Pediatric food allergy prevalence and trends by race and ethnicity

Abstract

Background:

In the United States, food allergies (FAs) disproportionately affect children based on their racial or ethnic heritage, though there are some gaps and discrepancies in the literature about the prevalence of FAs among different groups.

Objective:

The objective was to analyze clinical data to determine how FA prevalence varies based on race and ethnicity and to track FA diagnosis rates over time.

Methods:

Patient FA diagnoses, birth years, and self-reported races or ethnicities were analyzed for 298,476 pediatric patients in the Children’s Hospital of Philadelphia network. Odds ratios (OR) for FA were calculated for each race or ethnicity, controlling for sex, insurance type, and birth year. The percentages having each of the top 9 FAs and ORs for being allergic to a particular allergen and multiple allergens were calculated for each race or ethnicity using a subset of data for 20,582 food-allergic children. The percentages of each birth cohort diagnosed with FA were plotted.

Results:

Asian/Pacific Islander (A/PI) children had the highest odds of FA (OR, 1.92; CI, 1.80–2.05) and multiple FAs (OR, 1.55; CI, 1.37–1.76). The most common allergens varied between groups. All groups had higher odds than White children of allergies to fish and shellfish, but only A/PI children also had higher odds of peanut, egg, wheat, and sesame allergies. FA diagnoses appeared to increase until the 2004 birth cohort and remain steady through the 2008 birth cohort.

Conclusions:

Clinicians, researchers, and outreach coordinators should be aware of the potential for a significantly greater risk of FA for children of A/PI heritage in the United States. Additional studies that intentionally include patients from these demographics and monitor FA trends over time are warranted.

1. Introduction

Food allergies (FAs) can cause an array of reactions from relatively mild (itchiness, rash, nausea, or stomach upset) to severe (anaphylaxis or death) [1]. These reactions can lead to lower quality-of-life measures, more absences from school, and more frequent hospital visits, which can be emotionally and financially costly for those with FAs and their caregivers [2]. Proper diagnosis with outreach and education can help affected individuals obtain life-saving epinephrine prescriptions, allergen-avoidance help, social support services, and access to developing treatments like oral immunotherapy and IgE-blocking monoclonal antibodies. Awareness of FA rates in various demographic groups may increase the likelihood that FA is considered by healthcare providers as a possible diagnosis for presenting symptoms. In studies on the fatalities caused by anaphylactic reactions to food, it has been shown that this awareness and education is often lacking [3, 4]. Understanding the FA prevalence for demographic groups can also ensure that outreach and education are equitably distributed to affected groups [5, 6].

Previous work on FA epidemiology has identified differences in FA prevalence based on race or ethnicity, although the nature and magnitude of these differences sometimes vary considerably. One potential source of this variation is the type of research method used. Some studies have used surveys of households to identify FA prevalence. These studies have the advantage of potentially capturing cases of FA in families that are not willing or able to seek medical care but are limited by a lack of clinical data to support interviewee reports. However, data gathered in survey studies may be confounded by a lack of “participant honesty, introspective ability, and understanding of complex disease processes” [7] and likely overestimate the true number of FAs [8]. One survey-based study contacted a random sample of US households in 2009–2010 and found an overall self-reported pediatric FA prevalence of 8.0%, with higher odds for Black and Asian but not Hispanic children compared to White children [9]. Another survey conducted in 2015–2016 found a prevalence of 8.7% in children and higher adjusted odds ratios (ORs) in the combined adult and child cohort for Black, Asian, and Hispanic respondents compared to White respondents [10]. It should be noted that the authors of these studies did take steps to avoid the limitations of surveys by using an algorithm to assess patients responses to a variety of questions to make a determination of credible FA [9, 10]. An analysis of the National Health Interview Survey reported an overall FA prevalence of 3.9% among US children in 2007, with a significantly lower prevalence among Hispanic children (3.2%), compared to White (4.1%) or Black (4.0%) children [11]. Starting with the 2007–2008 report, the National Health and Nutrition Examination Survey (NHANES) has included questions about FAs. From these data, an overall prevalence rate of 6.5% has been calculated for children, with a significantly higher prevalence in Black non-Hispanic (NH) children (8.1%) compared to White NH (6.3%) and Hispanic (5.2%) children [12].

The 2005–2006 NHANES collected blood work from a large cross-section of selected US households, including IgE levels to specific food allergens. However, this iteration of the NHANES did not ask participants about their history with FAs, and, in the absence of a patient’s clinical history, positive serum IgE tests alone are not reliable in predicting the severity of FA symptoms [11, 13, 14]. Nevertheless, an analysis of the 2005–2006 NHANES data used serum IgE ranges to estimate “clinically significant” FA rates and found 4.3% of children 1–5 years old and 3.8% of children 6–19 years old fell into this category [15]. Black, but not Hispanic, children had higher odds of FA than White children in this model.

An increasing number of studies use provider-diagnosed FA data to draw conclusions about FA prevalence. A provider-based diagnosis has the advantage of combining the patient history with diagnostic testing and trained clinical judgment. One retrospective cohort study analyzed the medical records of 817 food-allergic children in Illinois or Ohio between 2008 and 2014 and found that rates of fish and shellfish allergy were significantly higher in Black and Hispanic children and rates of wheat and soy allergy were significantly higher in Black children compared to White children. The only allergen more common among White children was tree nut [16]. A nationwide study of nearly 24 million Medicaid records for patients aged 0 to 19 in 2012 found an overall FA prevalence of 0.6%. Those of Pacific Islander, Asian, or Black ancestry had significantly higher odds of FA (OR, 1.26, 1.24, 1.07, respectively), while those of Hispanic (OR, 0.85) and Native American (0.76) descent had lower odds compared to those identifying as White [17].

A prospective study of food-allergic White and Black children enrolled at 1 of 4 major medical centers in Ohio, Illinois, or Washington D.C. between 2017 and 2020 found no differences in the odds of having multiple allergies or of being allergic to any particular allergen, with the exceptions of fish (OR, 2.54) and shellfish (OR, 3.11), which were more common for Black children [18]. Similarly, a prospective study of 1,258 infants up to 36 months in age between 2003 and 2007 found no significant differences between Black children and those of other races in rates of milk, peanut, and egg allergy. The overall FA prevalences were 7.4% and 7.2% for Black and other racial groups, respectively, a nonsignificant difference [19].

The current report describes a retrospective, cross-sectional study that used a database of pediatric patient records from Children’s Hospital of Philadelphia to provide trend and prevalence data for provider-diagnosed FA among children of different races and ethnicities. It is anticipated that these data will help contribute to the effort to obtain accurate estimates of FA prevalence for those of different racial and ethnic backgrounds. Such estimates can be useful for supporting studies on the biological basis of FA as well as interventional efforts targeting FA awareness, education, and support services.

2. Methods

A publicly available dataset, published under a Creative Commons Zero v1.0 Universal license waiving copyright to the work, was used for the analysis [20]. The dataset contained information extracted from electronic patient records for a cross-section of 333,200 pediatric patients who had received care for at least 12 months in the Children’s Hospital of Philadelphia network between January 1, 2001 and December 31, 2013 [7]. The Children’s Hospital of Philadelphia network consisted of more than 30 sites providing both primary and specialty care to urban and suburban patients in Pennsylvania, New Jersey, and Delaware [7]. Health data for patients from this network has been previously shown to be reflective of the surrounding community [21]. The data collection procedure was reviewed by the Institutional Review Board of The Children’s Hospital of Philadelphia and found to be exempt from ethics approval [7].

Categories from the dataset that were analyzed included patient birth year, sex, self-assigned race or ethnicity, insurance type, and incidence of provider-diagnosed allergy to any of the following foods: shellfish, fish, milk, soy, egg, wheat, peanut, sesame, or any tree nut. Incidences of FA were collected from the hospital network’s patient electronic medical records for specific allergens listed within the “t5kmu6,” diagnostic category [7]. The organizers of the dataset categorized patients with codes for lactose intolerance and celiac disease as “non-allergic” to milk and wheat, respectively, to reduce the likelihood of FA misdiagnosis as a confounding factor [7].

Statistical analyses were performed using IBM SPSS Statistics, version 29.0.1.0, for logistic regression and comparative statistics and Microsoft Excel, version 2312, for descriptive statistics. Pearson chi-square tests were used for categorical data.

3. Results

3.1. Full cohort

Patient data where both race and ethnicity were unspecified were removed, leaving data for 298,476 children. The baseline demographics for these children are shown in Table 1. Participants who self-identified as Hispanic were categorized as Hispanic and not placed into a racial group; therefore, all race designations should be considered non-Hispanic. The prevalence of any FA was 6.9% and the percentage of the entire cohort who had multiple allergies was 2.2%. The prevalences, in decreasing order, of having any FA for each race and ethnicity were Asian/Pacific Islander (A/PI) NH (12.1%), Black NH (7.7%), White NH (6.3%), and Hispanic (5.7%). The percentage diagnosed with any FA in each birth year was graphed for each race and ethnicity (Fig. 1). Only those years in which there were at least 100 patients in the represented group were included in the graph. To allow adequate time for FAs to manifest in each cohort, patients who would have been less than 5 years old at the time of data collection (birth year later than December 31, 2008) were also not included in Figure 1. Holding constant birth year, insurance type, and sex, a logistic regression showed significantly higher odds of FA for A/PI NH (OR, 1.919; CI, 1.796–2.051) and Black NH (OR, 1.417; CI, 1.369–1.466) children, and significantly lower odds for Hispanic children (OR, 0.900; CI, 0.837–0.967). The proportion of males with FA (7.7%) was greater than that of females (6.0%), P < 0.001. The proportion of patients on Medicaid with FA (6.5%) was less than the proportion of patients on private insurance (7.0%), P < 0.001.

Table 1.

Baseline demographics of pediatric patients who had race or ethnicity data in a children’s hospital of Philadelphia Network Data Set

Characteristic	Main cohort Frequency, n (%) (Total n: 298,476)	FA subset Frequency, n (%) (Total n: 20,582)
Sex
Male	151,809 (51%)	11,757 (57%)
Female	146,667 (49%)	8,825 (43%)
Insurance Type
Medicaid	79,197 (27%)	5,176 (25%)
Nonmedicaid	219,279 (73%)	15,406 (75%)
Race/ethnicity
White, non-Hispanic	178,346 (60%)	11,211 (54%)
Black, non-Hispanic	95,715 (32%)	7,402 (36%)
Asian/PI, non-Hispanic	9,083 (3%)	1,099 (5%)
Hispanic	15,332 (5%)	870 (4%)
Birth year
1983–1992	36,691 (12%)	1,416 (7%)
1993–2002	124,032 (42%)	7,402 (36%)
2003–2012	137,753 (46%)	11,764 (57%)

Figure 1.

Percent of each birth year diagnosed with food allergy by race or ethnicity. The percentage of a cross-section of pediatric patients diagnosed with food allergy within the Children’s Hospital of Philadelphia network were graphed according to birth year and race/ethnicity. Patients who would have been less than age 5 at data collection and birth cohorts having fewer than 100 patients were not included (total n = 249,002).

3.2. Food allergic subset

From the main cohort, the subset of those with FA was 20,582 children. The demographic data for this subgroup is shown in Table 1. The prevalence with multiple FA was n = 6,579/20582, 32% (Table 2). Of the 9 food allergens analyzed, the maximum number of allergies per patient was 8. Each child had an average of 1.6 (median, 1; SD, 1.0) FAs. The percentages of patients in the FA cohort who were allergic to each of the top 9 food allergens and to multiple allergens is shown in Table 2, along with a breakdown by race and ethnicity. Holding constant birth year, insurance type, and sex, a logistic regression was used to generate ORs of being allergic to specific food allergens and to multiple allergens for each race or ethnicity, using data for White children as the reference (Fig. 2).

Table 2.

Prevalence of top 9 and multiple pediatric food allergies by race or ethnicity in a food-allergic cohort of the Children’s Hospital of Philadelphia Network

Food allergen	All n = 20,582	White, NH n = 11,211	Black, NH n = 7402	A/PI, NH n = 1099	Hispanic n = 870
Peanut	7,916 (38%)	4,531 (40%)	2,613 (35%)	494 (45%)	278 (32%)
Milk	6,611 (32%)	4,212 (38%)	1,747 (24%)	332 (30%)	320 (37%)
Egg	5,454 (26%)	3,250 (29%)	1,461 (20%)	478 (43%)	265 (30%)
Shellfish	4,893 (24%)	1,535 (14%)	2,891 (39%)	291 (26%)	176 (20%)
Soy	2,217 (11%)	1,429 (13%)	548 (7%)	140 (13%)	100 (11%)
Fish	1,667 (8%)	519 (5%)	952 (13%)	124 (11%)	72 (8%)
Tree nut	1,665 (8%)	1,089 (10%)	422 (6%)	96 (9%)	58 (7%)
Wheat	1,041 (5%)	606 (5%)	299 (4%)	87 (8%)	49 (6%)
Sesame	683 (3%)	458 (4%)	133 (2%)	68 (6%)	24 (3%)
Multiple	6,579 (32%)	3,759 (34%)	2,073 (28%)	488 (44%)	259 (30%)

A/PI, Asian/Pacific Islander; NH, non-Hispanic.

Figure 2.

Adjusted associations between race or ethnicity and allergy to top food allergens and multiple allergens. Odds ratios for race or ethnicity and allergy to each of the top food allergens, or to multiple allergens, were based on logistic regressions correcting for birth year, insurance type, and sex, with data for White children as the reference. Data was drawn from a cross-section of pediatric patients who had been diagnosed with at least one food allergy within the Children’s Hospital of Philadelphia network between 2001 and 2013. Total n = 20,582.

4. Discussion

The raw data used in this study were originally collected and analyzed by Hill et al. [7]. However, the race and ethnicity data were not used in the original analysis. In the present study, participants without a specified race or ethnicity were removed from the analysis. Despite this difference, the overall FA prevalence was similar (6.9% here vs 6.7% originally), and the order of the top 5 food allergens remained unchanged (Table 2) [7]. As expected, previous studies using surveys of self-reported FA generally showed higher estimates of overall FA prevalence, but the prevalence calculated from the 2007 to 2008 NHANES survey is closest to the one found in this study [12]. Compared to the results found here, the IgE measurements from the 2005 to 2006 NHANES resulted in lower estimates for overall FA rates [15]. Although a previous study of nationwide Medicaid claims gave a low overall FA prevalence relative to other studies, its conclusions about racial and ethnic prevalences matched the results here [17].

Analysis by racial or ethnic groups revealed all groups had higher odds of being allergic to fish or shellfish than White children, but A/PI children also had higher odds of being allergic to peanuts, eggs, wheat, and sesame. They were also more likely to be allergic to more than one food. The racial prevalence data in this study corresponded well with 2 self-reported surveys that have noted higher FA rates for A/PI and Black children, but not for Hispanic children [9, 12]. The results of a prospective study that found no differences between Black and White children in the odds of having multiple allergies or of being allergic to any particular allergen, with the exceptions of fish and shellfish, reconcile well with the results of the present study [18]. Among A/PI children, the prevalence of having any FA was higher than for any other racial or ethnic group and twice the FA prevalence for Hispanic children. More than 1 in 10 A/PI children had been diagnosed with a FA. This group also had higher odds of multiple FAs. These results support previous studies that have reported similar results [9, 17].

The causes of racial and ethnic differences in FA prevalences have not been determined, but some research suggests a mismatch between environmental factors and genetic heritage. For example, research conducted in Australia showed that first-generation Asian children born in Australia had higher rates of FA than Asian children who were born in Asia and subsequently moved to Australia. The authors suggested the cause may involve a deleterious interaction between genetic factors and the environment [22]. In contrast, Native Americans in the US had the lowest odds of FA among the groups studied in a Medicaid database [17]. Additional research along these lines may yield fruitful insights into the etiology of FAs.

The trend data reported here are consistent with data from the Centers for Disease Control and Prevention, which report a 50% increase in pediatric FA prevalence between 1997–1999 and 2009–2011 [23]. They are also consistent with a report finding that Hispanic children experienced an increase in FA prevalence of about 39% between 1997 and 2007 [11]. In the present study, there was a steady increase of FA diagnoses for each birth cohort until about 2004, at which point FA diagnoses remained steady through the 2008 birth cohort. It is possible the leveling of diagnoses near the end of the dataset is because the cohorts were younger at the time of data collection and therefore had less time to develop allergies, though children less than age 5 at data collection were not included in the analysis to help mitigate this problem. On the other hand, a study examining pediatric patient records in Olmstead County, Minnesota, revealed a decline and stabilization in FA cases between 2008 and 2013, when the patients on the later end of Figure 1 would have been between ages 0 and 17 [24]. When the data from the Minnesota study were later extended, they showed a subsequent increase in FA diagnoses between 2013 and 2018 [25]. It is noteworthy that our study of a population in the Northeast may have mirrored trends observed in Minnesota. The authors of the Olmstead County study suggested stabilization of FA rates between 2008 and 2013 may have reflected changing recommendations on early introduction of allergenic foods to infants by the American Academy of Pediatrics [25]. The possible presence of a transient slowdown in FA diagnoses and its potential causes may merit further research.

A strength of this study is the representative sample of A/PI children, who were not included in some of the previous studies on FA prevalence. Another strength is the data on FA diagnoses over time, as reports on trends in FA diagnosis are uncommon. This study also uses data from healthcare providers, which helps reduce some of the variability inherent in self-reports of FA. A limitation of the study is that it only includes patients who were willing or able to seek treatment from a healthcare provider. It also is limited to only 1 region of the country, which may not be reflective of demographic groups across the United States. Limitations associated with the original dataset include its dependence on diagnostic coding, which may be susceptible to error or variability; however, the dataset organizers manually analyzed the charts of a sample of 240 patients and found a high concordance between the coded diagnoses and current practice guidelines, though diagnoses were not typically based on the use of food challenge tests and made no distinction between IgE-mediated and non-IgE-mediated FAs [7]. While the database used in this study provided a generous breadth of FA data on a large number of pediatric patients, it did lack some information that may have been useful for adding depth and context to this analysis, including patient symptoms at presentation and the diagnostic tests applied by providers. It also primarily included patients from urban and suburban settings, meaning children from rural settings may not be accurately reflected in the dataset [7].

5. Conclusions

This analysis has shown that pediatric patients who are A/PI or Black have a higher prevalence of FA and that A/PI children have an increased risk of having multiple FAs compared to White children. The data also show a trend of increasing FA diagnoses, followed by a leveling off period, which might mirror other study results [25]. The information reported here may be useful for directing outreach efforts to diagnose, educate, and support individuals with FAs and for inspiring research aimed at providing a better understanding of the ongoing life-threatening epidemic of FA in children.

Conflicts of interest

The authors have no financial conflicts of interest.

Author contributions

Study conception: Tyler M. Rose. Acquisition, analysis and interpretation of the data: Francesca B. Rose and Tyler M. Rose. Drafting of the manuscript: Francesca B. Rose and Tyler M. Rose. Final approval of the manuscript: Francesca B. Rose and Tyler M. Rose.