National Vehicle Emissions Policies and Practices and Declining US Carbon Monoxide–Related Mortality FREE

Acute intoxication from carbon monoxide (CO) has been reported to contribute to more than 2000 poisoning deaths annually in the United States.^1- 4 Common sources of CO poisoning include the operation of motor vehicles in enclosed or semi-enclosed areas, malfunctioning home heating systems, and improperly vented combustion appliances.^2,5 Even in outdoor environments, combustion engine exhaust can produce localized concentrations of CO that are capable of producing unconsciousness and death in minutes.^6- 7

National surveillance for CO-related mortality was last updated over a decade ago, when it was reported that the annual rate of unintentional CO-related deaths decreased in the United States during 1979 through 1988.¹ One possible explanation for the observed decline in death rates is the enforcement of automobile emissions standards set by the 1970 Clean Air Act. To comply with these standards, the first catalytic converters were introduced for automobile use in the United States in 1975. This improved the completeness of engine combustion and reduced CO emissions from passenger vehicles.^8- 11 However, an evaluation of the influence of national vehicle emissions policies and practices on trends in CO-related mortality has not been undertaken. Existing surveillance reports have presented mortality data only for periods following the introduction of the catalytic converter in 1975, leaving preregulatory baseline rates of CO-related deaths unestablished. In addition, multivariate analyses have not been used to relate changes in national rates of CO-related deaths to annual indicators of CO emissions from automobiles. To address these issues, we examine 31 years (1968-1998) of national mortality and motor vehicle emissions data and also provide updated surveillance for CO-related deaths in the United States.

Death certificates contain demographic and geographic information of the decedent, as well as International Classification of Diseases (ICD)^12- 13 codes for up to 20 conditions that contributed to the death. All US death certificate information is compiled by the National Center for Health Statistics, Centers for Disease Control and Prevention into the Multiple Cause of Death Mortality Data Tapes. These cause of death codes are compiled in 2 forms: the entity axis, which contains conditions as stated on the death certificate; and the record axis, which is edited by a computer algorithm to revise some ICD codes to best describe the overall medical certification portion of the death certificate.¹⁴ We identified 31 years (1968-1998) of CO-related deaths in the record axis portion of these data. The ICD, Eighth Revision (ICD-8) was in effect for the years 1968-1978, and the ICD, Ninth Revision (ICD-9) was in effect for the years 1979-1998.

We searched the data tapes for records of all deaths during 1968 through 1998 containing the ICD 3-digit code N986 (toxic effect of CO) as a contributing cause of death and identified 115 987 deaths. A previous national surveillance report used this case definition for case ascertainment.¹ In a slight departure from the previous report, we also searched the data tapes for deaths where N986 was not listed on the death certificate, but where ICD external cause of injury codes (E-codes) exclusive to CO poisoning (ICD-8 = E874, E875, and E952.1; or ICD-9 = E868.3, E868.8, E868.9, E952.1, and E982.1), or poisoning from motor vehicle exhaust (ICD-8 = E873 or E952.0; or ICD-9 = E868.2, E952.0, and E982.0) were listed on the death certificate. Since CO is the only acutely poisonous gas in motor vehicle exhaust,⁸ we also considered these codes to indicate CO poisoning. This change in case definition led to the inclusion of 716 additional deaths during the 31-year period that would not have otherwise been captured (716/116 703 or 0.6% of all CO-related deaths). As data were only available from a 50% sample of death certificates for 1972, numbers of deaths for this year were weighted by a factor of 2.¹⁵

We classified each CO-related death by intent (unintentional, suicide, homicide, intent undetermined, and other) and mechanism (motor vehicle–related, non–motor vehicle–related, and mechanism undetermined), as defined by the E-codes listed on the death record (). Records containing E-codes representing more than 1 intent category (196/116 703 cases or 0.2% of all deaths) were classified as intent undetermined. Records containing E-codes representing both motor vehicle and non–motor vehicle (eg, gas from a domestic stove or fireplace, pipeline, or utility gas) mechanisms of death (5/116 703 cases or 0.01% of all deaths) were classified as mechanism undetermined. We excluded from the study any record containing codes for injuries related to fire and flames (E837, E890-E899), burns (N940-N949), and explosives (E923) as the public health prevention strategies pertaining to this group of deaths focus more on preventing house fires than CO poisoning per se.¹

We used US Census Bureau national resident midpoint population estimates for each year during 1968 through 1998 as denominators to calculate all national rates and trend lines. Log-linear negative binomial and Poisson regression analyses were used to estimate change, and 95% confidence intervals (CIs) for change in rates during the study period.¹⁶ Estimates of change in mortality rates by intent and mechanism of death were calculated for 1968-1998, and for the 3 study periods of 1968-1978 (11 years), 1979-1988, and 1989-1998. As the year 1978 was the final year of implementation of the ICD-8 coding scheme, there was no change in ICD coding within each of these 3 study periods.

We calculated trends in annual estimated CO emissions from automobiles using data provided by the US Environmental Protection Agency. During 1968-1996, the US Environmental Protection Agency collected vehicle model and vehicle age–specific CO emission rate measurements from 20 000 light-duty vehicles in the United States. The measurement of CO emissions was undertaken in a controlled setting during the cold start of the motor vehicle and also during 23 minutes of continuous stationary vehicle operation.¹⁷ This testing procedure resulted in estimates of total vehicle emissions (specified as grams/mile of CO per test mile) that are directly comparable across vehicles of different make and model years. We applied these emission rates (specific to vehicle age and model-year, within calendar year) to annual data on the composition of the US automobile fleet^18- 19 to produce annual estimates of CO emissions per motor vehicle in the United States.

Trend lines of these emissions data were calculated using a normally distributed linear regression model for the periods 1968-1974 and 1975-1996. Negative binomial regression models adjusting for the age, sex, and race of the decedent were used to calculate corresponding trends in rates of motor vehicle–related CO deaths. The cutoff of 1975 was used to examine any effect that the introduction year of the catalytic converter may have had on these trends.

We also used negative binomial regression models to directly examine associations between the annual automobile emissions estimates and annual rates of motor vehicle–related CO poisoning deaths. To estimate the number of unintentional poisoning deaths that would have occurred during 1975 to 1998 if unintentional CO-related death rates remained at pre-1975 levels, we calculated the average annual rate of unintentional CO-related deaths during 1968-1974, and then applied this rate to US population estimates during 1975-1998. SAS version 8 (SAS Institute, Inc, Cary, NC) was used and the a priori significance level for all statistical tests was set at P<.05.

To determine whether better medical treatment may have reduced mortality from CO poisoning, we classified deaths into 2 categories. First, deaths in home or nonclinical settings were classified as deaths without treatment. As these deaths likely involved no medical care, they could not have been prevented by improved hospital-based treatment. Second, persons who died during hospitalization, who died during an outpatient visit to an emergency department, or who were dead on arrival at a hospital, clinic, or medical center were classified as deaths possibly following treatment. We compared changes in the rate of deaths without treatment and deaths possibly following treatment. This analysis was undertaken only for 1979 to 1998, the years when location-of-death data were available on the death certificates.

During 1968 through 1998, non–fire-related CO poisoning was listed as a contributing cause for 116 703 deaths (Table 1). Motor vehicles were mentioned as the mechanism in 70.6% of these deaths. For every unintentional CO-related death that occurred (33 836 deaths), there were 2.2 CO-related suicides (73 940 deaths).

During 1968-1998, CO-related mortality rates in the United States declined by 57.8%. Declining trends in death rates were observed across selected categories of intent and mechanism of death (Table 2). However, greater declines were observed in CO-related unintentional death rates than suicide rates. (Table 2 and Figure 1). The annual rate of unintentional motor vehicle–related deaths did not change during 1968-1978, but this rate declined during 1979-1998. In comparison, unintentional non–motor vehicle–related deaths declined during each of the 3 study decades. Trends in unintentional deaths of undetermined mechanism more closely resembled motor vehicle than non–motor vehicle–related patterns of decline. Annual rates of CO-related suicides remained unchanged during the 1968-1978 period and increased during 1979-1988. As a result, the majority of the decline in CO-related suicides that occurred during 1968-1998 is attributable to declines that occurred during 1989-1998 (Table 2).

Declining trends in CO-related unintentional death rates were observed across age, sex, and race categories and were most pronounced in the 15- to 34-year-old group (Table 3). This was the case for both motor vehicle and non–motor vehicle–related mechanisms of death. Rates of unintentional deaths involving motor vehicles declined least sharply among persons aged 65 years and older. Annual rates of CO-related suicides decreased across age and sex categories. These declines were also greater among women, who had lower baseline rates than men. As with unintentional deaths, decreases in suicides involving motor vehicles were less pronounced among persons aged 65 years and older.

During 1979 to 1998, CO-related mortality rates declined by 66.7% (95% CI, –64.7% to –68.3%) in persons classified as having been possibly treated prior to death and by 47.3% (95% CI, −45.9% to −48.8%) in persons classified as not treated prior to death. As a result, the proportion of all CO-related deaths that occurred to persons who died possibly following treatment decreased from 23.9% to 16.2% during this period. Persons classified as possibly treated prior to death composed 20.9% of all decedents during 1979-1998, and declining deaths in this group accounted for 30.7% of the overall reduction in CO-related deaths rates that occurred since 1979.

Trends in unintentional motor vehicle–related CO death rates, and CO emissions from automobiles, follow nearly parallel trajectories (Figure 2 and BOX 2).^18- 20 Declines in each of these trends began in 1978, when 1975 and newer model-year cars made up 34% of the US passenger vehicle fleet.¹⁸ Annual rates of CO-related suicides involving motor vehicles began to decline in 1988, 10 years following initial declines in CO emissions estimates (Figure 2 inset).

During 1968-1974, CO emissions increased by an estimated 2.0 g/mile (95% CI, 1.3-2.8) or an estimated 2.2% of 1968 levels (95% CI, 1.4%-3.1%). Annual rates of motor vehicle–related CO suicides (14.6%; 95% CI, −33.3% to 96.8%) and unintentional deaths (−1.2%; 95% CI, −33.2% to 46.3%) did not change during 1968-1974. However, during 1975-1996, CO emissions decreased by an estimated 72.8 g/mile (95% CI, −78.2 to −67.2) or an estimated 76.3% of 1975 levels (95% CI, −82.0% to −70.4%). Corresponding rates of motor vehicle–related CO suicides (the annual death rate declined from 10.0 to 4.9, or by an estimated 43.3%; 95% CI, −57.5% to −24.3%) and unintentional deaths (the annual death rate declined from 4.0 to 0.9, or by an estimated 81.3%; 95% CI, −84.8% to −77.0%) also declined during the post-1975 period. Annual rates of non–motor vehicle–related CO deaths declined during both the 1968-1974 (−37.5%; 95% CI, −49.0% to −23.4%) and 1975-1998 (–75.5%; 95% CI, –78.5% to −71.9%) periods.

During 1975-1996, an annual decrease of 10 g/mile of estimated CO emissions from automobiles was associated with a 21.3% decrease (95% CI, –24.2% to –18.4%) in the annual unintentional motor vehicle–related CO death rate and a 5.9% decrease (95% CI, –10.0% to –1.8%) in the annual rate of motor vehicle–related CO suicides.

We estimated that an additional 19 008 unintentional deaths might have occurred from 1975-1998 had annual rates of CO-related deaths remained at pre-1975 levels. Of these deaths, as estimated 11 667 (61%) would have been motor vehicle–related, 4551 (24%) would have been non–motor vehicle–related, and 2790 (15%) would have been from an undetermined mechanism of exposure.

During 1968-1998, the US annual rate of non–fire-related CO poisoning deaths declined by almost 60%. Unintentional CO-related death rates remained unchanged during 1968 to 1978 but then declined by nearly 80% during the following 2 decades. Carbon monoxide–related suicides decreased by 40%, with nearly the entire decline occurring during 1988 to 1998. Although more than two thirds of the CO-related poisoning deaths involved motor vehicles, significant declines in both motor vehicle and non–motor vehicle–related mortality suggest that several factors may have contributed to the prevention of CO-related deaths. Possible explanations include the increasing use of CO alarms in homes, the increased safety of home appliances and consumer products, reduced CO emissions from automobiles, and a reduction in case-fatality rates associated with improved medical treatment of CO-exposed patients.

Of the CO-related poisoning deaths associated with consumer products that occurred during 1993 through 1997, an average of 76% involved indoor heating systems or heaters.⁵ Other consumer products involved in CO-related deaths included charcoal grills and burned charcoal (10%), gas water heaters (4%), camp stoves or lanterns (4%), gas ranges or ovens (4%), or other fuel-burning products (2%). Battery-operated residential CO alarms did not become widely available in the United States until 1992,²¹ when unintentional non–motor vehicle–related deaths had already stabilized at an annual rate of around 1 death per million persons. As a result, increased use of CO alarms cannot explain the observed decline in non–motor vehicle–related mortality, which began more than 20 years earlier. However, the US Consumer Product Safety Commission has issued at least 8 hazard warnings related to CO leakage in home and mobile home furnaces, resulting in subsidized inspection and maintenance for more than 500 000 faulty home furnaces.^22- 30 Unvented gas space heaters manufactured after 1982 were also equipped with oxygen sensors that shut off the system if the surrounding environment became oxygen depleted.^31- 32 In addition, a voluntary standard for home furnaces was introduced by the American National Standards Institute in 1987, requiring a blocked vent detection and safety shutoff mechanism on the draft hood of natural draft furnaces.³³

Although these and other consumer product–related interventions are noteworthy, more than 70% of CO-related deaths that occurred during the study were motor vehicle–related. In 1970, Congress established the Environmental Protection Agency and enacted the Clean Air Act, setting strict automotive emissions standards beginning with the 1975 model year.⁹ To help achieve these standards, the first catalytic converters appeared for use in vehicles in the United States in 1975, reducing CO emissions by removing products of incomplete combustion.^8- 10 The further introduction of 3-way catalysts, onboard computers, and oxygen sensors in new vehicles occurred in 1981, when new cars manufactured in the United States were required to meet a CO emissions standard that was 90% lower than the 1970 standard. Thus, the newest motor vehicles produce less than 5% of the CO emissions of pre-1975 vehicles.^8- 11

During 1968-1974, CO emission rates from light-duty vehicles increased slightly, and annual rates of motor vehicle–related CO poisoning deaths remained unchanged. However, following the 1975 introduction of the catalytic converter, emission rates and motor vehicle–related death rates began sharp and statistically significant declines. Unintentional motor vehicle–related CO deaths were most strongly associated with the emissions estimates, and both began concurrent declines in 1978, when one third of the US passenger vehicle fleet included vehicles made after 1975. In comparison, statistically significant declines in non–motor vehicle–related poisoning deaths were evident during both pre-1975 and post-1975 periods. During 1968 to 1996, an annual decrease of 10 g/mile of estimated CO emissions from automobiles was associated with a 21.3% decrease in the average annual unintentional motor vehicle–related CO death rate, independent of the age, sex, and race of the decedents.

Higher rates of CO poisoning deaths may occur in countries where emission standards for motor vehicles have not been implemented. In Great Britain, no decline in motor vehicle–related CO deaths was evident until 1993, when catalytic converters first were required.^34- 35 However, as of 1997, suicide rates from motor vehicle exhaust gas in Australia continued to rise, despite the introduction of catalytic converters to the Australian automobile fleet in 1986.³⁶ The United States set environmental limits on CO emissions from automobiles at 15.0 g/mile in 1975 and reduced this standard to 3.4 g/mile for automobiles manufactured in 1981 or later. Standards equivalent to 15.0 g/mile and 3.4 g/mile were not established in Australia until 1986 and 1997, respectively, and may help explain the decline in CO-related suicide rates observed only in the United States.³⁶

Improvements in hospital-based medical care for CO-poisoned patients may also have influenced the decline in CO-related death rates, as the proportion of all CO-related deaths that occurred to persons classified as possibly treated prior to death decreased during the study. However, only a small proportion of CO-related deaths occur at medical facilities.³⁷ In this case, we estimated that 21% of all of the CO-related deaths that occurred during 1979-1998 were possibly treated prior to death. If all of the deaths that were recorded as having occurred in a hospital or clinical care setting in 1979 had been prevented through improved treatment, the CO-related death rate that year would have decreased by 24%. While considerable, this is smaller than the more than 50% decrease in the annual rate of CO-related deaths that was observed since 1979.

Due to the ecologic nature of this study, we are limited in our ability to draw causal inferences about the association between individual exposures and CO-related deaths. However, the concurrent decline in motor vehicle–related emissions and poisoning deaths that only occurred following the first national intervention to reduce CO in automobile exhaust appears unlikely to be coincidental. While we were able to demonstrate that alternative explanations did not fit the data as well, individual-level measures of exposure would be needed to confirm these suggested etiologic relationships. Other limitations associated with CO data from death certificates include the absence of circumstantial detail, such as drug and alcohol use, the inconsistent assignment of mechanism, intent, and place of death, and the inconsistent application of ICD codes by nosologists.¹ However, as national awareness of CO as a source of fatal poisonings increased during the study, it appears unlikely that a tendency away from coding CO as a contributing cause of death could explain the observed decrease in CO-related death rates.

The federal regulation of motor vehicle emissions was associated with the decline in motor vehicle–related CO death rates that occurred in the United States after 1975. Similar to the rapid decline in population blood lead levels that followed the removal of lead from gasoline,³⁸ an estimated 11 700 unintentional motor vehicle–related poisoning deaths may have been averted since the year of introduction of the catalytic converter, reflecting what appears to be a public health benefit associated with the enforcement of the 1970 Clean Air Act. However, exposure to CO remains a leading cause of poisoning death, resulting in more than 1700 suicides and 500 unintentional deaths annually. Carbon monoxide–related deaths can and do occur with newer vehicle models³⁶ as the effectiveness of the catalytic converter decreases over its useful life,^39- 40 in oxygen-depleted environments,⁸ and during the cold start of vehicles.⁸ Localized intervention efforts, continued enforcement of emissions standards, and the identification and prevention of ongoing mechanisms of exposure remain important to continued public health success.

Intent of Death
Unintentional injury
ICD-8 and ICD-9: E800-E949
Suicide
ICD-8 and ICD-9: E950-E959
Homicide
ICD-8 and ICD-9: E960-E969
Intent undetermined
ICD-8 and ICD-9: E980-E989 or records including E-codes representing >1 category of intent of death (this occurred for 196/116 703 [0.2%] of all cases)
Other intent
ICD-8 and ICD-9: E970-E979 and E990-E999

Mechanism of Death
Fire-related (excluded from analyses)
ICD-8 and ICD-9: E837, E890-E899, E923, or N940-N949
Motor vehicle–related
ICD-8: E800-E807, E810-E825, E830-E836, E838, E840, E841, E843-E845, E873, E927, E940-E941, and E952.0; ICD-9: E800-E807, E810-E825, E830-E836, E838, E840, E841, E843-E847, E868.2, E929.0, E929.1, E952.0, E958.5, E958.6, E982.0, E988.5, and E988.6
Non–motor vehicle–related
ICD-8: E870-E872, E874, E920, E927, E928, E951, E981; ICD-9: E867, E868.0, E868.1, E868.3, E919, E920, E951, E981
Mechanism undetermined
All other E-codes or records including E-codes representing >1 category of mechanism of death (this occurred for 5/116 703 [0.01%] of all cases)

*ICD-8 indicates International Classification of Diseases, Eighth Revision; ICD-9, International Classification of Diseases, Ninth Revision.

1970: Congress enacts Clean Air Act. CO emissions standard at 34.0 g/mile.
1975: Catalytic converter introduced on new passenger cars to meet new CO emissions standard of 15 g/mile.
1978: The 1975 and newer model-year cars make up 34% of the US passenger vehicle fleet.
1980: New passenger cars required to meet new CO emissions standard of 7.0 g/mile. The 1975 and newer model-year cars make up 50% of US passenger vehicle fleet.
1981: New cars required to meet new CO emissions standard of 3.4 g/mile.
1990: The 1975 and newer model-year cars make up 91% of the US passenger vehicle fleet.
1992: Standards setting emission limits for CO at temperatures less than –7 °C are established. Oxygenated gasoline is introduced in cities with high CO levels.
*Taken from Environmental Protection Agency and EPA Fact Sheet OMS-12.²⁰