- Top of page
- Materials and methods
Background: Previous US population-based epidemiologic studies of anaphylactic deaths have been limited by small populations and/or few deaths. The objective of this study was to determine the 10-year incidence of death from anaphylaxis in Florida and its descriptive epidemiology.
Methods: Patients who died from anaphylaxis from 1996 to 2005 were identified from ICD-9 and ICD-10 codes on death certificates statewide. Age, race and gender-specific anaphylactic death rates were calculated.
Results: There were 89 deaths among Florida residents. The individuals with autopsy confirmed diagnoses, and those with clinical diagnoses only, did not differ with regard to race, anaphylactic triggers or the clinical variables of lung and heart disease. Annual death rate for anaphylaxis in Florida was 5.02/10 000 000. The relative risk of death from anaphylaxis was 14.09 for individuals ≥65 years old (P = 0.0000002) and 6.38 for individuals 35–64 years old (P = 0.0019) compared with those who were 5–14 years of age. Deaths among Florida residents that occurred in emergency rooms or outpatient settings were 2.11 times as likely to be anaphylactic deaths than deaths that occurred in inpatient settings (P = 0.0026). The ratios of anaphylactic deaths to total deaths in March and April and in July and August were greater than the ratios for the other bimonthly periods (P = 0.02).
Conclusion: Death from anaphylaxis in Florida was more likely to occur in older individuals, in an emergency department, and in the months of March and April and July and August.
Anaphylaxis is a potentially life-threatening syndrome which occurs as a result of mast cell and basophil mediator release. This cellular activation usually occurs following exposure to an antigen to which an individual has been sensitized and produced antigen-specific IgE but may also result from exposure to a nonspecific mast cell activator (1). Two US population-based studies of anaphylaxis have previously been reported. One based on a medical record review in Olmsted County, MN, reported one death during a 5-year period (2). The other study of 229 422 children and adolescents enrolled in a health maintenance organization reported no deaths during a 5-year period (3). A recent population-based study of hospitalization for anaphylaxis in Florida in 2001 found two deaths among hospitalized patients that year (4). Population-based European studies have reported one death (UK) (1994–1999) (5) and 24 deaths (Switzerland) (1996–1998) (6). One study of the entire UK population reported 214 deaths over a 10-year period (1992–2001) (7). The previous US studies analyzed only limited populations. We conducted a descriptive study and detailed analysis of a case series to determine the epidemiology of death from anaphylaxis in Florida. Our goals were to evaluate the incidence during a 10-year period of death from anaphylaxis, place of death, anaphylactic triggers, age, sex and clinical characteristics of the victims using information contained on their death certificates. This use of statewide death certificate data represents a novel approach in the study of death from anaphylaxis in the United States.
- Top of page
- Materials and methods
We report the average annual incidence of death from anaphylaxis in Florida to be 5.02 per 10 000 000 population from 1996 to 2005. This compares with one death from anaphylaxis in Olmsted County, MN (approximate population of 99 238 people in 1985) (16) over the 1983–1987 5-year interval (2) in which the average annual incidence of anaphylactic death was 20.15 per 10 000 000 population. However, because only one death occurred in Olmsted County, it is possible that the higher average annual incidence recorded there is a chance occurrence and does not represent the true incidence of anaphylactic death. Another possible explanation of this difference is that anaphylactic death is not diagnosed in 75% of cases in Florida. Underdiagnosis of anaphylaxis has been documented in other settings (17). However, the average annual incidence of anaphylactic death in Olmsted County was not different from that in Florida by chi-squared testing (Table 5) (P = 0.45). It is also believed that underdiagnosis of anaphylactic death as a result of food allergy may occur because these deaths may be coded as asthma deaths (18). A Canadian study revealed that 30 of 32 who died from anaphylaxis died with a respiratory condition listed as the cause of death (19). However, a Swedish study (20) did not find anaphylactic deaths to have been misdiagnosed as asthma deaths (T. Foucard, personal communication). A review of the epidemiology of anaphylaxis in the US estimated that there are approximately 1500 deaths annually with a fatality rate for anaphylaxis of 55/10 000 000 people (21). Our data suggest that this latter estimate may be inaccurately high.
Table 5. Comparison of the average annual incidence of death from anaphylaxis in Florida 1996–2005 with other published population-based anaphylactic mortality rates
|Region years evaluated (reference)||Number of cases||Total cumulative population for years indicated||Anaphylactic deaths/10 million population (95% CI)*||P (Fisher’s exact test)†|
|Florida 1996–2005||89||177 163 075||5.02 [4.08–6.18] || |
|Olmsted Co., Minn, 1983–1987 (2)||1||496 190||20.15 [3.56–114]||0.45|
|UK 1992–2001 (7)||214||573 337 000||3.73 [3.26–4.27]||0.24|
|UK 1994–1999 (5)||1||8 000 000||1.25 [0.21–7.08]||0.19|
|Switzerland 1996–1998 (6)||24||21,721,389||11.06 [7.41–16.44]||0.002|
Another comparison for the average annual incidence of death from anaphylaxis in Florida is the average annual incidence of death from anaphylaxis in the United Kingdom, 3.73/10 000 000 (7, 22) and 1.25/10 000 000 (5), and Switzerland, 11.06/10 000 000 (6). Table 5 shows the comparison of the average annual incidence of anaphylactic death in the previous studies with the average annual incidence in Florida. The average annual incidence in Switzerland (11.06/10 000 000) was significantly greater than that in Florida by the Fisher’s exact test (P = 0.002). The UK incidence rate reported using the U.K General Practice Research Database (5) (1.25/10 000 000) should be considered an estimate of the incidence proportion (personal communication, K. Wilcox Hagberg, MPH, Boston Collaborative Drug Surveillance Program). This finding and the other UK study were not different by the Fisher’s exact test (Table 5) (P = 0.19 and 0.24, respectively) from that found in Florida.
We report that the likelihood of death from anaphylaxis increases with increasing age. This is consistent with our earlier finding that the likelihood of hospitalization from anaphylaxis similarly increases with increasing age (4). The present study is the only study to report age-group specific anaphylactic death rates. The previous British study (7) did not report age-specific mortality rates.
We found no difference in the likelihood of death from anaphylaxis between men and women, and between White and non-White individuals. This finding is consistent with previous estimates of anaphylactic death. However, those reports did not account for the total population of each race and each sex. (7). This study is also the only one to report anaphylactic proportional mortality (see Methods) by location of death (i.e. hospital inpatient, emergency department/clinic, home, etc) (Table 3).
The information on the death certificates on anaphylactic triggers was available on only 42 of the 84 death certificates (50%). In the cases in which the anaphylactic trigger was specified, we found that 50% were because of medications, 20% because of Hymenoptera stings, and 16% and 14% because of foods and radiocontrast media, respectively. The British population-based study (7) found that 37% were because of medications, 22% because of Hymenoptera stings, 27% because of foods and 5% because of radiocontrast media reactions. Our findings approximate the British data. The proportion of food-related anaphylaxis that we observed was similar to that found in Sweden (20).
Men were more likely than women to have died from Hymenoptera anaphylaxis than from anaphylaxis from all other causes. This finding is consistent with our previous finding that men were more likely than women to have been hospitalized for Hymenoptera anaphylaxis (4).
We found that anaphylactic death in Florida was more likely to occur in the months of March and April and again in July and August. The increase in deaths in the summer months was consistent with our previous report that hospitalization for anaphylaxis in Florida in 2001 was increased in the third quarter of that year (23). That increase in hospitalizations was accounted for by Hymenoptera sting anaphylaxis (23). We are not able to make a similar association of a seasonal change in Hymenoptera sting anaphylaxis with the seasonal change in anaphylactic death. The Olmsted County study (2) found that the incidence of anaphylaxis was greatest in July, August and September in Minnesota. This was attributed by the authors to insect sting anaphylaxis (2).
This report has several strengths. The use of a series of statewide death certificates is a novel approach to study anaphylactic mortality. It is the first to study the epidemiology of anaphylactic death in the United States for a population of more than 16 000 000 people. A search of the PubMed/MEDLINE database did not reveal any similar large population-based studies in the United States. This study is also the first to report age-specific anaphylactic death rates. In addition, this is the first study to report anaphylactic deaths per population for gender and race. It is also the first to report anaphylactic proportional mortality by location and by month of death.
Our analysis was limited by our inability to confirm the accuracy of the complete death certificate record using another database. Underdiagnosis of anaphylaxis is believed to be common (17). To further evaluate the possibility of underdiagnosis, Florida in-hospital deaths from anaphylaxis in 2001 as determined from death certificates in the present study were compared with a review of all 10 diagnosis fields (as opposed to using only the principle diagnosis) of the hospital discharge database of the Florida Agency for Health Care Administration [see reference (4) for methods]. Four deaths were captured by both sources. Five in hospital deaths that were likely to be due to anaphylaxis were not so reported on the death certificates and three in hospital anaphylactic deaths as reported on the death certificates were not coded as anaphylaxis in the hospital discharge database. This limited assessment of the completeness of the death certificate data suggests that the actual number of anaphylactic deaths in Florida is higher than it is possible to determine on the basis of an analysis of only a single database.
The diagnosis of anaphylaxis was confirmed by autopsy in 40% of our cases. There was no difference in race, anaphylactic triggers or the clinical variables of lung disease (asthma and COPD) and heart disease between those with an autopsy confirmed diagnosis and those with a clinical diagnosis. However, those with an autopsy confirmed diagnosis were younger and more likely to be men than those with a clinical diagnosis. Nevertheless, we report a significantly higher death rate with increasing age. This suggests that underdiagnosis as a result of ascertainment bias (i.e. there were more anaphylactic deaths with increasing age although there were fewer autopsy confirmed diagnoses with increasing age) was not a significant factor in our analysis. We were unable to verify any of the diagnoses because of the retrospective nature of the study and because the deaths occurred throughout the state of Florida.
In conclusion, this large retrospective investigation of deaths from anaphylaxis in the State of Florida found an average annual incidence of 5.02 deaths per 10 000 000 people. People who were ≥65 years old and who were 35–64 years old, were 14.09 times and 6.38 times, respectively, as likely to die of anaphylaxis than those who were 5–14 years of age. Death from anaphylaxis was more likely to occur in an emergency department or clinic than in the hospital and was more likely to occur in March and April and in July and August than death from other causes.