PEDIATRICS Vol. 101 No. 5 May 1998, p. e12

ELECTRONIC ARTICLE:
Predictors of Mortality From Fires in Young Children

Seth J. Scholer^*, Gerald B. Hickson^*, Edward F. Mitchel Jr, and Wayne A. Ray

From the Departments of ^* Pediatrics and  Preventive Medicine, Division of General Pediatrics, Vanderbilt University, Nashville, Tennessee.

	ABSTRACT

Top Abstract Introduction Methods Results Discussion References

Background.  In the United States in 1994, fires claimed 3.75 lives per 100 000 child years and accounted for 17.3% of all injury deaths in children <5 years of age.

Objectives.  To conduct a historical cohort study that uses maternal demographic characteristics to identify young children at high risk of fire-related deaths, thus defining appropriate targets for prevention programs.

Methods.  The cohort consisted of children born to mothers who resided in the state of Tennessee between 1980 and 1995. Information was obtained by linking birth certificates, 1990 census data, and death certificates. Children were eligible for the study if they were <5 years of age at any time within the study period and if key study variables were present (99.2% of births).

Birth certificates provided information on maternal characteristics including age, race, education, previous live births, use of prenatal care, and residence (in standard metropolitan statistical area). Child characteristics included gender, gestational age, and birth type (singleton/multiple gestation). Neighborhood income was estimated by linking the mother's address at the time of birth to the 1990 census (block group mean per capita income).

The study outcome was a fire resulting in at least one fatality (fatal fire event) during the study period, identified from death certificates (coded E880 through E889 in the International Classification of Diseases, 9th rev). We calculated the fatal fire event rate corresponding to each stratum of maternal/child characteristics. We assessed the independent association between each characteristic and the risk of a fatal fire event from a Poisson regression multivariate analysis.

Results.  During the study period, 1 428 694 children contributed 5 415 213 child years to the cohort: there were 270 deaths from fire (4.99 deaths per 100 000 child years) and 231 fatal fire events. In the multivariate analysis, factors associated with greater than a threefold increase in fatal fire events included maternal education, age, and number of other children. Compared with children whose mothers had a college education, children whose mothers had less than a high school education had 19.4 times (95% confidence interval [CI], 2.6-142.4) an increased risk of a fatal fire event. Children whose mothers had more than two other children had 6.1 times (95% CI, 3.8-9.8) an increased risk of a fatal fire event compared with children whose mothers had no other children. Children of mothers <20 years of age had 3.9 times (95% CI, 2.2-7.1) increased risk of a fatal fire event compared with children whose mothers were 30 years old. Although both maternal neighborhood income and race were associated strongly with increased rates of fatal fire events in the univariate analysis, this association did not persist in the multivariate analysis. Other factors that were associated with increased risk of fatal fire events in the multivariate analysis were male gender and having a mother who was unmarried or who had delayed prenatal care.

The three factors associated most strongly with fire mortality were combined to create a risk score based on maternal education (16 years, 0 points; 13 to 15 years, 1 point; 12 years, 2 points; <12 years, 3 points); age (30 years, 0 points; 25 to 29 years, 1 point; 20 to 24 years, 2 points; <20 years, 3 points); and number of other children (none, 0 points; one, 1 point; two, 2 points; three or more, 3 points). The lowest-risk group (score <3) included 19% of the population and had 0.19 fatal fire events per 100 000 child years. In contrast, highest-risk children (score >7) comprised 1.5% of the population and had 28.6 fatal fire events per 100 000 child years, 150 times higher than low-risk children. Children with risk scores >5 contributed 26% of child years but experienced 68% of all fatal fire events. If the fatal fire event rate for all children had been equal to that of the low-risk group (risk score <3), then 95% of deaths from fires would not have occurred.

Discussion.  Maternal education, age, and number of other children had strong and independent associations with fire-related deaths among young children. Taken together, these factors defined a steep risk gradient, where children in the highest-risk group had a fire-related mortality rate that was 150 times that of the lowest-risk group. From a public health perspective, maternal factors clearly define children who would be good candidates for prevention programs. There is an urgent need to develop prevention programs that can be shown to reduce fire-related injury in high-risk children.

	INTRODUCTION

Top Abstract Introduction Methods Results Discussion References

Injuries account for >40% of deaths in children 1 to 4 years of age,¹ and one leading cause of injury death is fires. In the United States in 1994, fires claimed 3.75 lives per 100 000 child years and accounted for 17.3% (741/4286) of all injury deaths in children <5 years of age.² Acknowledging fires as an important public health problem, the Department of Human Services has made reduction in mortality from fires a major goal in the Healthy Children 2000 disease prevention objectives.³ This goal should be feasible because there are effective intervention programs demonstrated to decrease the risk of serious fire-related injuries.⁴ However, to be effective, such prevention programs must target high-risk groups.

We conducted a historical cohort study of young children born in Tennessee over a 16-year period using linked birth-death certificates. Our study sought to identify maternal characteristics that were associated independently with fire-related deaths in young children, and thus to better define children at high risk of death from fires.

	METHODS

Top Abstract Introduction Methods Results Discussion References

Cohort

The cohort consisted of children born to mothers who resided in the state of Tennessee between 1980 and 1995. Information was obtained by linking birth certificates, 1990 census data, and death certificates. Children were eligible for the study if they were <5 years at any time within the study period and if key study variables were present (99.2% of births). Follow-up began at the latter of the date the study began (January 1, 1980) or the date of the child's birth. Follow-up ended at the earliest of the date the study ended (December 31, 1995), the attainment of age 5, or the date of death.

Maternal/Child Characteristics

Birth certificates provided information on maternal characteristics including age, race, education, previous live births, use of prenatal care, and residence (in standard metropolitan statistical area). Child characteristics included gender, gestational age, and birth type (singleton/multiple gestation). We estimated the number of other children using previous live births, adjusting for birth type. A validation study found high concordance between characteristics recorded in birth certificates and medical charts.⁵ Neighborhood income was estimated by linking the mother's address at the time of birth to the 1990 census (block group mean per capita income).⁶

Outcomes

The study outcome was a fire resulting in at least one fatality (fatal fire event) during the study period, identified from death certificates (coded E880 through E889 in the International Classification of Diseases, 9th Rev). We studied fatal fire events rather than individual fire deaths because the latter could not be considered an independent outcome when multiple children died in the same fire. Although we did not capture deaths of children who were born in Tennessee and subsequently moved to another state, a previous assessment of intrastate migration suggests that interstate migration (and corresponding underestimate of mortality rates) does not exceed 10%.¹⁰

Analysis

We calculated the number of fatal fire events and child years corresponding to each stratum of maternal/child characteristics. The incidence density rates¹¹ for each stratum were calculated by dividing the number of fire events by the child years in the stratum. Univariate incidence density rate ratios were calculated by dividing each rate within a stratum by that of a reference category. We assessed the independent association between each characteristic and the risk of a fatal fire event from a Poisson regression multivariate analysis¹² with adjusted incidence density rate ratio estimates. All analyses were conducted using SAS version 6.11 under Window NT 3.51 on Intel Pentium.

	RESULTS

Top Abstract Introduction Methods Results Discussion References

During the study period, 1 428 694 children contributed 5 415 213 child years to the cohort; there were 270 deaths from fire (4.99 deaths per 100 000 child years) and 231 fatal fire events. Death was attributable to smoke or fumes (51.1%), burns (45.0%), and unspecified cause related to fire (3.9%).

The number of fatal fire events was associated strongly with maternal characteristics (Table 1). In the multivariate analysis, factors associated with greater than a threefold increase in fatal fire events included maternal education, age, and number of other children. Compared with children whose mothers had a college education, children whose mothers had less than a high school education had 19.4 times (95% confidence interval [CI], 2.6 to 142.4) increased risk of a fatal fire event. Children whose mothers had two or more other children had 6.1 times (95% CI, 3.8 to 9.8) increased risk of a fatal fire event compared with children whose mothers had no other children. Children of mothers <20 years of age had 3.9 times (95% CI, 2.2 to 7.1) increased risk of a fatal fire event compared with children whose mothers were 30 years old. Although both maternal neighborhood income and race were associated strongly with increased rates of fatal fire events in the univariate analysis, this association did not persist in the multivariate analysis. Other factors that were associated with increased risk of fatal fire events in the multivariate analysis were male gender and having a mother who was unmarried or who had delayed prenatal care.

The three factors associated most strongly with fire mortality were combined to create a risk score (Fig 1). Children received more points (associated with increased risk) if they were born to mothers who had less education (16 years, 0 points; 13 to 15 years, 1 point; 12 years, 2 points; <12 years, 3 points); younger age (30 years, 0 points; 25 to 29 years, 1 point; 20 to 24 years, 2 points; <20 years, 3 points); and more other children (none, 0 points; one, 1 point; two, 2 points; three or more, 3 points). The lowest-risk group (score <3) included 19% of the population, and had 0.19 fatal fire events per 100 000 child years. In contrast, highest-risk children (score >7) comprised 1.5% of the population and had 28.6 fatal fire events per 100 000 child years, 150 times higher than low-risk children. Children with risk scores >5 contributed 26% of child years, but experienced 68% of all fatal fire events. If the fatal fire event rate for all children had been equal to that of the low-risk group (risk score <3), then 95% of deaths from fires would not have occurred.

View larger version (26K): [in this window] [in a new window]   Fig. 1.   Fatal fire event rate for children 4 years of age, based on maternal education, age, and number of other children (1980 to 1995). *Risk score calculation is based on maternal education, age, and number of other children. **Numbers above bars indicate the fatal fire event rate (per 100 000 child years) for children with the corresponding risk score.

	DISCUSSION

Top Abstract Introduction Methods Results Discussion References

Maternal education, age, and number of other children had strong and independent associations with fire-related deaths among young children. Maternal education had a particularly pronounced association; children of mothers with less than a high school education had a 19-fold increased risk of death compared with children of mothers with a college education. Taken together, maternal education, age, and number of other children defined a steep risk gradient, where children in the highest-risk group had a fire-related mortality rate that was 150 times that of the lowest-risk group.

Previous research has emphasized the importance of race and income¹³ as risk factors related to mortality from fires. Interestingly, in the univariate analysis of our study, low income and African-American race were associated with a fourfold and twofold increased risk of fire mortality, respectively. In the multivariate analysis, income did show a nonsignificant trend, but neither income nor race had an independent effect on the risk of death from fire.

It is likely that the association of fire-related mortality with maternal characteristics is mediated by behavioral and environmental factors; these include cigarette use,¹⁷ alcohol consumption,^,, use of a space heater,^, lack of a smoke detector,¹⁷ and residence in a mobile home.¹⁷ Of these factors, Runyan found that the greatest increase in risk of fire-related mortality was associated with lack of a smoke detector (odds ratio = 3.4) and the presence of an alcohol-impaired person (odds ratio = 7.5).¹⁷ Given that the presence of a functioning smoke detector is a relatively inexpensive strategy to reduce risk of mortality from fires, a logical first step might be to increase the proportion of residential dwellings with working smoke detectors. However, even if the risk of fire-related mortality can be reduced three- to fourfold by ensuring functioning smoke detectors, this will not erase completely the disparity in mortality rates that we found. Our results indicate that a more comprehensive approach will be necessary to substantially narrow the gap in mortality rates between high- and low-risk children. As for what to include in a more comprehensive intervention program, there is currently no clear solution. Additional work is needed to determine which intervention strategies have the greatest likelihood of being relatively cost-effective.

From a public health perspective, maternal factors clearly define a very-high-risk group of children who would be good candidates for prevention programs. Also, these same risk factors identify children at increased risk of other major causes of injury mortality (motor vehicle trauma, drowning, suffocation/aspiration, and intentional), although not to the same magnitude.¹⁰ Because a recent study demonstrates that interventions targeting high-risk populations can succeed in reducing fire-related injury,⁴ it is extremely important to be able to accurately identify those at risk. In the state of Tennessee, our study suggests that it is possible to use three maternal characteristics to identify newborns who have alarmingly elevated rates of death from fire. If all children had a fire-related death rate equal to that of the low-risk group, 95% of deaths would not have occurred. Thus, there is an urgent need to develop prevention programs that can be shown to reduce fire-related injury in high-risk children.

	FOOTNOTES

Received for publication Oct 20, 1997; accepted Jan 27, 1998.

Reprint requests to (S.J.S.) Vanderbilt University Medical Center, 5028 Medical Center East, Nashville, TN 37232-8555.

	ABBREVIATIONS

CI, confidence interval.

	REFERENCES

Top Abstract Introduction Methods Results Discussion References

1. National Center for Health Statistics. Vital Statistics of the United States, 1984. Hyattsville, MD: US Dept of Health and Human Services, Public Health Service, NCHS; 1987
2. US Dept of Health and Human Services. National Summary of Injury Mortality Data. Washington, DC: National Center for Health Statistics, Vital Statistics. Available at: http://www.cdc.gov/ncipc/osp/us9491/ofire.htm. Accessed April 4, 1997
3. US Dept of Health and Human Services. Healthy Children 2000. Washington, DC: US Public Health Service; 1992
4. Mallonee S, Istre GR, Rosenberg M, Surveillance and prevention of residential-fire injuries. N Engl J Med. 1996; 335:27-31[Abstract/Free Full Text]
5. Piper JM, Mitchel EF Jr, Snowden M, Hall C, Adams M, Taylor P Validation of 1989 Tennessee birth certificates using maternal and newborn hospital records. Am J Epidemiol. 1993; 137:758-768[Abstract/Free Full Text]
6. Census of Population and Housing. Summary tape file 3 on CD-ROM for Tennessee. Washington, DC: Bureau of the Census; 1990
7. Census of Population and Housing. Summary tape file 3 on CD-ROM; Technical documentation. Washington, DC: Bureau of the Census; 1992
8. TIGER/Line Files. Washington, DC: Bureau of the Census; 1993
9. TIGER/Line Files. Technical documentation. Washington, DC: Bureau of the Census; 1992
10. Scholer SJ, Mitchel EF, Ray WA Predictors of injury mortality in early childhood. Pediatrics. 1997; 100:342-347[Abstract/Free Full Text]
11. Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic Research Principles and Quantitative Methods. New York, NY: Van Nostrand Reinhold Company, Inc; 1982
12. McCullagh P, Nelder JA. Generalized Linear Models. 2nd ed. New York, NY: Chapman and Hall; 1989
13. Fingerhut LA, Kleinman JC, Malloy MH, Feldman JJ Injury fatalities among young children. Public Health Rep. 1988; 103:399-405[Medline]
14. Gulaid JA, Onwuachi-Saunders EC, Sacks JJ, Roberts DR Differences in death rates due to injury among blacks and whites, 1984. MMWR. 1988; 37:25-31
15. Anonymous Leads from the MMWR. Differences in death rates due to injury among blacks and whites, 1984. JAMA. 1989; 261:214-216[CrossRef][Medline]
16. Ballard JE, Koepsell TD, Rivara FP, van Belle G Descriptive epidemiology of unintentional residential fire injuries in King County, WA, 1984 and 1985. Public Health Rep. 1992; 107:402-408[Medline]
17. Runyan CW, Bangdiwala SI, Linzer MA, Sacks JJ, Butts J Risk factors for fatal residential fires. N Engl J Med. 1992; 327:859-863[Abstract]
18. Chemichko L, Saunders LD, Tough S Unintentional house fire deaths in Alberta, 1985-1990: a population study. Can J Public Health. 1993; 84:317-320[Medline]
19. Barillo DJ, Goode R Substance abuse in victims of fire. J Burn Care Rehabil. 1996; 17:71-76[CrossRef][Medline]
20. Ballard JE, Koepsell TD, Rivara F Association of smoking and alcohol drinking with residential fire injuries. Am J Epidemiol. 1992; 135:26-34[Abstract/Free Full Text]
21. Mierley MC, Baker SP Fatal house fires in an urban population. JAMA. 1983; 249:1466-1468[Abstract]