a b s t r a c t

Background: Unintentional, non-fire related (UNFR) Carbon monoxide poisoning deaths are preventable. Surveillance of the populations most at-risk for unintentional, non-fire related (UNFR) Carbon monoxide poisoning is crucial for targeting Prevention efforts.

Objective: This study provides estimates on UNFR CO poisoning mortality in the United States and characterizes the at-risk populations.

Methods: We used 1999 to 2012 data to calculate Death rates. We used underlying and multiple conditions variables from death records to identify UNFR CO poisoning cases.

Results: For this study, we identified 6136 CO poisoning fatalities during 1999 to 2012 resulting in an average of 438 deaths annually. The annual average age-adjusted death rate was 1.48 deaths per million. Fifty four percent of the deaths occurred in a home. Age-adjusted death rates were highest for males (2.21 deaths per million) and non-Hispanic blacks (1.74 deaths per million). The age-specific death rate was highest for those aged >=85 years (6.00 deaths per million). The annual rate of UNFR CO poisoning deaths did not change substantially during the study period, but we observed a decrease in the rate of suicide and unintentional fire related cases.

Conclusion: CO poisoning was the second most common non-medicinal poisonings death. Developing and enhancing current public health interventions could reduce ongoing exposures to CO from common sources, such as those in the residential setting.

Introduction

Carbon monoxide (CO) is a toxic, colorless, odorless gas that is a product of combustion [1] Common symptoms of CO exposure include headache, dizziness, fatigue, nausea, vomiting, and chest pain. CO is particularly dangerous because it is imperceptible with a non-specific symptomatic presentation; victims can become sick or die before realizing they are exposed. However, CO poisoning related deaths are preventable.

CO poisoning causes approximately 50,000 people to visit the emergency department per year [2]. Of these, unintentional, non-fire

? Presentation of this material: Council of State and Territorial Epidemiology, June 2014.

?? Disclaimer: The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.

* Corresponding author. air pollution and Respiratory Health Branch, National Center for Environmental Health, Centers for Disease Control and Prevention, 4770 Buford High- way, NE, MS F-58, Chamblee, GA 30341. Tel.: +1 770 488 3384; fax: +1 770 488 1540.

E-mail address: [email protected] (K. Sircar).

related (UNFR) CO poisonings cause an estimated 21,000 people to visit an emergency department each year and result in hospitalization for more than 2300 [3]. Many CO poisoning cases occur in the home [4]. The most common sources of CO poisonings in homes are the use of gasoline-powered engines, such as electric generators, and malfunctioning heating and cooking appliances [1].

Several factors have affected CO poisoning death rates. Engineering innovations that decrease the amount of CO released from a source have been successful in reducing deaths from CO-poisoning. From 1968-1999, CO poisoning deaths declined due to advances such as the installation of catalytic converters in motor vehicles and the enforce- ment of standards set by the 1970 Clean Air Act [6,7]. Conversely, pre- vious studies have shown that circumstances that lead to greater use of CO-emitting sources, such as winter storms, increase the number of CO poisoning deaths. Loss of electric power and increased cold weather during winter storms or disasters promotes an increased use of gasoline-powered generators in unventilated areas and indoor use of alternative heating sources, such as charcoal grills, consequently increasing the number of deaths [4,8-12].

http://dx.doi.org/10.1016/j.ajem.2015.05.002 0735-6757/

Studies have postulated that interventions such as the installation of CO alarms can also reduce the number of deaths, possibly by half [13]. However, although many states have laws requiring CO alarms in new construction homes, older houses may not have a unit installed. Past studies have noted only 30% of households use CO alarms [14]. Even when homes have CO alarms, residents must remember to change bat- teries in the units regularly for them to work properly.

Although both unintentional and suicide related CO poisonings are preventable, public health intervention strategies differ. Interventions to prevent suicide-related poisonings focus on the sources of suicidal thoughts, while interventions to prevent unintentional poisonings focus on the source of the CO. For example, in unintentional fire- related CO poisoning, the fire is the source of the CO and therefore, fire prevention education and the use of flame retardant material are the focus. Unlike fire-related CO poisonings, UNFR CO poisoning victims do not sense the presence of a poisonous gas unless a CO alarm alerts them. Outreach and education for UNFR CO poisoning can range from increasing the use of CO alarms to educating consumers on the proper use of and care for related equipment and early sign and symptoms of CO poisoning. This analysis focuses on UNFR CO poisoning.

The objective of this study is to more comprehensively characterize the burden of CO poisoning-related mortality in the U.S and to charac- terize the populations most at-risk. Understanding the factors associated to UNFR CO poisoning death is a necessary component in the burden of both fatal and non-fatal CO poisoning cases and therefore a critical part for successful public health planning and prevention efforts.

Methods

To identify fatalities from CO poisoning, we used mortality data for multiple causes of death in the United States from the National Center for Health Statistic’s National Vital Statistics System for the years 1999 through 2012. The dataset for this period is a compilation of all death certificate data without any personal identifiers. Fatalities were limited to those who resided and died in the 50 states and the District of Columbia.

We calculated the frequency of death due to non-medicinal poisonings listed on the death record and defined by International Classification of Diseases, Tenth Revision (ICD-10) codes T51-T65, “Toxic effects of sub- stances chiefly non-medicinal as to source.”

Table 1 presents the definitions for the CO poisoning deaths. T58, toxic effect of CO, can only be a contributory cause of death and not an underlying cause. Therefore, we used code X47 (accidental poisoning by exposure to gases and vapors) to denote poisoning as either an un- derlying or contributory cause [15]. The manner of death, along with the ICD-10 code, distinguishes unintentional deaths (“accidental”) from suicides (“suicides” or “self-harm”). We calculated annual crude rates for suicides, UNFR, and unintentional fire and vehicular-related CO poisoning deaths.

We adjusted death rates (DRs) for age using the direct standardiza- tion method and the US standard age population for year 2000 [16]. De- nominator data for calculated rates came from the US Census population with bridged-race census counts or intercensal or postcensal estimates of the US resident population for July 1, 1999-July 1, 2012. For years 2000 and 2010, we used the census population count, and for all other years (1999, 2001-2009, 2011-2012) we used the intercensal or

postcensal estimates [17]. We calculated all annual rates in units of deaths per one million persons (deaths per million); we did not present rates based on 10 or fewer deaths.

We stratified age-adjusted UNFR CO poisoning DRs by sex, race/ ethnicity, educational attainment, marital status, urban/rural county of death classification, state, region, and year of death. For educational at- tainment analysis, we limited the population to persons 18 years and older and for marital status analysis, we limited the population to persons 15 years and older. Those categorized as “Widowed,” “Never married, single,” and “Divorced” were considered not married. In addition, we calculated age stratified rates and analyzed the average number of deaths per day by season, month, and day of the week. Finally, we computed frequencies for place of death and place of CO exposure.

We calculated 95 percent confidence intervals (95% CI) using the gamma method [18] and conducted analyses using SAS software (ver- sion 9.3; SAS Institute, Inc., Cary, NC). To determine if there was a statis- tical difference in the average number of deaths per day of the week, we used Poisson regressions using an alpha of 0.05.

Results

Among US residents in the 50 states and the District of Columbia, 34,215 CO poisoning deaths (ICD-10 code: T58) occurred from 1999 to 2012. CO poisoning is the second most frequent poisoning listed on death certificates (Table 2).

The annual average age-adjusted UNFR CO poisoning DR for the study

period (1999-2012) was 1.46 deaths per million (Table 3). Age-adjusted DRs were the highest for males (2.28 deaths per million) and non- Hispanic blacks (1.74 deaths per million). By age group, DRs were highest for those aged 85 years or greater (6.00 deaths per million) and lowest for those aged 5 to 14 years (0.25 deaths per million). Among decedents 18 years and older, 66% had a high school education or less. Finally, in 65% of the cases, UNFR CO decedents were not married (Table 3).

Forty-nine states had a sufficient number of deaths to permit calcu- lation of age-adjusted death rates by state (Fig. 1). The Western region had the highest DR (2.05 deaths per million [95% CI: 1.96, 2.15]) and the Northeast had the lowest DR (0.91 deaths per million [95% CI: 0.85, 0.98]). The three states with the highest age-adjusted UNFR CO poisoning DR were Wyoming (5.14 deaths per million [95% CI: 3.62, 7.08]), Montana (3.66 deaths per million [95% CI: 2.70, 4.84]), and Alas- ka (3.77 deaths per million [95% CI: 2.66, 5.18]) (Fig. 1). Among States with more than 10 deaths, the lowest age-adjusted DRs per million per- sons were Massachusetts (0.49 deaths per million [95% CI: 0.36, 0.65]), New Jersey (0.82 deaths per million [95% CI: 0.67, 0.99]) and California (0.68 deaths per million [95% CI: 0.60, 0.75]). Further geographic analy- sis revealed that age-adjusted rates of UNFR CO deaths were higher in rural counties (2.28 deaths per million [95% CI: 2.18, 2.38]) than in urban counties (1.24 deaths per million [95%CI: 1.20, 1.27]) (Table 4).

Over half of the fatalities (54%, n = 3341) occurred in the decedent’s home (Table 4), rather than in a medical center (18%, n = 1104. (Table 4) In 3744 (61%) of the deaths, the place of CO exposure was the home. In 549 (8%) of CO fatalities, exposure occurred in trade and service areas; 118 (1%) occurred in streets and highways, and 58 (1%) occurred in industrial and construction areas. Three percent (n = 193) of deaths were reported to be related to occupation.

Table 1 Definitions for UNFR, unintentional fire and vehicular and suicide carbon monoxide related death. Each death definition combines a Manner of Death code and an ICD-10 code

Table 2

Frequency of poisonings (Toxic effects of substances chiefly non-medicinal as to source) on the death certificates, 1999 to 2010

ICD-10 codes (either underlying or contributory cause)

Poisons Number of

deaths

CO

poisoning

UNFR T58 & X47 Exclude: X00-X09, X76, X97, Y26, and Y17

Unintentional vehicular T58& at least one from V01 to V99

Toxic effects of alcohol (ICD-10 code: T51) 77,082 704 (0.9%) Toxic effects of CO poisoning (ICD-10 code: T58) 34,215

Unintentional fire T58 & at least one from X00-X09 Suicide T58 & least one from X60 to X84

Smoke inhalation excluding tobacco smoke (ICD-10 code: T598)

34,037 5897 (17%)

Table 3

Characteristics of persons with fatal, unintentional, non-fire related carbon monoxide poisoning, United States, 1999 to 2012. N = 6136

In this study, we did not observe similar declines in UNFR CO poisoning fatalities as reported in previous studies [6,7]. In 1968, the crude death rate for all unintentional CO poisonings was 7.06 deaths per million and the

Variable n (%) Average annual crude rate?

Average annual age-adjusted rate?

95%

Confidence Interval

majority of deaths were due to motor vehicle exhaust [6]. In 1998, the rate dropped to 1.81 deaths per million [6]; from 1968 to 1998, deaths due to motor vehicle exhaust experienced the greatest decline. In our

Total (N) 6136 (100) 1.48 1.46 1.42-1.49

Age group (years)

b5	97 (1)	0.35	- -
5-14	141 (2)	0.25	- -
15-24	567 (9)	0.96	- -
25-34	715 (11)	1.27	- -
35-44	1038 (16)	1.72	- -
45-54	1157 (18)	1.96	- -
55-64	836 (13)	1.92	- -
65-74	557 (9)	2.00	- -
75-84	618 (10)	3.42	- -
>= 85	410 (6)	6.00	- -

Sex

Male	4518 (73)	2.21	2.28	2.21-2.34
Female	1618 (26)	0.77	0.72	0.69-0.76
ace/ethnicity White, non- 4427 (72) 1.59 1.46 1.42-1.50
Hispanic Black, non-	828 (13)	1.58	1.74	1.62-1.85
Hispanic Other, non-	203 (3)	0.88	0.90	0.78-1.03
Hispanic Hispanic	656 (10)	1.07	1.17	1.08-1.26
Unknown	22 (b1)	-	-	-

R

Educational attainment?? (n

= 5792)

study, the average age-adjusted UNFR CO death rate plateaued at 1.52 deaths per million over the 12 years.

Suicides and unintentional fire-related CO poisoning deaths continue to decline, and vehicle-related CO deaths have the lowest rate of all categories. The decline in suicides by CO poisoning is consistent with other studies [30]. Successes in fire safety may have caused the decline in unintentional fire-related deaths [29]. The rapid de- clines that were once seen with engineering changes in motor vehicles, including the use of the catalytic convertor and resulting increased fuel efficiency and compliance with standards in the Clean Air Act, reached a plateau [7].

Since 1999, the average annual rates of UNFR CO mortality did not change. Poisonings as a group remains a leading cause of residential deaths in the US [33]. As demonstrated in this study, most (64%) UNFR CO poisonings deaths occurred in the home. Residential settings contain many common CO sources, including poorly maintained or un- ventilated heating and cooking appliances, or motor vehicles running in an enclosed space. CO alarms in homes are, therefore, an important sec- ondary prevention strategy. However, according to national estimates during the study period, only 29% to 33% of American households reported having a working CO alarm [19,20].

Our characterization of the at-risk population for UNFR CO poisoning

Less than high school graduate High school graduate

1273 (22)

2534 (44)

fatalities is largely consistent with previous studies [8]. Studies have found that both male and elderly population subgroups have the highest rates of UNFR CO related hospitalization [5]. Males also typically

Some college 1025 (18)

College or more 682 (12)

Unknown 278 (5)

Marital status???? (n = 5898)

Married 2023 (34)

Not Married 3807 (65)

Unknown 68 (1)

* Rates per 1 million persons.

?? Educational attainment only for those persons >=18 years.

???? Marital status only for those >=15 years.

Fig. 2 shows the number of CO poisoning cases by year and manner of death. CO poisoning from suicides and unintentional fire cases de- creased during 1999 to 2012, while UNFR CO poisoning cases remained consistent over this period. The annual age-adjusted DR ranged from

1.21 deaths per million in 2010 to 1.68 deaths per million in 2005. The average daily number of deaths was 1.68 deaths for the entire period. The average daily number of deaths by month ranged from 0.99 deaths per day in July to 2.74 deaths per day in January. Analysis by day of the week showed that the daily average number of deaths was lower on weekdays (1.59 deaths per day) than on weekends (1.90 deaths per day) (P b .0001) (Table 4). Specifically, the average daily number of CO deaths was highest on Sundays (2.00 deaths) and lowest on Thurs- days (1.53 deaths) (P b 0.0001).

Discussion

From 1999 to 2012, CO poisoning was the second most common cause of death among non-medicinal poisonings (Table 2). The annual rate of UNFR CO poisoning deaths did not change substantially during this time, but we observed a decrease in the rate of suicide and uninten- tional fire related cases. The demographic groups with the highest age- adjusted DRs were among those who were 85 years or older, males, and non-Hispanic blacks.

represent a larger proportion of those who are severely poisoned. Males

may experience a higher dose of CO poisoning exposure due to the fre- quency of use of CO emitting equipment. For example, males may have higher DRs due to the increased likelihood of using fuel-burning equip- ment [8]. In contrast, females are more likely to receive medical treat- ment for less Severe poisonings, such as those that result in a visit to an emergency department [5].

Elderly population subgroups may be associated with other factors such as comorbid conditions and misdiagnoses which leads to repeated exposure. These factors may explain their higher death rate. For exam- ple, CO poisoning may exacerbate preexisting cardiopulmonary disease, even at low levels [31]. CO poisoning may be misdiagnosed, as symp- toms are nonspecific, which could result in the return of those exposed to a hazardous location [33]. Seeking medical care and rapid diagnosis and treatment increase the chance of a successful outcome [21,22] and may be why death is less frequent once CO poisoning victims are in the hospital [6].

Geographically, the US West and Midwest had the highest age- adjusted DRs, a finding that is consistent with other studies [5,6,23]. These regional distributions may be a reflection of the increase of UNFR CO deaths during winter. Although the Northeastern US has se- vere winters, it has the lowest geographical age-adjusted DR. Studies of CO morbidity have found higher regional rates of non-fatal UNFR CO poisoning in the Northeast [5,23] suggesting that residents leave the CO poisoned area and seek help. Another finding consistent with previous literature is that the incidence of CO poisoning for rural areas was higher than urban areas’ UNFR CO poisoning DR [5,24].

Temporally, our study found that the highest average daily number of CO related deaths occurred during the winter months, similar to ob- servations in previous studies [5,24,25]. Increases in such high-risk be- haviors as “warming up” motor vehicles, improperly maintaining home heating systems, using cooking equipment for heating, and using generators during winter storms leads to a greater rate of CO poi- soning during winter months [25]. In addition, CO related deaths due to disaster situations, such as hurricanes, may be attributable to factors

*Rates based on <=10 deaths in the numerator not presented(Hawaii and District of Columbia). Per million

Fig. 1. Age-adjusted UNFR CO Poisoning by state, 1999-2012, United States*. *Rates based on <=10 deaths in the numerator not presented (Hawaii and District of Columbia). Per million.

such as extended power outages that can lead to increased use of gene- rators and alternative heating and cooking methods [4,26,27].

Higher UNFR CO poisoning DRs on the weekend may be caused by prolonged exposure to CO, as compared to weekday exposure. During the weekends, people may be likely to engage in more leisure time

behaviors that could increase exposure to CO, such as using motorized equipment for home improvement or engine repair. More weekend hours may be spent at home than during the weekdays when many are at work. In addition, individuals may be more likely to consume alcohol on the weekend. In fact, previous studies have found that up

Table 4

Exposure circumstances for cases of fatal, unintentional, non-fire related CO poisoning, United States, 1999 to 2012. N = 6136

Variable n (%)	Age adjusted ten year average annual rate?	95% Confidence Interval	Daily average number of deaths
County Urban 4355 (71)	1.24	1.20-1.27	-
Rural 1781 (29)	2.28	2.18-2.38	-
Region
Northeast 733	0.91	0.85-0.98	-
Midwest 1581	1.68	1.59-1.76	-
South 1486	0.97	0.92-1.02	-
West 1932	2.05	1.95-2.25	-
Place of death
Decedent’s home 3341 (54)	-	-	-
Medical center 1104 (18)	-	-	-
Other/unknown 1691 (28)	-	-	-
Place of CO exposure
Home 3744 (61)	-	-	-
Other specified places 697 (11)	-	-	-
Trade and service area 549 (8)	-	-	-
Street and highway 118(1)	-	-	-
Industrial and construction area 58 (1)	-	-	-
Farm 39 (0)	-	-	-
Other/unknown 931 (0)	-	-	-
Yes 193 (3)	-	-	-
No 4885 (83)	-	-	-
Unknown 820 (14)	-	-	-
Season???	-
Winter 2383 (39)	-	-	2.64
Spring 1324 (22)		-	1.44
Summer 952 (16)	-	-	1.03
Autumn 1477 (24)	-	-	1.62
Day of the week****	-
Weekday 4149 (68)	-	-	1.59
Weekend 1984 (32)	-	-	1.90

Death at work??, n = 5898

*****P b .0001

* Average adjusted rate per 1 million persons.

?? Occupational relation only for those >=15 years.

??? Winter is defined as December, January, February; Autumn is defined as September, October, November; Spring is defined as March, April, May; Summer is defined as June, July, August.

Fig. 2. Crude rate of CO poisonings by intent, fire-, and vehicular-relatedness. Trend from 1999 to 2012, United States.

to 42% of those with unintentional, fatal CO poisoning were under the influence of alcohol [13], which impaired their ability to recognize symptoms and seek treatment.

While some of our demographic and temporal trends in mortality have been consistent with previous studies, we also examined previous- ly unpublished findings on populations potentially at-risk for UNFR CO mortality. We found that 69% of the fatalities had a high school educa- tion or less, suggesting that public health outreach needs to focus on this educational level. Furthermore, we found that 65% of those with fatal UNFR CO exposures were not married. One possibility is that no one was present to assist in recognizing the hazard and intervening or seeking help among the non-married individuals with fatal CO exposures.

Overall, maintaining a high index of suspicion is important to help recognize and identify CO cases that go to the emergency department, especially among those most at-risk, and during the winter season and around disaster events. The emergency department not only can play an important role in the early recognition of CO poisoning cases, but can also serve as an opportunity for intervention. For example, if an patient arrives at the emergency department with altered mental status, checking the patient’s Carboxyhemoglobin level in addition to other factors may provide additional insight on their condition. There may also be an opportunity to educate a patient on the potential source of the CO and steps that need to be taken to ensure that it is safe to return to the area, as well as the importance of CO alarms. [28,29,32]

This study is subject to a few limitations. While the main purpose of this paper is to more comprehensively characterize the burden of CO poisoning-related mortality in the US, some cases are missed. For exam- ple, among cases that die in the emergency department or hospital set- ting, if the CO poisoning is not recognized before death, the case can be misdiagnosed and missed, leading to an underestimation of the DR. In addition, because we based the case definition largely on ICD-10 coding, some intentional and fire related cases may be included in the UNFR analysis through reporting and coding errors or vice versa. However, this misclassification may be minimal.

One strength of this study is that poisoning deaths are likely to be reviewed by a Medical Examiner and an autopsy is conducted, thus in- creasing confidence that the event was related to CO [7]. Because of cur- rent vital statistics practices, the number of missing cases should be

minimal. Additionally, the case definition in this study requires both ICD-10 codes T58 and X47, thereby increasing the likelihood that the cases captured in the analysis were truly UNFR CO cases.

Conclusions

The findings of this analysis do not demonstrate a decreasing trend in UNFR CO poisoning deaths since 1999. Opportunities to reduce these deaths can include Public health programs focused on populations with the highest rates of UNFR CO mortality and emphasizing the im- portance of properly installing CO alarms; these approaches have shown to be an effective use of prevention resources and reduce the se- verity of CO exposure and prevent UNFR CO deaths [13,24]. Further- more, as the majority of American households do not have a working CO alarm [19], this is an area where public health interventions can have significant impact.

Acknowledgements

Shahed Iqbal. No external funding was used for this analysis.

References

1. United States Environmental Protection Agency. An Introduction to Indoor Air Quality (IAQ). Carbon Monoxide (CO). http://www.epa.gov/iaq/co.html.
2. Hampson NB, Weaver LK. Carbon monoxide poisoning: a new incidence for an old disease. Undersea Hyperb Med 2007;34:163-8.
3. Raub JA, Mathieu-Nolf M, Hampson NB, Thom SR. Carbon monoxide poisoning-a public health perspective. Toxicology 2000;145:1-14.
4. Iqbal S, Clower JH, Hernandez SA, Damon SA, Yip FY. A review of disaster-related carbon monoxide poisoning: surveillance, epidemiology, and opportunities for prevention. Am J Public Health 2012.
5. Iqbal S, Law HZ, Clower JH, Yip FY, Elixhauser A. hospital burden of unintentional carbon monoxide poisoning in the United States, 2007. Am J Emerg Med 2012;30: 657-64.
6. Mott JA, Wolfe MI, Alverson CJ, Macdonald SC, Bailey CR, Ball LB. National vehicle emissions policies and practices and declining US carbon monoxide-related mortality. JAMA 2002;288:988-95.
7. Cobb N, Etzel RA. Unintentional carbon monoxide-related deaths in the United States, 1979 through 1988. JAMA 1991;266:659-63.
8. Centers for Disease Control and Prevention. Carbon monoxide-related deaths- United States, 1999-2004. MMWR Morb Mortal Wkly Rep 2007;56:1309-12.
9. Hampson NB, Stock AL. Storm-related carbon monoxide poisoning: lessons learned from recent epidemics. Undersea Hyperb Med 2006;33:257-63.
10. Centers for Disease Control and Prevention. Carbon monoxide exposures after Hurricane Ike - Texas, September 2008. MMWR Morb Mortal Wkly Rep 2009;58:845-9.
11. Broder J, Mehrotra A, Tintinalli J. Injuries from the 2002 North Carolina ice storm, and strategies for prevention. Injury 2005;36:21-6.
12. Houck PM, Hampson NB. Epidemic carbon monoxide poisoning following a winter storm. J Emerg Med 1997;15:469-73.
13. Yoon SS, Macdonald SC, Parrish RG. Deaths from unintentional carbon monoxide poisoning and potential for prevention with carbon monoxide detectors. JAMA 1998;279:685-7.
14. Iqbal S, Clower JH, Saha S, Boehmer TK, Mattson C, Yip FY. Residential carbon monoxide alarm prevalence and ordinance awareness. J Public Health Manag Pract 2012;18:272-8.
15. National Center for Health Statistics. Mortality all-county microdata. Vital statistics; 1999-2008 [editor].
16. Klein RJ, Schoenborn CA. Age adjustment using the 2000 projected U.S. population. Healthy People 2012 Stat Notes 2001:1-10.
17. National Center for Health Statistics. Estimates of the resident population of the United States, by county, single-year of age (0, 1, 2, …, 85 years and over), bridged race, Hispanic origin, and sex. Prepared under a collaborative arrangement with the

    U.S. Census Bureau. nchs/nvss/bridged_race.htm. [Last accessed February 15, 2013].

    Fay MP, Feuer EJ. Confidence intervals for directly standardized rates: a method based on the gamma distribution. Stat Med 1997;16:791-801.

18. US Census Bureau, American Housing Survey, Table 988. Housing Units-Characteristics by Tenure and Region: 2009. http://www.census.gov/compendia/statab/2012/tables/ 12s0988.pdf. [Last accessed May 21, 2013].
19. Runyan CW, Johnson RM, Yang J, Waller AE, Perkis D, Marshall SW. Risk and protective factors for fires, burns, and carbon monoxide poisoning in U.S. households. Am J Prev Med 2005;28:102-8.
20. Kao LW, Nanagas KA. Carbon monoxide poisoning. Emerg Med Clin North Am 2004; 22:985-1018.
21. Tomaszewski C. Carbon monoxide poisoning. Early awareness and intervention can save lives. Postgrad Med 1999;105:39-40 [3-8, 50].
22. Centers for Disease Control and Prevention. Carbon monoxide exposures-United States, 2000-2009. MMWR Morb Mortal Wkly Rep 2011;60:1014-7.
23. Clower JH, Hampson NB, Iqbal S, Yip FY. Recipients of Hyperbaric oxygen treatment for car- bon monoxide poisoning and exposure circumstances. Am J Emerg Med 2012;30:846-51.
24. Centers for Disease Control and Prevention. Nonfatal, unintentional, non-fire- related carbon monoxide exposures-United States, 2004-2006. MMWR Morb Mortal Wkly Rep 2008;57:896-9.
25. National Oceaniac and Atmospheric Administration. NOAA Reviews Record-Setting 2005 Atlantic Hurricane Season. http://www.noaanews.noaa.gov/stories2005/ s2540.htm. [Last accessed May 20, 2013].
26. Lutterloh EC, Iqbal S, Clower JH, Spiller HA, Riggs MA, Sugg TJ. Carbon monoxide poi- soning after an ice storm in Kentucky, 2009. Public Health Rep 2011;126(Suppl 1): 108-15.
27. Centers for Disease Control and Prevention. Use of carbon monoxide alarms to pre- vent poisonings during a power outage-North Carolina, December 2002. MMWR Morb Mortal Wkly Rep 2004;53:189-92.
28. Centers for Disease Control and Prevention. Deaths resulting from residential fires and the prevalence of smoke alarms-United States, 1991-1995. MMWR Morb Mortal Wkly Rep 1998;47(38):803-6.
29. Gunell D, Coope C, Fearn V, Wells C, Chang SS, Hawton K, et al. Suicide by gases in England and Wales 2001-2011: Evidence of the emergence of new methods of suicide. J Affect Disord 2014;170C:190-5.
30. Omaye ST. Metabolic modulation of carbon monoxide toxicity. Toxicology 2002; 180:139-50.
31. Krenzelok E, Roth R, Full R. Carbon monoxide…the silent killer with an audible solution. Am J Emerg Med 1996;14(5):484-6.
32. Mack KA, Rudd RA, Mickalide AD, Ballesteros Michael F. Fatal unintentional injuries in the home in the U.S., 2000-2008. Am J Prev Med 2013;44(3):239-46.