Identifying heat-related deaths by using medical examiner and vital statistics data: Surveillance analysis and descriptive epidemiology — Oklahoma, 1990–2011

Environmental Research 150 (2016) 30–37

Contents lists available at ScienceDirect

Environmental Research journal homepage: www.elsevier.com/locate/envres

Identifying heat-related deaths by using medical examiner and vital statistics data: Surveillance analysis and descriptive epidemiology — Oklahoma, 1990–2011 Matthew G. Johnson a,b,n, Sheryll Brown c, Pam Archer c, Aaron Wendelboe d, Sheryl Magzamen d,e, Kristy K. Bradley f a

Epidemic Intelligence Service, Centers for Disease Control and Prevention, Atlanta, GA, USA Acute Disease Service, Oklahoma State Department of Health, Oklahoma City, OK, USA Injury Prevention Service, Oklahoma State Department of Health, Oklahoma City, OK, USA d College of Public Health, University of Oklahoma Health Sciences Center, Oklahoma City, OK, USA e College of Veterinary Medicine and Biomedical Sciences, Colorado State University, Fort Collins, CO, USA f Office of the State Epidemiologist, Oklahoma State Department of Health, Oklahoma City, OK, USA b c

art ic l e i nf o

a b s t r a c t

Article history: Received 11 September 2015 Received in revised form 3 May 2016 Accepted 19 May 2016

Objectives: Approximately 660 deaths occur annually in the United States associated with excess natural heat. A record heat wave in Oklahoma during 2011 generated increased interest concerning heat-related mortality among public health preparedness partners. We aimed to improve surveillance for heat-related mortality and better characterize heat-related deaths in Oklahoma during 1990–2011, and to enhance public health messaging during future heat emergencies. Methods: Heat-related deaths were identified by querying vital statistics (VS) and medical examiner (ME) data during 1990–2011. Case inclusion criteria were developed by using heat-related International Classification of Diseases codes, cause-of-death nomenclature, and ME investigation narrative. We calculated sensitivity and predictive value positive (PVP) for heat-related mortality surveillance by using VS and ME data and performed a descriptive analysis. Results: During the study period, 364 confirmed and probable heat-related deaths were identified when utilizing both data sets. ME reports had 87% sensitivity and 74% PVP; VS reports had 80% sensitivity and 52% PVP. Compared to Oklahoma's general population, decedents were disproportionately male (67% vs. 49%), aged Z 65 years (46% vs. 14%), and unmarried (78% vs. 47%). Higher rates of heat-related mortality were observed among Blacks. Of 95 decedents with available information, 91 (96%) did not use air conditioning. Conclusions: Linking ME and VS data sources together and using narrative description for case classification allows for improved case ascertainment and surveillance data quality. Males, Blacks, persons aged Z65 years, unmarried persons, and those without air conditioning carry a disproportionate burden of the heat-related deaths in Oklahoma. Published by Elsevier Inc.

Keywords: Heat Mortality Vital statistics Medical examiner Oklahoma

1. Introduction A record heat wave in the southcentral United States during summer 2011 generated increased interest in heat-related mortality among public health officials in Oklahoma. The Centers for Disease Control and Prevention (CDC) estimates approximately 660 deaths occur annually in the United States directly associated n Correspondence to: Infectious Diseases Fellow, Duke University Medical Center, DUMC 102539, 315 Trent Drive, Durham, NC, 27710 USA. E-mail address: [email protected] (M.G. Johnson).

http://dx.doi.org/10.1016/j.envres.2016.05.035 0013-9351/Published by Elsevier Inc.

with excess natural heat (CDC, 2013). This figure is likely a substantial underestimate of the actual number of deaths as no widely accepted consensus exists regarding specific criteria for defining heat-related mortality, deaths from other causes known to be exacerbated by heat (e.g., cardiovascular disease) might not be classified as heat related, and a medical examiner (ME) might not consider heat as a cause or contributor of death during non-heat wave periods (Basu and Samet, 2002; Ostro et al., 2009). Estimates of 22,000
45,000 heat-related deaths across Europe associated with a two week heat wave during August 2003 underscore the effects of excessive heat on human health (Basara et al., 2010; Patz

M.G. Johnson et al. / Environmental Research 150 (2016) 30–37

et al., 2005). Among the most vulnerable populations to heat-related illness are older adults and infants, persons with cardiovascular and psychiatric disease, bedbound persons, and those with lower socioeconomic status; the strongest protective factor is having a working air conditioner in the home (Basu, 2009; Bouchama et al., 2007; Naughton et al., 2002; Semenza et al., 1996). If heat preparedness partners target precautionary interventions to susceptible populations and behaviors, heat-related morbidity and mortality can be reduced among those at greatest risk. Climate model projections of hotter, longer, and more frequent heat waves indicate greater scrutiny is needed regarding this problem (Davis et al., 2003) Heat-related illness and death are preventable, and simple steps can save lives during hot summer months (CDC, 2005). In Oklahoma, where the climate ranges from humid subtropical in the east to semiarid in the west (Oklahoma Climatological Survey, 2016), the Office of the Chief Medical Examiner (OCME) is responsible for investigating all sudden, violent, unexpected, unattended, and suspicious deaths, including heat-related deaths. The OCME works with law enforcement officials to conduct a medicolegal investigation of circumstances concerning the death, and the ME completes sections of the death certificate that indicate the cause, manner, and circumstances of death. Thus only a small fraction of all deaths are evaluated by the OCME. Therefore most death certificates are either completed by clinicians in the hospital or funeral directors and are forwarded to the vital statistics (VS) division of the Oklahoma State Department of Health (OSDH), where trained personnel enter the cause and manner of death into a VS database (Oklahoma State Department of Health, 2016). The cause of death is later coded according to the International Classification of Diseases, Tenth Revision (ICD-10) by using nosology software (WHO, 2007). Heat-related mortality literature often relies on ICD codes to identify heat-related deaths, which likely underestimates the actual burden of disease. However, to better understand the true scope of heat-related mortality in Oklahoma, we employed a novel approach wherein we reviewed ME data along with VS data to identify heat-related deaths. Narrative descriptions in ME records were used for both case finding and case confirmation. Using ME narratives identified additional cases not found by using ICD codes alone and also served as the ultimate standard by which the sensitivity and specificity of our data sources were measured, providing additional confidence in the accuracy of our cases meeting the case definition. In this study, all heat-related deaths in Oklahoma during 1990– 2011 identified by using Oklahoma's ME database and VS records were compiled and reviewed. Specific heat-related ICD codes, cause-of-death nomenclature, or narrative description of the circumstances concerning death were reviewed to determine case inclusion and yield a more thorough approximation of heat-related deaths. To our knowledge, the OCME has used a consistent reporting practice for heat-related mortality during the study period. A comparison of the sensitivity and predictive value positive (PVP) for identifying heat-related mortality cases by using ME or VS data alone was performed. A descriptive epidemiologic analysis was also performed of heat-related mortality in Oklahoma, a region whose population is hypothesized to be more acclimatized to higher temperatures and thus at lower risk for heatrelated mortality as evidenced by Anderson and Bell's (2009) demonstration that Oklahoma cities are in the bottom 20% of cities nationwide ranked by heat impact on all-cause mortality. Results of this study will be used to guide public health messaging and early warning heat advisories from OSDH and other heat preparedness partners during future heat waves. A similar epidemiologic evaluation and prevention strategy might be considered

31

in regions of the world with comparable risk for heat-related morbidity and mortality.

2. Materials and methods According to the National Association of Medical Examiners, a heat-related death is a death in which exposure to high ambient temperature either caused the death or was a substantial contributor. Heat-related death is based on a history of exposure to high ambient temperature and the reasonable exclusion of other causes of hyperthermia. The diagnosis can be established from circumstances concerning the death, investigative reports concerning environmental temperatures, or a measured antemortem body temperature at the time of collapse ( Z105°F or lower if cooling was attempted before arrival at the hospital or a clinical history of mental status changes and elevated liver or muscle enzymes is noted) (Donoghue et al., 1997). Heat-related deaths were identified and classified by using a tiered strategy on the basis of specific heat-related ICD codes, cause-of-death nomenclature, and the narrative description of circumstances concerning the death gathered from VS and ME databases. VS data include ICD codes and cause-of-death nomenclature; ME data include cause-of-death nomenclature and a narrative description of the circumstances concerning death from scene investigation reports or final autopsy reports. For certain cases, ME data were unavailable or only partial ME data were available without a final autopsy report. Queries of VS data for specific heat-related International Classification of Diseases, Ninth Revision (ICD-9) and ICD-10 codes during 1990–2011 in Oklahoma were performed; the ICD-9 coding scheme changed to ICD-10 in 1999 (WHO, 2007). Non-specific heat-related ICD codes such as “fever of other and unknown origin” (ICD-10 R50), in addition to environmental heat-specific heat-related ICD codes such as “exposure to excessive natural heat” (ICD10  30), were included to ensure broad case ascertainment (Supplemental Material, Table S1). Heat-specific ICD codes did not have to be the primary ICD codes relating to the cause of death; codes could be in any position in the VS database for inclusion in the study. VS and ME data were also queried for specific heat-related cause-of-death nomenclature, namely heat, hot, hyperthermia, and sun. After generating a list of all potential heat-related deaths from the ME and VS databases, the narrative description of the circumstances concerning death was used to classify cases as confirmed, probable, or not a case. Confirmed cases had a narrative description that was compelling for a heat-related death, along with heat-related ICD codes or cause-of-death nomenclature. Probable cases included (a) a narrative description that was compelling for a heat-related death but did not have any heatrelated ICD codes or cause-of-death nomenclature or (b) a narrative description that was unclear, but a heat-related ICD code or cause-of-death nomenclature was present. In addition, if ME data were unavailable for confirmation, two inclusion criteria were required to be considered a probable heatrelated death. First, the death must have occurred during the warmer months of May–September. Second, environmental heatspecific heat-related ICD codes had to be present. Cases with a narrative description that was not compelling for a heat-related death or a narrative description that was unclear without any heat-related ICD codes or cause-of-death nomenclature were excluded. Case classification was reviewed and validated by OSDH public health officials. Complete case classification criteria are included in the text box. Classification criteria for confirmed, probable, and non-cases of heat-related death in Oklahoma, 1990–2011.

32

M.G. Johnson et al. / Environmental Research 150 (2016) 30–37

Confirmed: ME narrative description was compelling for a heat-related death, and heat-related ICD codes or cause-ofdeath nomenclature was present.  Final autopsy report concluded a heat-related death.  If no autopsy was performed or no final autopsy report was available, then the ME narrative description must be compelling for heat-related death; also, must contain heat-specific ICD code or heat-related ME nomenclature.  If no VS report was available, the ME narrative description must be compelling for a heat-related death and must contain heatrelated ME nomenclature. Probable: ME narrative description was compelling for a heatrelated death but heat-related ICD codes or cause-of-death nomenclature were not present, or the ME narrative description was unclear but heat-related ICD codes or cause-ofdeath nomenclature was present.  If final autopsy report was unclear or unavailable, or no autopsy was performed, additional information was considered as follows: ○ ME narrative description was compelling for a heat-related death, but no heat-specific ICD code and no heat-related ME nomenclature was present. ○ Heat-specific ICD code or heat-related ME nomenclature was present, but ME narrative description for heat-related death was unclear.  If no VS report was available, ME narrative description must be compelling for a heat-related death or must contain heat-related ME nomenclature. Not a Case:  Final autopsy report concluded death was not heat-related.  If no autopsy was performed or no final autopsy report was available, additional information was considered as follows: ○ ME narrative description was not compelling for heat-related death, regardless of ICD code and ME nomenclature. ○ Heat-specific ICD code and heat-related ME nomenclature were not present, with an unclear ME narrative description for heat-related death.  Death was because of heat exposure of a manmade origin.  Death was involving fever not related to ambient temperature.  Death was considered to have been due to malignant hyperthermia.  Death was associated with suicide or homicide.  Case had involved coding error (e.g., miscoded as hyperthermia instead of hypothermia).

Variables, including demographic information, time and location of heat injury, education, occupation or industry, alcohol or drug screen results, past medical history, medications, and air conditioner use, were abstracted from ME and VS reports. We used marital status as a marker for social isolation. Additional details were obtained from the ME narrative description. All cases were entered using Epi Info™ 7.0 software (Centers for Disease Control and Prevention, Atlanta, Georgia, USA). Census estimates were used to compare demographic variables among decedents with Oklahoma's total population (US Census Bureau, 2013). Analyses s were conducted by using SAS Version 9.2 (SAS Institute, Cary, North Carolina, USA). Confirmed and probable heat-related deaths were grouped for analysis. Sensitivity and PVP of using ME data alone and VS data alone for heat-related mortality surveillance were calculated. Out of the total number of heat-related deaths using both databases, sensitivity was defined as the proportion of confirmed and probable heat-related deaths identified as heat-related deaths by using

either ME or VS data alone (true positives divided by the sum of true positives and false negatives). PVP was defined as the proportion of heat-related deaths identified by using either ME or VS data that were confirmed or probable heat-related deaths (true positives divided by the sum of true positives and false positives). The standard used in our analysis was the sum of confirmed and probable cases identified with both databases as defined by the aforementioned criteria.

3. Results 3.1. Surveillance analysis A total of 719 potential heat-related deaths were detected by searching ME and VS databases during 1990–2011; 364 heat-related deaths were identified (163 confirmed cases and 201 probable cases), and 355 deaths were not considered heat-related deaths and excluded from analysis (Supplemental Material, Table S2). Searching with ME reports alone detected a total of 430 potential heat-related deaths; 318 were cases, resulting in a sensitivity and PVP of 87% (318/364) and 74% (318/430), respectively. Searching with VS reports alone detected a total of 557 potential heat-related deaths; 292 were cases, resulting in a sensitivity and PVP of 80% (292/364) and 52% (292/557), respectively. Among the 364 total cases, 246 (68%) were discovered by both VS and ME query, 72 (20%) by ME query alone, and 46 (13%) by VS query alone. Heat-related cause-of-death nomenclature was present in 340 (93%) heat-related deaths, but a heat-specific ICD code was present in only 246 (68%) deaths. The most common heat-related cause-of-death nomenclature were hyperthermia and heat, which were documented in 202 (55%) and 140 (38%) heat-related deaths, respectively. These categories were not exclusive from one another; some decedents may have had both hyperthermia and heat used in the VS and ME databases. The most common ICD codes were E900 (ICD-9) and X30 (ICD-10) (excessive heat) and E992 (ICD-9) and T67 (ICD-10) (effects of heat and light) and were represented in 235 (65%) and 122 (34%) heat-related deaths, respectively. 3.2. Descriptive epidemiology A total of 364 confirmed and probable heat-related deaths occurred in Oklahoma during 1990–2011. The years with the highest rate of heat-related deaths were 1998, 2006, and 2011, with a steadily increasing trend line during the past two decades (Fig. 1). Of 364 confirmed and probable heat-related deaths, 168 (46%) deaths occurred during July; 60 (16%) deaths occurred during April–June, and 136 (37%) deaths occurred during August–September (Fig. 2). The total number of confirmed and probable heat-related deaths by age and sex categories is displayed in Fig. 3. Basic demographic information and heat-related mortality rates for all confirmed and probable heat-related deaths are listed in Table 1. Heat-related mortality rates were highest among males, Blacks, and persons aged Z 65 years. The rates of heat-related deaths steadily increased with age, with the highest rates occurring among persons Z85 years. The relative risk of heat-related deaths among decedents Z 65 years compared to persons o65 years was 5.4 (95% CI 4.4–6.7); the relative risk among Black versus white race was 1.7 (95% CI 1.2–2.2). Among 346 decedents aged Z 16 years (legal age of marriage in Oklahoma), 270 (78%) were single, divorced, widowed, or separated. Among all decedents, 46% were aged Z65 years (range, 2 months–95 years); 67% were male; and 78% were unmarried (versus 14%, 49%, and 47%, respectively, for

M.G. Johnson et al. / Environmental Research 150 (2016) 30–37

1.6

Table 1. Demographic characteristics of confirmed and probable heat-related deaths in Oklahoma, 1990–2011 (n¼ 364).

1.4

Rate per 100,000 persons

33

1.2

No. (%)

Rate per 100,000 persons

Gender Male Female

245 (67) 119 (33)

0.65 0.31

Race White Black American Indian Asian

286 (79) 47 (13) 30 (8) 1 ( o1)

0.46 0.76 0.43 0.08

Hispanic

15 (4)

0.35

Age at death, by age groups (years) 0–4 5–18 19–50 51–64 65–74 75–84 Z 85

12 (3) 9 (2) 95 (26) 82 (23) 45 (12) 79 (22) 42 (12)

0.23 0.06 0.28 0.74 0.82 2.31 3.44

62 years (range, 2 months–95 years)

–

1 0.8 0.6 0.4 0.2 0 1990

1993

1996

1999

2002

2005

2008

2011

Year Fig. 1. Rate of confirmed and probable heat-related deaths by year (with trend line represented by dotted line) in Oklahoma, 1990–2011. 180 160

Number of deaths

140 120 100 80 60 40 20

Age at death (median)

0 April

May

June

July

August

September

Month Fig. 2. Number of confirmed and probable heat-related deaths by month of death* in Oklahoma, 1990–2011 (n¼ 364). *Probable case definition restricted occurrence of cases to May–September if Medical Examiner data was not available for confirmation. 100 90

Number of Deaths

80 70

Male

Female

60 50 40 30 20 10 0 0–4

5–9

10–14

15–24

25–34

35–44

45–54

55–64

65+

Age Categories Fig. 3. Number of confirmed and probable heat-related deaths by age and sex in Oklahoma, 1990–2011 (n ¼364).

Oklahoma overall). Among 95 decedents for whom explicit information was available from ME reports, 91 (96%) did not use air conditioning. Thirty-five (10%) heat-related deaths were work-related. The most common industries were construction (n ¼8), agriculture (n ¼6), oil and gas (n ¼6), and landscaping (n ¼4). The most common occupations were laborer (n ¼7), farmer or rancher (n ¼ 6), oilfield worker (n ¼5), and truck driver (n ¼4). Among all deaths, 152 (41%) occurred inside a home, and 41 (11%) involved exposure in an enclosed vehicle (Table 2). Among all adults found in vehicles, median age was 56 years (range, 20–89 years) and included younger decedents with suspected substance abuse as well as older decedents with chronic medical conditions.

On the basis of probable cause of death and other past medical history from VS and ME reports, the most common medical comorbidities, by body system, of the decedents were as follows: cardiac (47%), psychiatric (27%), neurologic (16%), pulmonary (10%), and endocrine (9%) (Supplemental Material, Table S3). The most common medications, by medication class, listed in decedent's ME report were cardiac (46%), psychiatric (38%), pain or anti-inflammatory (35%), endocrine (21%), and neurologic (15%) (Supplemental Material, Table S4). Among 169 decedents with documented body mass index (BMI) aged Z 20 years, 50 (30%) were overweight (BMI 25.0–29.99 kg/m2) and 37 (22%) were obese (BMI Z30.0 kg/m2). Median BMI was 25.4 kg/m2. Among decedents aged Z14 years, 224 ethyl alcohol screens were performed; 35 (16%) were positive. A total of 102 drug screens were performed; 58 (57%) were positive (Supplemental Material, Fig. S1). Family members were most likely to initially inquire about the decedent's well-being, as evidenced by 99 (49%) of 204 cases where the initial party who instigated the investigation was a family member; the initial party who instigated the investigation was unknown for 160 decedents. Other parties to inquire about the decedent's well-being were neighbors (15%), friends (10%), colleagues (8%), police (1%), and mail carriers (1%).

4. Discussion 4.1. Surveillance analysis Heat-related illness represents a spectrum of physiologically diverse conditions that can affect every system in the body (Becker and Stewart, 2011). The definition of a heat-related death is multifactorial and often lacks anatomic evidence; therefore, capturing a true record of all heat-related mortality in a population is difficult (Shen et al., 1998). Some previous studies of heat-related

34

M.G. Johnson et al. / Environmental Research 150 (2016) 30–37

Table 2. Confirmed and probable heat-related deaths, by type of exposure or activity, in Oklahoma, 1990–2011 (n¼ 364). Type of exposure or activity In vehicle Adult found in vehicle Child left in vehicle Child accidentally locked self in vehicle Adult incapacitated in vehicle because of significant medical condition Person locked or left in vehicle, history of cerebral palsy or mental retardation Unknown

Inside home House or home Residence, not otherwise specified Trailer or mobile home Apartment Nursing home Prison or jail cell Unknown

No. of deaths (%) 41 (11) 19 (5) 7 (2) 7 (2) 4 (1) 3 (1) 1 (o 1)

152 (42) 93 (26) 31 (9) 12 (3) 8 (2) 4 (1) 3 (1) 1 (o 1)

Outside home Found in yard Found in driveway Known fall outside residence Found on porch or patio Found in garden Sunbathing outside residence Unknown

35 (10) 12 (3) 4 (1) 4 (1) 4 (1) 3 (1) 2 (o 1) 6 (2)

Inside other Warehouse Friend's house Gas station bathroom Motel

5 (1) 2 (1) 1 (o 1) 1 (o 1) 1 (o 1)

Outside other Found along roadway Found in pasture or field Found in yard Found along railroad track Found in forest or woods Found by lake or pond Other

51 (14) 13 (4) 11 (3) 5 (1) 3 (1) 2 (1) 2 (1) 15 (4)

Type of exertion Mowing lawn Yardwork Farming-related activities Oil industry-related activities Construction Roofing High school sporting event Working on car or truck Military or firefighting training Other

51 (14) 10 (3) 10 (3) 5 (1) 4 (1) 4 (1) 3 (1) 3 (1) 3 (1) 3 (1) 6 (2)

Unknown

Total

29 (8)

364 (100)

illness have only relied on searching for specific heat-related ICD code or cause-of-death nomenclature; this technique is subject to coding error and omissions that do not provide an accurate estimate of the overall burden of heat-related death (Ostro et al., 2009; Bouchama et al., 2007; Naughton et al., 2002; Semenza et al., 1996). The diagnosis of heat-related mortality primarily is based on investigative information; ICD codes and cause-of-death nomenclature alone are often inadequate (Donoghue et al., 1997). In contrast, many analyses do not even attempt to classify heatrelated ICD codes or cause-of-death nomenclature and rather rely on surrogate markers such as total mortality or elderly mortality to estimate the true effect of heat on mortality (Harlan et al., 2014; Honda et al., 2014). As heat as a risk factor for death may have never come to the attention of the clinician or the person completing the death certificate, studies which rely on heat-specific ICD codes likely underestimate the true burden of heat-related mortality. However, studies using elderly mortality or cause-specific mortality (e.g., cardiovascular disease mortality) are less specific than studies using heat-related codes and consequently suffer from false positives. Our primary aim was to create an improved reference standard by using both VS and ME data instead of relying on VS data alone. As such, we implemented a novel and comprehensive case finding and case classification approach to ascertain a more reliable estimate of heat-related deaths in Oklahoma by employing rigorous methods to identify confirmed and probable cases and exclude miscoded cases. Mortality data from ME and VS sources have been compared in general and for other public health surveillance purposes (e.g., injury surveillance), but no studies have examined the correlation between ME and VS coding of heat-related deaths (Comstock et al., 2005; Graitcer et al., 1987). Despite inherent problems with these data sources, including selection and information bias, data entry errors, limited data accuracy and timeliness, and inadequate nosology training, both remain useful and fundamental data sources for public health surveillance (Comstock et al., 2005; Graitcer et al., 1987; Hanzlick and Parrish, 1996; Moyer et al., 1989). Linking the ME and VS databases together allows for more complete case ascertainment and higher data quality than using any single data source (Comstock et al., 2005; Hanzlick and Parrish, 1996). In Oklahoma, ME reports had higher sensitivity and PVP for detecting heat-related deaths and provided more detail than VS data alone. This is not surprising, because ME cases represent a select group of deaths that are specifically referred to OCME because of their sudden, violent, unexpected, unattended, or suspicious nature. Information contained in the ME report narrative is invaluable for classifying cases and collecting additional details about the circumstances concerning the death. For example, 28 (4%) deaths were identified as heat-related and coded as such; however, when reviewing the narrative description, a determination was made that the death involved exposure to heat of manmade origin, which excluded the case from analysis. By using both databases to search for cases, followed by case confirmation with narrative description from ME data, the combination provided a higher degree of quality control that was crucial for this analysis. Searching for heat-related deaths with heat-related nomenclature captured the majority of cases. Hyperthermia was the most frequently employed heat-related nomenclature term, which is consistent with a 2006 Morbidity and Mortality Weekly Report that demonstrated the inclusion of hyperthermia as a search term increased the total number of heat-related deaths during 1999–2003 in the United States by 54% (CDC, 2006). The most common heatspecific ICD codes were E900 (ICD-9) and X30 (ICD-10) (excessive heat) and E992 (ICD-9) and T67 (ICD-10) (effects of heat and light). However, other inappropriate ICD codes, including 780.6 (ICD-9) and R50.9 (ICD-10) (fever, unspecified), were periodically assigned to heat-related deaths. Heat-specific ICD codes were present

M.G. Johnson et al. / Environmental Research 150 (2016) 30–37

among only 68% of cases, emphasizing the need for more accurate coding of death certificates by nosologists and using literal text from death certificates in searching for heat-related deaths. 4.2. Descriptive epidemiology More heat-related deaths have been documented in Oklahoma in recent years, with peaks occurring during heat waves in 1998, 2006, and 2011. Studies have suggested that only cooler temperate regions are at risk for heat-related illness because persons in those regions are not adapted to hot weather and might not use air conditioning as often as persons in other regions (Anderson and Bell, 2009). However, heat-related mortality is a concern for hot climates and active work is being done to mitigate the potential harm, such as in Maricopa County, AZ (Maricopa County Department of Public Health, 2016). In our study, a pattern of nonuse of air conditioning was observed among decedents. Whether the air conditioner was turned off, not present, or malfunctioning, the vast majority of heat-related deaths that occurred indoors in Oklahoma involved not using air conditioning. Heat-related deaths also occurred among outdoor workers (e.g., laborers, farmers or ranchers, and oilfield workers) who often do not have access to climate-control measures. We hypothesize that persons residing in hotter regions might be maladapted because of their dependence on climate control; persons can have an inability to cope with climate extremes during system failures (e.g., power outages or a malfunctioning air conditioner). Despite the increased availability of air conditioning and its protective effect, a substantial and growing burden of heat-related illness in Oklahoma should be addressed to prevent unnecessary deaths (Bobb et al., 2014). Approximately half of the deaths occurred during July. Only 16% of all heat-related mortality occurred during April–June, and 37% occurred during August–September. However, acclimatization to higher temperatures through physical adaptation, housing characteristics, or behavioral patterns would predict greater heat-related mortality earlier during the summer when the body has not had the opportunity to adjust yet (Patz et al., 2005; Anderson and Bell, 2009). An alternative explanation involves mortality displacement, wherein already frail individuals are “harvested” by an exposure (in this case, heat) early in the season which leaves behind a stronger population that is less susceptible to the exposure later in the season (Basu and Malig, 2011). However, we observed more deaths later in the summer than earlier in the summer. Given the lack of thermal variability data in our study, it is difficult to draw any clear conclusions other than to state that continued awareness of heat risk not just at the start of summer, but throughout the summer season, is important. Males, Blacks, persons aged Z65 years, and unmarried persons are disproportionately represented among heat-related deaths in Oklahoma. These risk factors have been previously described among other populations (Basu and Samet, 2002; Basu, 2009; Bouchama et al., 2007; Semenza et al., 1996; CDC, 2002). Males can have a propensity to do more strenuous outdoor activity than females, placing them at greater risk for heat-related illness. Fig. 3 displays the higher frequency of heat-related mortality among males aged 15–64 years, the age range in which they would likely perform more strenuous outdoor activity. Other studies have not consistently reported gender differences regarding vulnerability to heat-related mortality; thus, no consensus regarding gender exists (Basu and Samet, 2002; Basu, 2009; Semenza et al., 1996). Likewise, literature on race and ethnicity as a risk factor for heat-related mortality is conflicting. Some studies cite Blacks as being more vulnerable to excess heat, while others find no association between race/ethnicity and heat-related illness (Gronlund, 2014). Some factors which have been proposed to underlie this potential association include genetic differences to heat tolerance, lower

35

income and socioeconomic status, less air conditioning ownership, poorer overall health, and more social isolation (Gronlund, 2014). Older persons are thought to have an impaired physiologic response to heat stress with less of a thirst response and might have less ability to avoid hot environments and obtain adequate fluid intake, often in addition to the presence of chronic medical conditions (Basu, 2009; CDC, 2002). Approximately 80% of decedents were single, divorced, widowed, or separated; socially isolated persons often do not have a robust support network to assist them during heat-wave periods, placing them at higher risk for death (Semenza et al., 1996). Chronic medical conditions (e.g., cardiovascular or psychiatric disease) are other well-known risk factors for heat-related mortality (Basu and Samet, 2002; Basu, 2009; Bouchama et al., 2007; Semenza et al., 1996). However little research has been conducted regarding other medical comorbidities. After cardiovascular disease, the most common underlying causes of death with hyperthermia as a contributing factor in the United States during 1999–2003 were external causes (e.g., unintentional poisonings) (29%); respiratory disease (3%); endocrine, nutritional, and metabolic disorders (3%); and mental and behavioral disorders (2%) (CDC, 2006). In our study, we identified a higher proportion of heat-related deaths associated with psychiatric, neurologic, pulmonary, and endocrine disease, as well as substance abuse. Not only do these medical comorbidities directly compromise the overall health of the decedents, but certain medications (e.g, diuretics and antidepressants) can interfere with the body's heat regulatory abilities (Semenza et al., 1996). Using a comprehensive case review of all available records (e.g., ME and VS data, autopsy reports, and toxicology screening) allowed for a more thorough assessment of the decedent's medical history and important contributing factors to heat-related mortality. As for our observation of increasing rates of heat-related mortality in Oklahoma, our data challenge recent evidence from national and international studies that the population has become more resilient to heat over time. Based on our case definition, we have found a positive trend in heat-related deaths over the study period. In a recent analysis of all-cause chronic disease (e.g., nonaccidental) mortality from 1987 to 2005, Bobb et al. (2014) report a negative, non-significant change in deaths for Tulsa, a positive, non-significant change in deaths for Oklahoma City, and an overall significant decrease in heat-related deaths nationally (Bobb et al., 2014). In an international study of all-cause chronic disease mortality, Gasparrini et al. (2015) also found attenuation in the mortality risk from heat in the United States for 1985–2006. Several other national and regional studies support the assertion of a decline in heat-related mortality (Anderson and Bell, 2009; Sheridan et al., 2009; Barnett, 2007; Petrovka et al., 2014). One primary reason for these discrepancies in our study compared to those in the primary literature likely relates to case-ascertainment. In the United States, age-adjusted cardiovascular disease mortality, which comprises the largest category of all-cause chronic disease mortality (CDC, 2015), and primary risk factors for cardiovascular disease (with the exception of diabetes mellitus) have decreased over time (Cooper et al., 2000; Gregg et al., 2005). These decreases have been largely attributed to behavioral changes (e.g., smoking cessation, active lifestyle) and improved medical treatments (e.g., hypertension control) (Gregg et al., 2005). The time-series methodology implemented in these large national studies do not account for the secular trends in these risk factors that contribute to cardiovascular disease mortality. This is especially critical in Oklahoma, which currently ranks 48th in tobacco use prevalence and has not seen the concomitant reductions in tobacco use compared to other states (CDC, 2012). Though Anderson and Bell (2009) found Oklahoma City and Tulsa to be in the lowest 20% of cities in the US impacted by heat-related mortality, the study period of

36

M.G. Johnson et al. / Environmental Research 150 (2016) 30–37

1987–2000 covered a time when tobacco use prevalence was relatively high and fairly consistent, which may have strongly confounded the comparisons to other metropolitan areas during that time (Anderson and Bell, 2009; American Lung Association, 2011). Further, there is acknowledged difficulty in creating national estimates for a large country with heterogeneous meteorological conditions as well as highly variable sociodemographic covariates and infrastructure (Gasparrini et al., 2015). During the 2008 heatwave in Oklahoma City, a consistent urban heat island effect was observed in which temperatures in the urban core were approximately 0.5 °C warmer in the day, and 2 °C warmer at night compared to rural areas (Basara et al., 2010). The evidence for a decrease in heat-related mortality also lie in contrast to future projections, which suggest that future heat-related mortality will increase given climate change scenarios (Hondula et al., 2015). Though several studies based on evaluation of past data suggest that adaptation may contribute to the observed decrease in heat-related mortality, there is little to no empirical evidence that exists to examine the role of adaptation in mitigating the effects of extreme temperatures on health (Hondula et al., 2015; Deschenes, 2014). As cities and states move to adapt climate change plans, there may be a growing research base to test theories of adaptation (Wheeler, 2008). Limitations of this study include miscoding of death certificate data and subjective narrative descriptions and cause-of-death nomenclature used by different medical examiners in Oklahoma, which might have resulted in bias, misclassification, and missed cases. For instance, among the 719 cases initially identified during 1990–2011, 14 cases were coding errors upon further review and deemed non-cases (Supplemental Material, Table S2). In addition, we have likely underestimated the true number of heat-related deaths because of under-recognition by health care professionals and the inaccuracy of death certificates, in addition to inconsistent criteria for the classification of heat-related deaths in the literature (Mirchandani et al., 1996). Detailed meteorological data were not included in the analysis and thus we are unable to draw specific conclusions regarding the time of death and meteorological variables such as temperature, ozone, and particulate matter. Demographic data and other clinical information of the decedents were not consistently reported in ME and VS data. Finally, using marital status might not be a relevant proxy for social isolation, but other collateral history was lacking. Limiting months of eligibility for probable heat-related mortality cases to May–September is not felt to be a limitation as this restriction only applied to those cases without ME data to corroborate the association. As there are likely fewer heat-related deaths during October–April, we felt this strategy would limit those cases with heat-related coding errors. During summer months, especially during heat waves, increased public health response and social services outreach efforts are critical for preventing heat-related deaths. Efforts to reach older adults, socially isolated persons, those lacking air conditioning, outdoor workers, persons with chronic illnesses, and the homeless are of particular importance. Family members, neighbors, and organizations that provide home visiting services to older adults should conduct more frequent checks on these individuals, assessing the working status of air conditioners in the home, encouraging the use of air conditioning, and if not available, directing persons to an air conditioned environment (e.g., senior/ veterans centers, public libraries, shopping malls, homeless shelters, or public health sponsored relief shelters) (Basu and Samet, 2002; Semenza et al., 1996; CDC, 2002). The importance of a working air conditioner needs to be stressed to those without air conditioning; fans do not provide the same benefit (Becker and Stewart, 2011). Working with media to advertise the availability and locations of community cooling stations, as well as transportation available to assist people in traveling to these stations, will

help prevent deaths. Facilitating social contact for persons who are unmarried or live alone has been demonstrated to reduce the risk for heat-related mortality (Semenza et al., 1996). In our study, approximately 75% of all heat-related deaths were initially investigated by family, friends, or neighbors, and approximately half of all heat-related deaths occurred in the home; thus it is important to frequently check in on vulnerable friends and family members at home to prevent heat-related deaths. Physicians and pharmacists should discuss with patients the risks of heat-related illness in relation to certain chronic medical conditions and medications. Home health care providers can be important sources of preventive health messaging for persons with chronic illnesses or other medical conditions, including psychiatric disease; visiting medical and non-medical organizations, including those providing mobile meals and other similar networks, can take advantage of preexisting outreach connections to frail or chronically ill populations (Semenza et al., 1996). Information on first aid for heat-related illnesses should also be publicized. Electric companies should incorporate policies to suspend disconnection of service due to non-payment when the temperature or heat index is very high (Oklahoma Gas and Electric, 2014). Businesses that employ outdoor workers should ensure access to an adequate fluid supply, shade for rest and recovery, and limit strenuous activity during the hottest daytime hours (Becker and Stewart, 2011; Jackson and Rosenberg, 2010). This is particularly important for states such as Oklahoma, with many outdoor laborers in the oil and natural gas industry as well as farmers and ranchers. State, local, and municipal emergency preparedness plans need to be developed in partnership with public health, social services, faith-based organizations, community coalitions, businesses, health care entities, and others to respond to extreme heat emergencies, particularly for the aforementioned vulnerable populations, to reduce heat-related morbidity and mortality (Bernard and McGeehin, 2004).

5. Conclusions A tiered and comprehensive search strategy was used to capture a more accurate estimate of the total burden of heat-related mortality in Oklahoma during the past two decades. Miscoding and data entry errors remain a problem, but linking ME and VS data and using ME narrative description for case classification improves case ascertainment and surveillance data quality. Searching for heat-related nomenclature (e.g., hyperthermia) in addition to heat-specific ICD codes proved to be a superior strategy for identifying heat-related deaths; searching for heat-specific ICD codes alone was a less effective strategy. Heat-related mortality is not just a concern in northern, temperate regions of the United States. Other parts of the country, even among areas with a preponderance of air conditioning, can be at risk during heat waves. In addition, physiological adaptation does not always reduce the risk for death at the end of summer; maintaining preventive efforts during the entire summer is important. Cardiopulmonary and psychiatric diseases are not the only chronic medical comorbidities that can place a person at risk for heat-related death; neurologic disease, endocrine disease, and substance abuse can also be important contributing factors. A coordinated, comprehensive municipal heat response plan in place in every city is helpful to communicate safety messaging to the public and populations at high risk to reduce heat-related mortality during heat waves. Financial interests declaration None.

M.G. Johnson et al. / Environmental Research 150 (2016) 30–37

Funding sources None.

Note The findings and conclusions in this report are those of the author(s) and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Acknowledgements We thank Tabitha Garwe, Heather Basara, Jeffrey Basara, Rachel Jantz, and the staff of the Office of the Chief Medical Examiner and the Oklahoma State Department of Health Center for Health Statistics for their assistance.

Appendix A. Supplementary material Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.envres.2016.05.035.

References American Lung Association, 2011. Trends in tobacco use. Research and program services, epidemiology and statistics unit. Available: 〈http://www.lung.org/as sets/documents/research/tobacco-trend-report.pdf〉 (accessed 27.04.16). Anderson, B.G., Bell, M.L., 2009. Weather-related mortality: how heat, cold, and heat waves affect mortality in the United States. Epidemiology 20, 205–213. Barnett, A.G., 2007. Temperature and cardiovascular deaths in the US elderly: changes over time. Epidemiology 18, 369–372. Basara, J.B., Basara, H.G., Illston, B.G., et al., 2010. The impact of the urban heat island during an intense heat wave in Oklahoma City. Adv. Meteorol. 2010, 1–11. Basu, R., 2009. High ambient temperature and mortality: a review of epidemiologic studies from 2001 to 2008. Environ. Health 8, 40–53. Basu, R., Samet, J., 2002. Relation between elevated ambient temperature and mortality: a review of the epidemiologic evidence. Epidemiol. Rev. 24, 190–202. Basu, R., Malig, B., 2011. High ambient temperature and mortality in California: exploring the roles of age, disease, and mortality displacement. Environ. Res. 111, 1286–1292. Becker, J.A., Stewart, L.K., 2011. Heat-related illness. Am. Fam. Physician 83, 1325–1330. Bernard, S.M., McGeehin, M.A., 2004. Municipal heat wave response plans. Am. J. Public Health 94, 1520–1522. Bobb, J.F., Peng, R.D., Bell, M.L., et al., 2014. Heat-related mortality and adaptation to heat in the United States. Environ. Health Perspect. 122, 811–816. Bouchama, A., Dehbi, M., Mohamed, G., et al., 2007. Prognostic factors in heat waverelated deaths. Arch. Intern. Med. 167, 2170–2176. CDC, 2002. Heat-related deaths—four states, July–August 2001, and United States, 1979–1999. Morb. Mortal. Wkly. Rep. 51, 567–570. CDC, 2005. Heat-related mortality—Arizona, 1993–2002, and United States, 1979– 1999. Morb. Mortal. Wkly. Rep. 54, 628–630. CDC, 2006. Heat-related deaths—United States, 1999–2003. Morb. Mortal. Wkly, Rep. 55, 796–798. CDC, 2013. Heat-related deaths after an extreme heat event—Four state, 2012, and United States, 1999–2009. Morb. Mortal. Wkly. Rep . 62, 433–436. CDC (National Center for Health Statistics), 2015. Underlying cause of death 1999– 2013 on CDC WONDER online database, released 2015. Data are from the Multiple Cause of Death files, 1999–2013, as compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative Program: 〈http://wonder.cdc.gov/ucd-icd10.html〉 (accessed 03.02.15). CDC, 2012. Smoking and tobacco use: state highlights. Available: 〈http://www.cdc. gov/tobacco/data_statistics/state_data/〉 (accessed 25.04.16). Comstock, R.D., Mallonee, S., Jordan, F., 2005. A comparison of two surveillance

37

systems for deaths related to violent injury. Inj. Prev. 11, 58–63. Cooper, R., Cutler, J., Desvigne-Nickens, P., et al., 2000. Trends and disparities coronary heart disease, stroke, and other cardiovascular risk diseases in the Unites states: findings of the National Conference on Cardiovascular Disease Prevention. Circulation 102, 3137–3147. Davis, R.E., Knappenberger, P.C., Michael, P.J., et al., 2003. Changing heat-related mortality in the United States. Environ. Health Perspect. 111, 1712–1718. Deschenes, O., 2014. Temperature, human health, and adaptation: a review of the empirical literature. Energy Econ. 46, 606–619. Donoghue, E.R., Graham, M.A., Jentzen, J.M., et al., 1997. Criteria for the diagnosis of heat-related deaths: National Association of Medical Examiners. Position paper. National Association of Medical Examiners Ad Hoc Committee on the definition of heat-related fatalities. Am. J. Forensic Med. Pathol. 18, 11–14. Gasparrini, A., Guo, Y., Hashizume, M., et al., 2015. Temporal variation in heatmortality associations: a multicountry study. Environ. Health Perspect. 123, 1200–1207. Graitcer, P.L., Williams, W.W., Finton, R.J., et al., 1987. An evaluation of the use of medical examiner data for epidemiological surveillance. Am. J. Public Health 77, 1212–1214. Gregg, E.W., Cheng, Y.J., Cadwell, B.L., et al., 2005. Secular trends in cardiovascular disease risk factors according to body mass index in US adults. JAMA 293, 1868–1874. Gronlund, C.J., 2014. Racial and socioeconomic disparities in heat-related health effects and their mechanism: a review. Curr. Epidemiol. Rep. 1, 165–173. Hanzlick, R., Parrish, R.G., 1996. The role of medical examiners and coroners in public health surveillance and epidemiologic research. Annu Rev. Public Health 17, 383–409. Harlan, S.L., Chowell, G., Yang, S., et al., 2014. Heat-related deaths in hot cities: estimates of human tolerance to high temperature thresholds. Int. J. Environ. Res. Public Health 11, 3304–3326. Honda, Y., Kondo, M., McGregor, G., et al., 2014. Heat-related mortality risk model for climate change impact projection. Environ. Health Prev. Med 19, 56–63. Hondula, D.M., Balling Jr., R.C., Vanos, J.K., et al., 2015. Rising temperatures, human health, and the role of adaptation. Curr. Clim. Change Rep. 1, 144–154. Jackson, L.L., Rosenberg, H.R., 2010. Preventing heat-related illness among agricultural workers. J. Agromed. 15, 200–215. Maricopa County Department of Public Health, 2016. Surviving Arizona's Extreme Heat. Available: 〈http://www.maricopa.gov/PublicHealth/Programs/Heat/〉 (accessed 01.02.16). Mirchandani, H.G., McDonald, G., Hood, I.C., et al., 1996. Heat-related deaths in Philadelphia—1993. Am. J. Forensic Med. Pathol. 17, 106–108. Moyer, L.A., Boyle, C.A., Pollack, D.A., 1989. Validity of death certificates for injuryrelated causes of deaths. Am. J. Epidemiol. 130, 1024–1032. Naughton, M.P., Henderson, A., Mirabelli, M.C., et al., 2002. Heat-related mortality during a 1999 heat wave in Chicago. Am. J. Prev. Med 22, 221–227. Oklahoma Climatological Survey, 2016. Climate of Oklahoma. Available: 〈http://cli mate.ok.gov/index.php/site/page/climate_of_oklahoma〉 (accessed 01.02.16). Oklahoma Gas and Electric, 2014. What Residential Customers Should Know About OGþ E Electric Services' Disconnection Policy. Available: 〈https://oge.com/wps/ wcm/connect/5ac2786d-8b25-44c9-86c4-d607ba96ca16/Disconnect þPolicy. pdf?MOD¼ AJPERES&CACHEID ¼5ac2786d-8b25-44c9-86c4-d607ba96ca16〉 (accessed 01.02.16). Oklahoma State Department of Health, 2016. Death Registration Training. Available: 〈https://www.ok.gov/health/Birth_and_Death_Certificates/ROVER/Death_Regis tration_Training/Index.HTML〉. (accessed 19.04.16). Ostro, B.D., Roth, L.A., Green, R.S., et al., 2009. Estimating the mortality effect of the July 2006 California heat wave. Environ. Res. 109, 614–619. Patz, J.A., Campbell-Lendrum, D., Holloway, T., et al., 2005. Impact of regional climate change on human health. Nature 438, 310–317. Petrovka, E., Gasparrini, A., Kinney, P.L., 2014. Heat and mortality in New York City since the beginning of the 20th century. Epidemiology 25, 554–560. Semenza, J.C., Rubin, C.H., Falter, K.H., et al., 1996. Heat-related deaths during the July 1995 heat wave in Chicago. N. Engl. J. Med 335, 84–90. Shen, T., Howe, H., Ato, C., et al., 1998. Towards 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., Kalkstein, A.J., Kalkstein, L.S., 2009. Trends in heat-related mortality in the United States, 1975–2004. Nat. Hazards 50, 145–160. US Census Bureau, 2013. US Census Bureau Homepage, US Department of Commerce, US Census Bureau Washington, DC. Available: 〈http://www.census.gov/ en.html〉 (accessed 04.12.14). Wheeler, S.M., 2008. State and municipal climate change plans: the first generation. J. Am. Plan. Assoc. 74, 481–496. WHO, 2007. International Classification of Diseases. 10th Revision, WHO, Geneva, Switzerland. Available: 〈http://www.who.int/classifications/icd/en/〉 (accessed 04.12.14).