence may improve preparedness for heat events.

Extreme heat is the top cause of weather-related fatalities in the United States. From 2003 to 2012, the annual average number of heat deaths exceeded the combined annual average number of deaths due to floods, lightning, tornadoes, and hurricanes combined.¹ Epidemiologists have found a consistent relationship between heightened morbidity and mortality and extreme heat.² High temperatures adversely affect the human body by interfering with its ability to dissipate heat and thermoregulate, leading to heat exhaustion and possibly heat stroke, a condition characterized by a core temperature >105°F and central nervous system dysfunction.³ Exposure to high temperatures has also been shown to exacerbate chronic preexisting illnesses.⁴

Heat waves are expected to increase in frequency and severity with climate change⁵ and have been cited as the most pressing related health impact of climate change.^6,7 Global temperature averages climbed by 1 to 2 degrees Fahrenheit in the last century and are projected to increase 2 to 11 degrees Fahrenheit by 2100.⁸ Heat waves have resulted in a rising number of crises and mass mortality events domestically and internationally, including in 1993 in Philadelphia, where 118 people died; in Chicago in 1995, where approximately 800 people died; in Western Europe in 2003, where an estimated 70,000 people died;^9-13 in Western India in 2009, where thousands of people died;¹⁴ and in 2010, when 55,000 people died in Eastern Europe.¹³

Given the substantial impacts of heat on mortality and the likelihood of climate change to exacerbate those effects, it is critical for public health leaders to develop effective heat death prevention strategies. While research has suggested a number of community- and individual-level characteristics as risk factors for heat death, these efforts, as well as those to validly estimate the number of deaths attributable to excess heat, have been complicated by heat not being recorded as the underlying or contributing cause of death on death certificates.¹⁵ Many heat deaths may be recorded as deaths attributable to preexisting conditions that actually increased the deceased's susceptibility to heat death, resulting in considerable underestimates of heat death incidence.¹⁶ At present, the factors contributing to this under-diagnosis and the extent to which under-diagnosis occurs are poorly understood. The only available estimates of heat death under-diagnosis are based on comparing all-cause mortality during periods of extreme heat and average temperatures and counting all excess deaths as heat deaths. This method has suggested the actual number of heat deaths may be 2 or 3 times higher than the number of officially diagnosed heat deaths.¹⁷ For example, the Centers for Disease Control and Prevention (CDC) calculated that there are on average 618 heat deaths annually in the United States.¹⁸ In contrast, a recent epidemiologic study identified 400 excess deaths during heat events per year in Manhattan alone in the 1980s, with more projected in the 2020s.¹⁹

Improving Heat Death Identification

Heat death estimations are a charged issue since they reflect the capacity of public health institutions to manage climate impacts. Accurately identifying who dies from heat exposure is a critical measure to improve public health preparedness and avert mass casualty events like those that have been occurring in recent history. In addition, mitigation measures depend on a correct identification of who is dying of heat and why they die. Current thinking about heat risk posits that cities are particularly vulnerable to heat waves because of their dense environments that lack greenspace.^20,21 Urban heat islands²¹ occur in cities where temperatures spike 8 to 10 degrees above average.²² Therefore, identifying who is exposed and how these exposures can be mitigated in these areas could play a role in mitigating risk. Heat waves have been thought to disproportionately affect older adults and people of low socioeconomic status;²³ therefore, understanding what the particular risks and exposures are of these groups could prove useful in reducing heat impacts.

Yet, uncovering the reasons why heat deaths may be misclassified is complicated because of the lack of standardized procedures for identifying heat as a cause of death. Medical examiners determine causes of death in collaboration with law enforcement and other experts, conducting scene investigations, performing autopsies and lab tests, triangulating evidence, and applying inductive reasoning and probabilistic decision making.²⁴ Evidence of exposure to heat may be difficult to capture, given that the environmental conditions at the scene may change between the time of death and the discovery of the body, the core body temperature of the deceased may drop over time, and many physical signs detected during an autopsy (eg, skin slippage, pleural petechiae, etc) may be equivocal in the absence of environmental evidence.²⁵ Therefore, it is necessary to identify what types of evidence are commonly collected during the investigation of suspected heat deaths and to assess how different pieces of evidence influence the likelihood of heat death diagnosis.

The objective of our study was to identify evidentiary practices contributing to the underestimation of heat deaths by analyzing medical examiner and vital statistics records for deaths during heat waves in New York City. We describe demographic, socioeconomic, and environmental factors that differ between officially classified heat deaths and deaths that occurred during heat waves but were not officially classified as being due to heat.

Data and Methods

Our study population consisted of decedents who were residents of the 5 boroughs of New York City (the Bronx, Brooklyn, Manhattan, Queens, and Staten Island) during the warm season (May-September) of 2010 to 2012. Subjects were selected from people who died during an identified heat wave period, defined as at least 3 consecutive days with maximum temperatures of at least 90°F,²⁶ and from the 2 days immediately following the heat wave, as previous studies have indicated many health impacts occur within 0 to 3 days following exposure to high temperatures.^4,19-21,27 Daily maximum temperature data were obtained from the LaGuardia Airport Weather Station via the National Climatic Data Center's Climate Data Online search tool.²⁸ We also recorded when in a heat wave each death occurred (ie, day 1, day 2, day 3, etc).

We classified deaths reported by the New York City Office of the Chief Medical Examiner (OCME) as an official heat death or possible heat death on the basis of the International Classification of Diseases, 10^th revision (ICD-10) codes listed for underlying and contributing causes of death on death certificates.²⁹ The CDC defines deaths attributable to heat as those with an underlying cause of death coded as exposure to excessive natural heat (X30), a contributing cause of death coded as due to effects of heat and light (T67), or both. We used this definition to identify all officially diagnosed heat deaths (oHDs). These cases are ones in which heat was clearly a stressor and was acknowledged as such by the medical examiner.

Possible heat deaths (pHD) excluded oHDs and included all other deaths during the same period that were coded as deaths caused by conditions that could have been exacerbated by heat to the point of death. These conditions included all diseases of the circulatory system (I00-I99), volume depletion (E86), acute renal failure (N17), mental and behavioral disorders (F00-F99), cerebral palsy and other paralytic syndromes (G80-G83), and obesity (E65-E68).³⁰ We considered these cases “possible heat deaths” since they represented deaths that were not classified as related to the heat, even though they may have been caused by it. Only deaths investigated by the OCME were eligible to be included, since deaths not investigated by the OCME typically have death certificates available only through the Office of Vital Statistics (OVS), and these documents do not provide sufficient data to assess how the cause of death was identified. Since there is a lack of records for all deaths, we cannot conclude that a lack of records reflects an under-estimation of heat deaths. Rather, it is necessary to conduct further analysis of the records that are available to extrapolate to records that are not included in the database.

We used the records in the medical examiner's files to detail the evidentiary practices used to determine the causes of death and the demographic characteristics of each case.³¹ Prior to our investigation, the precise contents of the OCME's case management system files were unknown, since a detailed description of this database was not publicly available. We identified 4 types of records with evidence useful for assessing how and why heat deaths may be under-diagnosed: (1) the scene investigation form, (2) the supplemental investigation report, (3) the death certificate, and (4) the autopsy report. The scene investigation form consisted of the standardized paperwork completed by the OCME's medicolegal staff responsible for collecting evidence at the scene of death. This form provided details on the discovery, identification, and condition of the body, the surrounding environment, and the decedent's relationship to the scene and past medical history. The supplemental investigation report served as the accompanying narrative description of the scene, the sources used, and the types of evidence collected. The death certificate provided standardized data on socioeconomic and demographic characteristics of the deceased, as well as the location and official causes of death. When an autopsy was performed on the deceased, additional details on the physical condition of the body were contained in the autopsy report.

From the 4 types of records, we obtained detailed data for each of the cases (see Table 1). The variables chosen for analysis were hypothesized to play a role in heat death diagnosis, and they enabled us to assess the types and sources of evidence collected by the OCME and the methods by which the evidence was obtained. Our analysis included variables describing demographic, socioeconomic, and social characteristics of the case, the environmental conditions at the scene of death, bodily evidence of heat exposure, and the basis of the evidence collected by the OCME. Our variables for evidence of heat exposure included the existence of an autopsy report in the deceased's file (yes or no), signs of decomposition (yes or no), and body temperature. We recorded a positive indication of signs of decomposition if any post-mortem body changes were listed in the scene investigation form or described in the supplemental investigation report. Body temperature was extracted from the scene investigation form, where it was reported either quantitatively (°F) or qualitatively (cool or warm). For the environmental conditions at the scene of death, we extracted the place of death, the existence of cooling methods, room temperature (reported quantitatively, °F, or qualitatively, cool or warm), and the presence and use of air conditioning, fans, and windows from the death certificate and scene investigation form.

Table 1.

Variables Extracted from the Death Records for Each Case and Control

Characteristics of Deceased and Investigation in Death Records	Variables Extracted
Demographic determinants of death	Gender (male, female)
	Age (years)
	Race (white, black, Asian, other)
	Ethnicity (non-Hispanic, Hispanic)
Socioeconomic determinants of death	Education (high school or less, more than high school)
Bodily evidence of heat exposure	Body temperature, quantitative (°F)
	Body temperature, qualitative (cool, warm)
	Signs of decomposition (yes, no)
	Autopsy report (yes, no)
Environmental conditions of death	Place of death (home, hospital, outside, nursing home, other)
	Floor of building where body found (floor number)
	Air conditioning (not present, present)
	Fan (not present, present)
	Windows (closed, open)
	Room temperature, quantitative (°F)
	Room temperature, qualitative (cool, warm)
Social determinants of death	Living alone (yes, no)
Basis of the evidence	Police (high, medium, low, not a source)
	Family (high, medium, low, not a source)
	Landlord/neighbor (high, medium, low, not a source)
	Past medical history (high, medium, low, not a source)
	Room temperature method (thermometer, perception)
	Body temperature method (thermometer, touch)

We conducted a bivariate analysis, using Pearson's χ², Fisher's Exact, and Mann-Whitney Rank Sum tests, to compare characteristics of oHDs and pHDs (see Table 2). We also sought to assess the existence of evidence to support the official diagnosis. The OCME's definition requires that at least 2 of the following 3 conditions be met:

• Bodily evidence: “pathologically elevated core body temperature of the decedent, usually >105°F (40.6°C) at the time of or immediately after death”;

• Environmental evidence: “substantial environmental or circumstantial evidence of heat as a contributor to death (e.g., decedent found in a room without air conditioning, all windows closed, and a high ambient temperature)”; and/or

• Other explanations: “decedent in a decomposed condition without evidence of other cause of death.”³²

Table 2.

Bivariate Analysis of Continuous Variables

Characteristic	n	oHDs Median (IQR)	pHDs Median (IQR)	p Value^a
Age (years)	1,093	61 (48-73)	66 (55-78)	0.0202
Body temperature, quantitative (°F)	356	106 (102-108)	92 (86-96)	0.0000
Room temperature, quantitative (°F)	524	92 (90-96)	80 (76-86)	0.0000
Daily maximum temperature (°F)	1,095	92.48 (89.96-98.96)	91.04 (87.98-95.00)	0.0052
Floor of building (floor number)	526	4 (3-6.5)	4 (3-7)	0.7868

Abbreviations: oHDs = officially diagnosed heat deaths, pHDs = possible heat deaths, IQR = interquartile range.

Mann-Whitney Rank Sum test.

We compared the quantity of evidence meeting OCME criteria in oHDs and pHDs. All statistical analyses were performed using Stata^® 12 Data Analysis and Statistical Software (StataCorp, 2011).

Results

Our review of the National Climatic Data Center temperature data revealed that 11 heat wave periods, lasting a total of 93 days, occurred in New York City during the summers (May through September) of 2010, 2011, and 2012. The heat wave periods ranged in length from 5 to 17 days, with the average heat wave lasting 8.5 days (±4.1 days). The results of our inquiry with the Office of Vital Statistics indicated that, during these periods, a total of 1,095 deaths occurred that met the criteria for inclusion in our study. Of these, we classified 1,049 as pHDs and 46 as oHDs.

There were several important differences between oHDs and pHDs. A greater percentage of oHDs (54.35%) than pHDs (37.89%) were black as compared to all other races combined (p = 0.025; n = 1,078). The median age of oHDs (61 years) was significantly younger than the median age of pHDs (66 years) (p = 0.0202; n = 1093).

Overall, there was a lack of evidence to substantiate whether a heat death had occurred. This was especially true of environmental evidence. In both oHDs and pHDs, there was very little data on windows, fans, body temperature, or room temperature. In the case of oHDs, signs of decomposition were reported on in some cases, but 35 of 46 (79.55%) had no signs (Table 3). At least 1 environmental condition was reported for 73% of oHDs, whereas 60% of pHDs had at least 1 condition reported. Ninety-seven percent of oHDs had reported biological evidence (body temp or decomposition), whereas 86% of pHDs had such evidence. Evidence from family members was used much more often than information from police or landlords in both oHDs and pHDs. A scene investigation report was used in 70% of all oHDs and 78% of pHDs, so clearly it is not used to distinguish between them. Medical records were used more frequently than family, policy, or landlords. Only 82.61% of oHDs and 47.57% of pHDs had evidence collected related to 2 or more of the OCME's criteria for identifying a heat death (see Table 2).

Table 3.

Bivariate Analysis of Categorical Variables

Characteristic	oHDs n (%)	pHDs n (%)	p Value
Scene Investigation Form (n = 1,095)
Yes	32 (69.57)	816 (77.79)	0.192^a
No	14 (30.43)	233 (22.21)
Demographic Determinants of Death
Gender (n = 1,095)
Male	29 (63.04)	663 (62.25)	0.913^a
Female	17 (36.96)	396 (37.75)
Race (n = 1,078)
Black	25 (54.35)	391 (37.89)	0.025 ^a
White, Asian, other combined	21 (45.65)	641 (62.11)
Ethnicity (n = 1,053)
Non-Hispanic	37 (84.09)	784 (77.70)	0.317^a
Hispanic	7 (15.91)	225 (22.30)
Socioeconomic Determinants of Death
Education (n = 996)
High school or less	31 (75.61)	646 (67.64)	0.284^a
More than high school	10 (24.39)	309 (32.36)
Bodily Evidence of Heat Exposure
Body temperature, qualitative (n = 228)
Cool	0 (0.00)	154 (67.84)	0.325^b
Warm	1 (100.00)	73 (32.16)
Signs of decomposition (n = 942)
Yes	9 (20.45)	330 (36.75)	0.028 ^a
No	35 (79.55)	568 (63.25)
Autopsy report (n = 1,094)
Yes	40 (86.96)	541 (51.62)	0.000 ^a
No	6 (13.04)	507 (48.38)
Environmental Conditions of Death
Place of death (n = 1,095)
Home	26 (56.52)	740 (70.54)	0.020 ^b
Hospital	19 (41.30)	243 (23.16)
Outside, nursing home, other combined	1 (2.17)	66 (6.29)
Air conditioning (n = 144)
Not present	11 (73.33)	30 (23.26)	0.000 ^b
Present	4 (26.67)	99 (76.74)
Fan (n = 92)
Not present	4 (20.00)	11 (15.28)	0.733^b
Present	16 (80.00)	61 (84.72)
Windows (n = 79)
Closed	7 (53.85)	115 (69.28)	0.250^a
Open	6 (46.15)	51 (30.72)
Room temperature, categorical (n = 121)
Cool	0 (0.00)	24 (20.34)	1.000^b
Warm	3 (100.00)	94 (79.66)
Basis of the Evidence
Police (n = 1,095)
High	2 (4.35)	44 (4.19)	0.156^b
Medium	10 (21.74)	194 (18.49)
Low	2 (4.35)	163 (15.54)
Not a source	32 (69.57)	648 (61.77)
Family (n = 1,095)
High	11 (23.91)	76 (7.24)	0.000 ^a
Medium	12 (26.09)	212 (20.21)
Low	5 (10.87)	157 (14.97)
Not a source	18 (39.13)	604 (57.58)
Landlord/neighbor (n = 1,095)
High	3 (6.52)	13 (1.24)	0.048 ^b
Medium	1 (2.17)	61 (5.82)
Low	6 (13.04)	198 (18.88)
Not a source	36 (78.26)	777 (74.07)
Past medical history (n = 1,095)
High	7 (15.22)	450 (42.90)	0.000 ^a
Medium	22 (47.83)	210 (20.02)
Low	10 (21.74)	204 (19.45)
Not a source	7 (15.22)	185 (17.64)
Body temperature method (n = 587)
Thermometer	29 (93.55)	312 (56.12)	0.000 ^b
Touch	2 (6.45)	244 (43.88)
Room temperature method (n = 625)
Thermometer	26 (89.66)	459 (77.01)	0.168^b
Perception	3 (10.34)	137 (22.99)
Social Determinants of Death
Living alone (n = 635)
Yes	3 (10.71)	303 (49.92)	0.000 ^b
No	25 (89.29)	304 (50.08)
Composite Indicators
Evidence regarding environmental conditions (n = 1,095)
At least one piece of evidence	34 (73.91)	633 (60.34)	0.065^a
None	12 (26.09)	416 (39.66)
Evidence regarding OCME heat death criteria (n = 1,095)
Evidence for fewer than 2 criteria	8 (17.39)	550 (52.43)	0.000 ^a
Evidence for 2 or more criteria	38 (82.61)	499 (47.57)

Abbreviations: oHDs = officially diagnosed heat deaths, pHDs = possible heat deaths, OCME = Office of the Chief Medical Examiner.

Pearson's χ² test.

Fisher's Exact test.

There was also a significant difference in the timing of when oHDs and pHDs occurred during the heat wave periods studied. The median day of death for oHDs was day 9, compared to day 5 for pHDs (p = 0.001).

Discussion

Assuming strict implementation of the official definition of heat stroke death, we anticipated that an official heat death diagnosis would demonstrate statistically significant associations with the existence of data pertaining to each of the 3 criteria. More specifically, we hypothesized that oHDs would demonstrate significantly greater odds than pHDs of having a quantitative body temperature, environmental evidence showing a lack of cooling methods (ie, air conditioning not present, fans not present, windows closed, higher room temperatures), and signs of decomposition. Based on the existing literature on risk factors for heat death, we further hypothesized that oHDs would demonstrate significantly greater odds of the following personal and environmental characteristics: female gender, older age, minority status, lower educational attainment, higher daily maximum and body temperatures, living on higher floors, living alone, and dying at home. However, we anticipated that rigid adherence to either preconceptions of heat death risk or diagnostic criteria could cause some heat deaths to be misclassified.

Given the variable nature and inconsistent frequency with which some of the data were reported in the death records, we also reasoned that the availability and quality of evidence could contribute to heat death under-diagnosis. We, therefore, hypothesized that oHDs would demonstrate significantly greater odds of the existence of a scene investigation form, the existence of an autopsy report, critical evidence from sources (police, family, landlord/neighbor, or past medical history), and quantitative temperature data (body or room) obtained via thermometer.

Our findings suggest several central conclusions. First, evidence to substantiate heat death is not being sufficiently collected, and/or the availability and quality of evidence is likely playing an important role in the under-diagnosis of heat death. Fulfillment of 2 of the 3 criteria specified by New York City to identify a heat death is often impossible. This is particularly the case with regard to environmental evidence. We find that medical evidence is the most heavily used, which would indicate that for cases where medical records, an autopsy, or other medical information is unavailable, a case may be misidentified.

Second, medical examiners may be less likely to collect evidence related to heat exposure for specific kinds of individuals. Older adults are less likely to be oHD and more likely to be a pHD (see Tables 4 and 5). While our data do not indicate why this is the case, it is possible that older deaths are more expected than younger deaths, so they were less frequently investigated and therefore had less evidence to support the death's being caused by the heat. This means that even for older adults who are typically most likely to die of heat death, the risk attributed to heat may be underestimated. Similarly, black people are more likely to be an oHD than a pHD. It could be that investigators and medical examiners expect black people are the more typical case of heat death and collect evidence to support that. In sum, social norms and biases could play a role in evidence gathering.

Table 4.

Initial logistic regression models constructed to assess the relative importance of the Office of Chief Medical Examiner's diagnostic criteria and the use of various sources of evidence to identification of heat deaths

Predictor Variable	Model 1 OR (95% CI)	Model 2 OR (95% CI)	Model 3 OR (95% CI)	Model 4 OR (95% CI)
OCME heat death criteria^a
Body temperature, quantitative	15.74 (5.87, 42.23)
Environmental evidence	0.39 (0.15, 1.04)
Signs of decomposition	0.52 (0.24, 1.15)
Records available^b
Scene investigation form		0.63 (0.31, 1.30)	0.74 (0.37, 1.46)	0.75 (0.39, 1.45)
Autopsy report		6.07 (2.54, 14.49)	6.07 (2.52, 14.62)	6.13 (2.57, 14.60)
Sources mentioned (high, medium, or low usage rankings)^c
Police		0.96 (0.49, 1.91)
Family		2.27 (1.21, 4.28)
Landlord/neighbor		1.03 (0.48, 2.20)
Past medical history		1.13 (0.48, 2.63)
Sources of critical evidence (high usage ranking only)^d
Police			0.80 (0.17, 3.63)
Family			5.05 (2.29, 11.16)
Landlord/neighbor			3.70 (0.79, 17.45)
Past medical history			0.22 (0.09, 0.50)
At least one				0.90 (0.49, 1.65)
n	1,095	1,094	1,094	1,094
Hosmer-Lemeshow χ² (p value)	17.20 (0.0041)	14.27 (0.0750)	18.19 (0.0027)	5.51 (0.3571)

Abbreviations: OR = odds ratio, CI = confidence interval.

Existence of evidence for criterion compared to absence of evidence for criterion.

Existence of record compared to absence of record.

High, medium, and low usage rankings combined compared to “not a source.”

High usage ranking compared to medium, low, and “not a source” combined.

Table 5.

Logistic regression models constructed to improve the fit with the data and to identify the most effective set of predictors of heat death diagnosis

Predictor Variable	Model 5 OR (95% CI)	Model 6 OR (95% CI)	Model 7 OR (95% CI)	Model 8 OR (95% CI)
Significant variables from models 1-4^a
Body temperature, quantitative^b	13.65 (1.50, 123.98)	264.87 (1.14, 61809.23)	16.30 (1.87, 142.04)	14.46 (1.74, 120.07)
Autopsy report^c	25.47 (0.86, 752.18)	515.73 (1.55, 1.72×10⁵)	40.86 (1.37, 1215.79)	29.28 (1.48, 579.66)
Source of critical evidence: Family^d	37.16 (2.60, 531.35)	33,860.92 (2.28, 4.14×10⁸)	24.63 (2.45, 247.30)	22.58 (2.37, 215.22)
Source of critical evidence: Past medical history^d	0.15 (0.01, 2.46)
Additional variables
Age^e	0.94 (0.88, 1.01)	0.88 (0.79, 0.98)	0.93 (0.88, 0.99)	0.94 (0.88, 0.99)
Race^f	0.57 (0.11, 2.94)
Education^g		323.94 (0.45, 2.32×10⁵)
Presence of air conditioning^h	0.07 (0.01, 0.53)	1.76×10⁻⁵ (1.72×10^−9, 0.18)	0.04 (0.01, 0.26)	0.04 (0.01, 0.25)
Daily maximum temperature (°F)^e			1.06 (0.90, 1.24)
n	144	129	144	144
Hosmer-Lemeshow χ² (p value)	8.27 (0.4080)	0.98 (0.9984)	1.17 (0.9969)	1.15 (0.9971)

Abbreviations: OR = odds ratio, CI = confidence interval.

Family mentioned as a source was not included as a predictor variable in order to avoid multicollinearity with family as a source of critical evidence.

Existence of evidence compared to absence of evidence.

Existence of record compared to absence of record.

High usage ranking compared to medium, low, and “not a source” combined.

Continuous variable.

More than high school compared to high school or less.

Black compared to white, Asian, and other combined.

Presence compared to absence.

Third, place of death may have played a role in the availability of different types of evidence. For instance, oHDs were more likely to have died in a hospital and to have an autopsy report on file, suggesting that these 2 characteristics might be related. Since pHDs were significantly more likely to have died at home, medical examiners may have been less inclined to perform autopsies for these deaths. We also noticed, informally, that deaths in hospitals seemed not to receive scene investigations.

Based on these findings, we recommend that medical examiners and coroners train their staff to collect evidence more accurately and comprehensively in order that cases can be properly identified. This training could be used to raise awareness of preconceptions of heat death risk that may influence diagnostic outcomes, stress the importance of the collection of environmental evidence, and help promote consistent evidentiary standards regardless of the decedent's place of death.

Limitations

The main limitations of our study include the temperature measure we used to identify heat waves, the lack of direct observation of medical examiners' evidentiary practices, and the restricted external validity of our results. We used New York City's definition of a heat wave, which is based on daily maximum ambient temperatures. Using a single maximum temperature for all study subjects does not account for microscale variations in heat exposure due to differences in urban geometry, differences in indoor versus outdoor temperature, and the distribution of green space across the city. Maximum temperature, in comparison measures such as the heat index, also does not incorporate the effect of humidity. We may have excluded from our study days where the perceived temperature was 90°F or higher even though the maximum real temperature did not cross this threshold. This limitation may have reduced our study sample and the statistical power of our analyses.

We used records from the OCME's Case Management System database to extract data on all of our predictor variables, and we did not perform direct observations of medical examiners engaged in evidence collection and evaluation processes. Consequently, we were not able to compare the data in official death records to medical examiners' evidentiary and decision-making practices in the field. Any error in the classification of the characteristics of our study subjects because of this lack of validation would have most likely been nondifferential and would have biased associations with heat death diagnosis toward the null. Lastly, the absence of consistent diagnostic criteria across medical examiners' offices means our results are applicable only to New York City.

Conclusions

New York City, like many other cities across the country, faces the challenge of effectively identifying deaths from heat exposure in order to mitigate heat risk. In order for public health leaders to develop appropriate preparedness and response plans for extreme heat events, it is essential that they have an accurate understanding of who is at risk. Our analysis offers specific suggestions for how localities may adjust the training of medical examiners to improve identification of heat deaths. Additional research involving interviews with medical examiners is needed to further clarify how they value and interpret specific types of evidence and to refine our assessment of the factors contributing to heat death under-diagnosis. Once improved assessments of heat risk are developed, preparedness measures can match these risks.

Additional research is needed to understand how medical examiners approach the scene of a death, what types of evidence they prioritize during their investigations, and how they weight different pieces of evidence when diagnosing the cause of death. In-depth interviews and shadowing of medical examiners may reveal that other factors not considered during our study play an important role in heat death diagnosis. Our methodology can be revised to focus on types of evidence that most accurately reflect the medical examiners' decision-making process, and it can be used to assess heat death under-diagnosis in other locations and to identify best practices for heat death identification. As estimates of the incidence of heat death are improved, they can be used to refine analyses of trends in heat deaths over time and develop more accurate projections of how climate change may influence mortality.

References

National Weather Service. Natural hazards statistics. Weather fatalities. NOAA website. August 1, 2012. http://www.nws.noaa.gov/om/hazstats.shtml. Accessed March 14, 2016.

O'Neill

, Hajat

, Zanobetti

, Ramirez-Aguilar

, Schwartz

. Impact of control for air pollution and respiratory epidemics on the estimated associations of temperature and daily mortality. Int J Biometeorol, 2005; 50(2):121-129.

Bouchama

, Knochel

. Heat stroke. N Engl J Med, 2002; 346(25):1978-1988.

Halonen

, Zanobetti

, Sparrow

, Vokonas

, Schwartz

. Relationship between outdoor temperature and blood pressure. Occup Environ Med, 2011; 68(4):296-301.

IPCC. Summary for policymakers. In: Field

, Barros

, Stocker

, et al, eds. Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation. Cambridge, UK, and New York: Cambridge University Press; 2012:1-19.

McGeehin

, Mirabelli

. The potential impacts of climate variability and change on temperature-related morbidity and mortality in the United States. Environ Health Perspect, 2001; 109(Suppl 2):185-189.

Ebi

, Meehl

. Heatwaves & Global Climate Change: The Heat Is On: Climate Change & Heatwaves in the Midwest. Arlington, VA: Pew Center on Global Climate Change; 2007.

America's Climate Choices: Panel on Advancing the Science of Climate Change; Board on Atmospheric Sciences and Climate; Division on Earth and Life Studies; National Research Council. Advancing the Science of Climate Change. Washington, DC: National Academies Press; 2010.

Klinenberg

. Heat Wave: A Social Autopsy of Disaster in Chicago. Chicago, IL: University of Chicago Press; 2002, 2015.

10.

Poumadère

, Mays

, Le Mer

, Blong

. The 2003 heat wave in France: dangerous climate change here and now. Risk Anal, 2005; 25(6):1483-1494.

11.

Kovats

, Edwards

, Hajat

, Armstrong

, Ebi

, Menne

. The effect of temperature on food poisoning: a time-series analysis of salmonellosis in ten European countries. Epidemiol Infect, 2004; 132(3):443-453.

12.

Nicholls

, Skinner

, Loughnan

, Tapper

. A simple heat alert system for Melbourne, Australia. Int J Biometeorol, 2008; 52(5):375-384.

13.

Osborn

. Moscow smog and nationwide heat wave claim thousands of lives. BMJ, 2010; 341:c4360.

14.

Samra

. Changes in extreme climate events and their management in India. In: Solh

, Saxena

, eds. Food Security and Climate Change in Dry Areas. Aleppo, Syria: International Center for Agricultural Research in the Dry Areas; 2010:13-23.

15.

Shen

, Howe

, Alo

, Moolenaar

. 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, 1998; 19(2):113-118.

16.

Basu

, Samet

. Relation between elevated ambient temperature and mortality: a review of the epidemiologic evidence. Epidemiol Rev, 2002; 24(2):190-202.

17.

Ostro

, Roth

, Green

, Basu

. Estimating the mortality effect of the July 2006 California heat wave. Environ Res, 2009; 109(5):614-619.

18.

Centers for Disease Control and Prevention. QuickStats: number of heat-related deaths, by sex—National Vital Statistics System, United States, 1999-2010. MMWR Morb Mortal Wkly Rep, 2012; 61(36):729.

19.

, Horton

, Kinney

. Future projections of seasonal patterns in temperature-related deaths for Manhattan. Nat Clim Chang, 2013; 3:717-721.

20.

Luber

, McGeehin

. Climate change and extreme heat events. Am J Prev Med, 2008; 35(5):429-435.

21.

Metzger

, Ito

, Matte

. Summer heat and mortality in New York City: how hot is too hot?. Environ Health Perspect, 2010; 118(1):80-86.

22.

Hodges

, Parshall

, Rosenzweig

, Solecki

. Skin of the Big Apple: Characterizing the Surface Heat Island of New York City & Integration with MM5 Climate Model. Presentation on EPA conference call. January 26, 2005. https://www.epa.gov/sites/production/files/2014-07/documents/surfaceuhi_epa4.pdf. Accessed March 15, 2016.

23.

Morello-Frosch

, Pastor

, Sadd

, Shonkoff

. The Climate Gap: Inequalities in How Climate Change Hurts Americans & How to Close the Gap. Berkeley, CA: University of California Press; 2009.

24.

Timmermans

. Suicide determination and the professional authority of medical examiners. Am Sociol Rev, 2005; 70:311-333.

25.

Byard

, Tsokos

. The challenges presented by decomposition. Forensic Sci Med Pathol, 2013; 9(2):135-137.

26.

New York City Office of Emergency Management. 2014. NYC hazards: Extreme heat basics.

27.

Corobov

, Sheridan

, Opopol

, Ebi

. Heat-related mortality in Moldova: the summer of 2007. Int J Climatol, 2013; 33(11):2551-2560.

28.

National Centers for Environmental Information. Climate data online. NOAA website. 2014. http://www.ncdc.noaa.gov/cdo-web/. Accessed March 17, 2016.

29.

Centers for Disease Control and Prevention. International Classification of Diseases, tenth revision, clinical modification (ICD-10-CM). Update October 29, 2015. http://www.cdc.gov/nchs/icd/icd10cm.htm. Accessed March 17, 2016.

30.

Basu

, Pearson

, Malig

, Broadwin

, Green

. The effect of high ambient temperature on emergency room visits. Epidemiology, 2012; 23(6):813-820.

31.

NYC Office of Chief Medical Examiner. Authority and responsibilities. NYC website. 2014. http://www.nyc.gov/html/ocme/html/about/authority.shtml. Accessed March 5, 2014.

32.

New York City Department of Health and Mental Hygiene. Deaths associated with heat waves in 2006. NYC Vital Signs Investigation Report. November 2006. http://www.nyc.gov/html/doh/downloads/pdf/survey/survey-2006heatdeaths.pdf. Accessed March 17, 2016.