Background
Lower respiratory illnesses (LRIs) and associated complications are the leading cause of death in very young children [
1], and LRIs and asthma are associated with increased morbidity [
2]. In addition, asthma is among the most common childhood diseases in the United States (US), with 1,406,000 children under the age of 5 years diagnosed with asthma [
3]. This is particularly important because the lungs of very young children are still developing, making them more susceptible to respiratory health risks such as air pollution, compared to adults [
4,
5]. Additionally, very young children who have LRIs are more likely to develop respiratory issues later in life [
6‐
8].
Asthma and LRI morbidity are influenced by complex relationships among social and environmental factors such as socio-economic status (SES), housing conditions, and ambient air pollution, in addition to genetic predispositions, lifestyle habits, and psychological stressors [
9‐
13]. Past studies have linked children’s asthma and LRI emergency department (ED) visits and hospital admissions to a substantial list of social and environmental factors, including but not limited to lower household income, minority race [
14], older housing structures [
15], household crowding [
16], and increased ambient air pollution [
17‐
20]. Further, relationships among these factors and respiratory diseases have been found to vary across regions and spatial scales (e.g., census tract, county, state, etc.) [
21‐
23].
While independent social and environmental risk factors have been identified for these diseases, few studies have examined multiple risk factors in children at the community level [
12,
14,
15,
24]. Of those studies that have, none have focused exclusively on children <5 years of age, who are more physiologically vulnerable and spend more time at home (66% – 77% on average) compared to older children [
25‐
27]. Combined with the impact these diseases in early childhood may have on respiratory health later in life [
6‐
8], it is crucial to better understanding these relationships. By better understanding the complex, multi-factorial relationships between community-level predictors and these respiratory disease outcomes, we can better design public health interventions for specific areas or communities to reduce their prevalence in children under 5 years [
28]. In this study, our goal was to better understand how socio-economic and housing characteristics, ambient air pollution levels, and geographic variables related to asthma and LRI ED visits and hospitalization rates in children under 5 years at a community level for years 2005–2009 in Maricopa and Pima Counties in Southern Arizona.
Discussion
While independent social and environmental risk factors have been identified for these diseases, few studies have examined their relationship to multiple predictors at the neighborhood level [
12,
14,
15,
24,
46]. Of those studies that have, none have focused exclusively on children <5 years, who are more physiologically vulnerable and spend more of their time at home (ranging from 66 to 77% on average) compared to older children [
25‐
27]. In our study, we investigated the relationships between socio-economic and housing characteristics, ambient air pollution levels, and geographic variables and asthma and LRI ED visits and hospitalization rates in children under 5 years at a community level in Maricopa and Pima Counties in Southern Arizona. Socio-economic characteristics and ambient air pollutant levels were combined into unitless indices (i.e., lower SES and increased air pollution levels) using PCA, accounting for 56% and 72% of variance, respectively. Housing characteristics variables did not exhibit moderate-to-high correlations (i.e., ρ > 0.30 and
p < 0.05 in a Bonferroni-corrected Pearson correlation) and thus were not combined using PCA. Nearly all predictors, except household crowding, were significantly related to one or more outcomes in the multiple regressions. Notably, lower SES and reduced population density were associated with asthma hospitalization and both LRI outcomes. Living in Pima County was protective against asthma and LRI hospitalization, when compared to Maricopa County, which had significantly higher air pollution levels, among other factors (Table
2). Multiple regression model residuals did not exhibit spatial autocorrelation, satisfying the regression assumption of independent observations. Our results have provided more information on the complex, multi-factorial relationships between asthma and LRIs and geographic factors at the neighborhood level based off home census tract. By better understanding these relationships, it may be possible to design more focused public health interventions at the community level to prevent and better control these diseases in children under 5 years, who spend most of their time at home and are more physiologically vulnerable compared to older children.
When predicting ED visits and hospitalization rates for asthma and LRIs, we found lower SES was significantly related to all outcomes except for asthma ED visit rates, which bordered on significance (Table
3). Other studies have associated asthma hospitalizations with various measures of lower SES, including minority race, household income, unemployment, adult education levels, and proportion of non-English language speaking persons [
14,
15,
47,
48]. We also found lower SES was associated with LRI hospitalization rates, as has been shown previously [
24,
49,
50]. Both findings suggest a lack of financial resources to obtain regular medical care or controlling medications, thus waiting till symptoms become so severe that the child will be hospitalized [
15,
51]. Another potential contributor to these relationships could be the association of parental and household smoking and increased risk of LRIs [
52]. Although tract-level data for smoking rates were not available for this study, it is known that lower SES [
53] and education levels are associated with smoking rates [
54]. In addition, minorities or those with limited English language skills may receive substandard care during and after hospitalization, potentially leading to repeat hospitalizations [
55]. Future analyses accounting for hospital readmissions may better elucidate these complex relationships. Nevertheless, our findings indicate that very young children living in census tracts with fewer financial or health resources to control asthma or LRIs are more likely to be hospitalized for these diseases.
Reduced population density was significantly related to all outcomes but asthma ED visit rates, suggesting that rural areas may have more severe cases of respiratory diseases that may result from parents waiting until the symptoms become so extreme that a hospital visit is necessitated. Pesek et al. [
56] found that children living in rural Arkansas have more undiagnosed and severe asthma symptoms compared to children in urban areas after accounting for other factors [
56]. In another study of very young children from a Medicaid cohort in Tennessee, rural participants were more likely to have asthma and visit the ED for asthma-related incidents and were less likely to use asthma medications (inhaled corticosteroids), compared to their urban counterparts [
57]. Valet et al. also found that rural children were more likely to have mothers who smoked [
57], which is associated with more severe respiratory symptoms [
58]. This contradicts other studies which found children in rural non-farm areas were less likely to develop asthma than those in urban settings [
59] or were no different than those in urban areas [
60]. However, these studies were not conducted with older children, whose lung function and development of asthma were likely already determined by unknown early life exposures [
28,
61]. Our findings suggest that urban areas have more health resources to control asthma and LRIs (e.g., more transportation options and quicker ambulance response times [
62]) before they become so severe they require hospitalization. Future studies should incorporate more information about these potential explanatory variables such as household transportation options, insurance status, and distance and effort to access regular health care to better understand the complex relationships between population density and respiratory diseases. Moving forward, telemedicine as a means to access care providers and specialists, has shown to be an effective means of reaching respiratory specialists [
63] and reducing asthma intensity in underserved communities [
64]. Another potential solution may be coordinating care among clinics, child care facilities, and caregivers to improve asthma outcomes [
65].
In addition to population density, county of residence was also related to respiratory disease outcomes. We found that living in Pima County was negatively associated with asthma and LRI hospitalization rates, even after accounting for differences in predictors by county (Table
2). Maricopa County had significantly more air pollution (Interquartile range [IQR] = −0.97–5.39, Coefficient of Variation [CV] = 236) and greater proportions of attached homes (IQR = 1.93%–42.4%, CV = 89.3) and household crowding (IQR = 0.70%–7.70%, CV = 130) and population density (IQR = 2,900–6,770 persons/sq. mile, CV = 62.9), while Pima County had significantly larger proportions of mobile homes (IQR = 0%–9.4%, CV = 190), home gas heating use (IQR = 47.8%–69.1%, CV = 27.8), and homes built before 1940 (IQR = 0%–1.90%, CV = 274). Other explanations of Maricopa County’s increased respiratory disease rates could be explained by a number of factors, such as higher pregnancy rates among females 18–19 years of age during the study period [
66]. Children of younger mothers are more likely to have wheezing LRIs than those born to older ones [
67], and children born to younger mothers are more susceptible to environmental factors, such as diesel traffic related air pollution [
68]. Another potential explanation could be that Maricopa County has ten times the proportion of land used for crop production compared to Pima County, which may lead to increased pesticide exposure in nearby residences, promoting childhood respiratory diseases [
69]. Further, Maricopa County has more industrial livestock operations, which have been linked to childhood asthma and other respiratory issues [
70,
71]. Another potential explanation could be that, in areas where employers are primarily agricultural, they may not offer insurance, leading to delays in seeking out care [
72].
Interestingly, increased air pollution was significantly related to all outcomes in the simple regressions but was only related to LRI ED visit rates when accounting for other factors. Darrow et al. [
73] also found this same relationship, specifically during abrupt increases in traffic-related air pollution due to meteorological changes. This could be a feasible explanation for findings in our study area, however one which we do not have the temporal resolution in air pollution data to address. In addition, because air pollution concentrations are based on emissions inventories, this may result in exposure misclassification, leading to underestimating the relationship of air pollution and respiratory disease outcomes. This relationship may increase in strength and significance, if exposure estimates have more spatial variability [
74]. It is also important to note that allowable concentrations of criteria air pollutants governed by the US National Ambient Air Quality Standards were reduced after our study [
75‐
78]. While this is beyond the scope of our project, it might further reduce air pollution’s significance as a predictor of respiratory disease when accounting for other factors. By accounting for lower levels of air pollutants in future studies, it might be possible to further elucidate these complex relationships among factors and outcomes.
For home characteristics, the proportion of home gas heating use was associated with asthma and LRI hospitalization rates, yet negatively associated with asthma ED visit rates. Other studies have found that exposure to gas combustion sources, whether for heating or cooking, have led to increased LRIs [
79‐
81]. However, our study also showed that the proportion of gas heating use was negatively related to asthma ED visit rates, running contradictory to associations with hospitalization rates for asthma and LRIs. A similar but not significant relationship was found with LRI ED visit rates and home gas heating use (IRR = 0.98; 95% CI = 0.95–1.02). Meanwhile, hospitalization outcomes were significantly related to the proportion of home gas heating. These contrasting relationships between ED and hospitalization rates and proportion of gas heating may result from 826 tracts having ED outcome data, compared to just 805 tracts for hospitalization outcomes. The 21 tracts without hospitalization data have significantly increased SES and reduced air pollution levels and proportions of household crowding and population density compared to the other 805 tracts (Wilcoxon rank-sum test;
p < 0.001 for all variables). This suggests that while home gas heating use relates to severity of respiratory diseases, it has a more nuanced relationship to care access in certain areas. This may be elucidated with future study into factors such as transportation options, insurance status, and availability of respiratory diseases specialists.
The proportions of mobile and attached homes were both associated with LRI ED visit rates. The relationship between LRI hospitalization rates and proportions of mobile homes may result from moisture build up from poorer ventilation [
82], however there are no recent studies examining these relationships. This could also suggest that residents in areas with high proportions of mobile homes (more common in rural areas) wait until LRI symptoms are so severe that the children require hospitalization. Also, the proportion of mobile homes was significantly related to LRI hospitalizations, while attached homes had a similar but insignificant relationship (IRR = 1.03; 95% CI = 0.99–1.08). This may suggest that residents of areas with greater proportion of attached homes (more common in more urban areas) also wait until symptoms are severe, but simply because they are closer to care, they take less time to reach the ED, and as a result, have less severe symptoms compared to those traveling a further distance in rural areas. Our results are similar to findings of increased health disparities for residents in rural areas with lower population density (increased proportion of mobile homes) and in old urban cores of Phoenix and Tucson with higher population densities (increased proportion of attached homes) [
83].
Similarly, homes built before 1940 were related to asthma ED visit rates, and the top 25% of tracts with high proportions of older homes were in rural areas and old urban cores of Phoenix and Tucson (Fig.
2). Again, this may indicate areas lacking care access either due to geographic proximity (rural areas) or lack of insurance (old urban cores). Questions of access may be answered in the future with more comprehensive GIS datasets of health care provider locations and transportation options. The proportion of homes with incomplete plumbing, likely an indicator of poor sanitation, was also related to LRI hospitalization rates. This same relationship between plumbing status and LRIs in young children has also been shown, albeit in a very different environment (i.e., Alaska) [
84]. Interestingly, household overcrowding, which Luijk et al. [
85] identified as a predictor for asthma and LRI symptoms, was related to all outcomes in the simple regressions but none in the multiple regressions. This may suggest that, while crowding is a potential factor, it is overtaken by others in our study area or may be indicative of other characteristics, such as parity [
86] or bed-sharing [
85].
Our study has several limitations, notably that this study was completed prior to the implementation of the Affordable Care Act, which may now alter relationships among geographic factors and outcomes due to a changed insurance landscape. In addition, the population in our study area grew from 2005 to 2009, which could increase the chances of model mis-specification. While asthma should not be diagnosed until after age 5 or 6 [
87], because of the high number of children in the transient wheeze phenotype before age 6 years who do not go on to develop asthma [
61,
88], we felt it important to assess factors for children <5 years because this may predict the expression of asthma and lung function in childhood and beyond [
28,
61]. Instead, asthma diagnoses before age 5 or 6 may reflect measures of care quality and disease severity (ED and hospitalization visits, respectively) [
89]. Despite these shortcomings, our paper has numerous strengths including the use of PCA to reduce many correlated predictor variables into unitless indices. This let us assess multiple known risk factors for childhood respiratory diseases in very young children for a large area (2 counties with 4.8 million residents). We also included spatial variables (population density and county), which helped to decrease the chance for residuals to exhibit spatial autocorrelation. As a result, our models meet the assumption of independent observations, while also accounting for natural spatial relationships among observations (i.e., census tracts) [
90]. Furthermore, we studied respiratory diseases in children under 5 years, who are more susceptible to these factors yet not studied with these outcomes and predictors at the neighborhood scale.