Increased healthcare resource utilization among Latino infants with acute respiratory illness compared with other racial/ethnic groups

Abstract

OBJECTIVE

To examine healthcare resource utilization for Latino infants with acute respiratory illness compared with other racial/ethnic groups.

STUDY DESIGN

We studied 674 term, previously healthy infants presenting for an unscheduled healthcare visit for acute respiratory illness. The predictor variable was infant race/ethnicity, and the primary outcome was healthcare resource utilization, adjusted for age and disease severity.

RESULTS

The cohort was 14% Latino, 52% white, 22% African-American, and 12% other races/ethnicities. Mothers of Latino infants were 37% Spanish-speaking. Bronchiolitis severity score was higher (indicating more severe disease) in white infants (median 6.0, interquartile range 3.0–9.0, on a 0–12 point scale) versus Latinos (3.0, 1.0–6.0) and African-Americans (3.5, 1.0–6.0), p<0.001 for the comparison of all groups. Disease severity was similar between Latino and African-American infants (p=0.96). Latino infants were the group most likely to receive antibiotics (58% of Latinos, 47% of whites, 34% of African-Americans, p=0.005) and have body fluid cultures drawn. Latino infants were also more likely than African-Americans to have chest x-rays and respiratory virus rapid antigen testing (p≤0.01). Latino infants from Spanish-speaking families, compared with those from English-speaking families, had increased receipt of RSV testing (76% versus 51%, p=0.016).

CONCLUSIONS

Providers caring for Latino infants with acute respiratory illness ordered more antibiotics and diagnostic testing for this group, particularly compared with African-Americans, who have similar disease severity and socioeconomic disparities. This suggests that the language barrier may be a potential explanation for observed differences.

Introduction

Latino children face a number of barriers to obtaining healthcare, including high rates of poverty, lack of insurance coverage, parental non-citizenship, and linguistic and cultural barriers.^1–3 There is limited evidence that, in order to compensate for the language barrier, providers utilize healthcare resources in the acute care setting at a higher rate in caring for Latino children compared with other groups. In a large (N=1847) survey of parents of Latino children with asthma, children from Spanish-speaking families were more than twice as likely to have been hospitalized for their asthma, despite having a lower prevalence of both asthma and measures of uncontrolled asthma.⁴ Latino children were found in one study that used administrative data to be more likely to present to the emergency department (ED) with bronchiolitis and be admitted for bronchiolitis; however, disease severity and possible explanations for this finding were not determined.⁵ A study of pediatric ED visits for fever, vomiting, or diarrhea showed that non-English speaking cases (of whom nearly all were Spanish-speaking) not seen by a language concordant doctor or with an interpreter were the group that had the highest testing costs and were most likely to be admitted. Importantly, the cost difference, but not increased admission rate, disappeared with the use of a professional interpreter.⁶

The above evidence suggests that in the acute care setting, more healthcare resources are used to care for Latino children versus other groups, at least partly to compensate for language barriers. However, whether there is increased healthcare resource utilization for Latino infants with acute respiratory illness (the most common reason for infant acute care and hospitalization), as well as potential explanations for any differences in utilization, is not known. To address this question, we studied Latino infants presenting with acute respiratory illness compared with infants of other races/ethnicities, comparing disease severity and interventions ordered by treating clinicians. We hypothesized that healthcare resources would be utilized at a higher level in caring for Latino infants and that this difference would not be explained by disease severity.

Patients and Methods

We studied infants enrolled in the Tennessee Children’s Respiratory Initiative (TCRI), which is a prospective longitudinal cohort study of the effect of viral and host factors during infancy on the development of asthma. A group of 7632 infants was screened to determine an eligible group of 2986 infants, of whom 674 infants along with their biological mothers were enrolled in TCRI. Infants were enrolled during four respiratory viral seasons (2004–2008) from the pediatric acute care clinic, pediatric emergency department, and inpatient pediatric ward of the Vanderbilt Children’s Hospital. Among the 2312 eligible infants who were not enrolled, 61% refused or did not have time for the visits, 20% had a conflict with another study in which they were participating, and 18% were ineligible for other reasons (principally that the mother was not present or there was a language barrier apart from Spanish). Enrolled versus non-enrolled infants did not differ by age or gender, and there was effective recruitment of the Latino population (Latinos were 15% of the enrolled cohort and 7% of the non-enrolled cohort). Detailed information about study design, objectives, and recruitment is contained in reference ⁷.

To be included in the study, infants had to be born at term and previously healthy, and had to have symptoms consistent with an acute respiratory illness, which included at least one of three major symptoms (wheeze, retraction, dyspnea) and at least two minor symptoms (fever, rhinorrhea, cough, otitis media, hoarse cry, vomiting after cough, respiratory syncytial virus [RSV] testing ordered by the treating clinician). Infants were excluded if their biological mother was not available or could not communicate in English or Spanish, or if the infant was over 1 year old, had a birth weight <2275 grams or gestational age <37 weeks, had a prior history of mechanical ventilation or of any major chronic health problems (children with gastroesophageal reflux, febrile seizure, otitis media, or prior wheezing were not excluded), or had received RSV immunoglobulin.

Charts were abstracted, and mothers of the infants underwent face-to-face interviews in English or Spanish according to subject preference, with Spanish-language interviews performed by a bilingual researcher (R.H.). Infant and family demographic information was ascertained, as was information about severity of respiratory illness and inpatient healthcare resource utilization.

Measures of severity of respiratory illness included calculation of a total bronchiolitis severity score, which ranged from 0 to 12 points, with a higher score indicating greater disease severity.^8–9 The total bronchiolitis severity score is a validated modified Tal score^8–9 that incorporates flaring, wheezing, oxygen saturation, and respiratory rate, as documented in the chart at the time of each infant’s initial clinical encounter. Because bronchiolitis severity score is a continuous variable describing the full spectrum of respiratory disease severity, it was applied to all infants (those with bronchiolitis as well as those with URI or other less common respiratory diagnoses).

Measures of healthcare resource utilization were determined from the chart, and included medication usage (antibiotics, oral or parenteral steroids, racemic epinephrine, albuterol), testing for respiratory illness (chest x-ray, rapid RSV antigen, rapid influenza antigen), and testing for other infections (cultures of blood, cerebrospinal fluid [CSF], and urine). Study investigators had no role in ordering any of these interventions, which were undertaken entirely at the discretion of the treating clinicians. This study was approved by the Institutional Review Board of Vanderbilt University, and parents provided written informed consent for both their and their child’s study participation.

Statistical analysis

Demographic features, bronchiolitis severity score, and measures of healthcare resource utilization were presented as frequency and proportion, median and interquartile range (IQR), or mean and standard deviation (SD) as appropriate, and were compared by infant race/ethnicity using the Kruskal-Wallis test for continuous measures and Chi-square test for proportions. The main independent variable was maternally reported infant race/ethnicity (white, African-American, Latino, or of other races/ethnicities). Dependent variables were health care resource utilization variables. To assess the adjusted association of infant race with healthcare resource utilization variables, we used separate multivariable logistic regression models including a priori selected covariates (infant age, gender, bronchiolitis severity score, and insurance type [as a measure of socioeconomic status]). Analyses were uncorrected for multiple comparisons because the dependent variables are typically correlated.¹⁰ Statistical analysis was performed using R version 2.12.1 (http://www.R-project.org).

Results

Of the 674 infants, 14% were Latino, 52% were white, 22% were African-American, and 12% were of other races/ethnicities. The proportion of Latino infants that was hospitalized (66%) was intermediate between the proportion of whites (79%) and African-Americans (46%); p for the overall comparison <0.001. The median age at presentation for Latino infants was 12.5 weeks (IQR 5.0, 26.2) and was intermediate between the median age for whites (11.0 weeks, IQR 6.0, 22.0) and African-Americans (20.0 weeks, IQR 8.0, 36.2). Thirty-seven percent of the mothers of Latino infants spoke Spanish as their primary language. Compared with other groups, Latino infants were more likely to be breastfed, less likely to be exposed to tobacco smoke, and less likely to be enrolled in childcare outside of the home, and their mothers were less likely to work outside the home and had a lower level of educational attainment. Latino infants had a lower prevalence of family history of atopic diseases compared with other groups. (Table 1)

Table 1.

Demographic characteristics of 674 term, previously healthy infants with respiratory symptoms, and their mothers, who were enrolled in the Tennessee Children’s Respiratory Initiative. Unless otherwise stated, data are presented as frequency (percentage).

	Latino, N=92	African-American, N=148	White, N=351	Other+, N=83	P value*
Infant characteristics
Male gender	55 (60)	84 (57)	191 (54)	50 (60)	0.68
Infant age at presentation in weeks, median (IQR)	12.5 (5.0,26.2)	20.0 (8.0,36.2)	11.0 (6.0,22.0)	19 (6.5,38.0)	<0.001
Infant’s birth weight in grams, mean ± SD	3413 ± 447	3224 ± 481	3363 ± 454	3343 ± 463	<0.001
Infant’s gestational age in weeks, mean ± SD	37.3 ± 10.0	38.8 ±4.2	38.4 ± 4.6	39.3 ± 1.5	<0.001
Infant was breastfed	69 (75)	63 (43)	196 (56)	58 (70)	<0.001
Infant has no other medical problems at presentationψ	73 (79)	116 (78)	274 (78)	70 (85)	0.55
Family and household characteristics
Mother works outside the home	12 (13)	58 (39)	132 (38)	32 (39)	<0.001
Infant enrolled in childcare outside the home	8 (9)	50 (34)	84 (24)	20 (24)	<0.001
Maternal education in years, mean ± SD, N=490**	9.2 ± 3.0	12.4 ± 1.6	13.4 ± 2.4	12.4 ± 2.9	<0.001
Maternal age in years, mean ± SD	26.3 ± 5.1	24.3 ± 5.3	26.8 ± 6.0	25.5 ± 4.8	<0.001
Maternal primary language					<0.001
English	55 (60)	147 (100)	342 (100)	68 (85)
Spanish	34 (37)	0 (0)	0 (0)	5 (6)
Other/unknown	3 (3)	0 (0)	0 (0)	7 (9)
Secondhand smoke exposure in the home or childcare environment	19 (21)	80 (54)	213 (62)	50 (61)	<0.001
Parent with a history of asthma	11 (12)	53 (37)	113 (33)	25 (31)	<0.001
Parent with a history of allergic rhinitis	22 (24)	49 (35)	185 (55)	28 (35)	<0.001

Abbreviations: SD=standard deviation, IQR=interquartile range

^*The Pearson Chi-square test was used for proportions and the Kruskal-Wallis test for continuous variables. Comparisons are among the 4 groups.

^**Maternal education was ascertained in only a limited subset of the cohort.

⁺Denotes infants of other races/ethnicities or of more than one race/ethnicity

^ψinfants were excluded from the study if they had prior episodes of mechanical ventilation, history of RSV immunoglobulin or palivizumab treatment, or serious medical problems (infants with a history of gastroesophageal reflux, febrile seizure, otitis media, or prior wheezing were not excluded).

Bronchiolitis severity

Latino and African-American infants had similar bronchiolitis severity, whereas white infants had more severe disease than these groups (Figure 1, panel A). The total bronchiolitis severity score, which ranges from 0–12, was highest in white infants (indicating more severe disease) at 6.0 (IQR 3.0–9.0) points. The median total score for Latino infants was 3.0 (1.0–6.0) points, and for African-American infants 3.5 (1.0–6.0) points (p<0.001 for the overall comparison; p=0.96 for the comparison of Latino versus African-American). Fifty-eight percent of Latinos and 56% of African-Americans were diagnosed with bronchiolitis versus 79% of whites; of the remaining infants, nearly all were diagnosed with an upper respiratory infection (34% of Latinos, 40% of African-Americans, and 15% of whites; p<0.001 for the overall comparison, and p=0.23 for the comparison of Latino versus African-American). Among the subset of Latino and African-American infants that was admitted, these groups did not differ in their length of stay or in their receipt of supplemental oxygen, ICU care, or mechanical ventilation (data not shown).

Figure 1.

Total bronchiolitis severity score (BSS) by race/ethnicity for all infants (panel A) and infants who were admitted (panel B). BSS is a composite score (including flaring, wheezing, O2 saturation, and respiratory rate) with up to 12 points possible, with a higher score indicating more severe bronchiolitis.

Measures of healthcare utilization

In spite of similar clinical characteristics compared with African-American infants, Latino infants had increased receipt of several types of interventions. First, among the subgroup of infants that was hospitalized, Latinos had a lower bronchiolitis severity score at initial clinical encounter than did infants of other races/ethnicities (for Latinos vs. African-Americans OR=0.49, 95%CI: 0.27, 0.90; for Latinos vs. whites OR=0.25, 95%CI: 0.15, 0.43; adjusted p value =<0.001, controlling for infant age, gender, and insurance type), suggesting a lower threshold for hospital admission (Figure 1, panel B). Latinos were more likely to receive antibiotics (Table 2), which occurred in 58% of Latino infants versus 34% of African-Americans (p<0.001 for the comparison of Latino versus African-American), even though the proportion of infants with fever prior to presentation (73% of Latinos, 67% of African-Americans, 70% of whites, p=0.74), and with otitis media (7% of Latinos, 13% of African-Americans, and 17% of whites, p=0.087) did not differ significantly by race/ethnicity. The proportions of Latinos and African-Americans who received medications (systemic corticosteroids, racemic epinephrine, albuterol) typically used to treat acute respiratory illness did not differ significantly, and Latinos were less likely than whites to receive racemic epinephrine or albuterol. However, Latinos were the group most likely to undergo cultures of blood, urine, and cerebrospinal fluid (p<0.01 for all of these variables for the comparison among all racial/ethnic groups). Although white infants were the group most likely to have chest x-rays and rapid testing for RSV and influenza ordered, providers were more likely to order these tests for Latino infants than for African-American infants (p≤0.01 in all cases, for the comparison of Latino versus African-American). Among those infants who had body fluid cultures obtained, there was no difference in the rate of culture positivity by race/ethnicity (p=0.46 for blood cultures, p=0.54 for urine cultures, and p=0.43 for cerebrospinal fluid cultures).

Table 2.

Healthcare resource utilization among infants with acute respiratory illness, by race/ethnicity. All data are presented as number (percentage).

	Latino, N=92	African-American, N=148	White, N=351	Other, N=83	P*, overall	P*, Latino vs. African-American	OR (95%CI), Latino vs. African-American§
Treatments	N (%)	N (%)	N (%)	N (%)	N (%)	N (%)
Infant received parenteral steroids	14 (18)	18 (13)	63 (18)	10 (12)	0.38	0.39
Infant received racemic epinephrine	12 (13)	15 (10)	81 (23)	8 (10)	<0.001	0.50
Infant received albuterol	24 (31)	48 (36)	187 (54)	27 (33)	<0.001	0.45
Infant received antibiotics	46 (58)	46 (34)	162 (47)	36 (44)	0.005	<0.001	2.69 (1.50–4.84)
Otitis media present	6 (7)	19 (13)	59 (17)	12 (14)	0.087
Diagnostic testing
Chest x-ray obtained	46 (50)	47 (32)	214 (63)	36 (44)	<0.001	0.006	2.36 (1.29–4.31)
Blood culture obtained	38 (41)	28 (19)	113 (32)	21 (25)	0.001	<0.001	2.58 (1.36–4.89)
Positives among those cultured (N=214)	1 (2)	0 (0)	7 (6)	1 (4)	0.46
Urine culture obtained	38 (41)	29 (20)	119 (34)	23 (28)	0.002	<0.001	2.41 (1.25–4.63)
Positives among those cultured (N=222)	2 (5)	0 (0)	7 (6)	1 (4)	0.54
Cerebrospinal fluid culture obtained	26 (28)	15 (10)	62 (18)	11 (13)	0.003	<0.001	2.61 (1.14–5.94)
Positives among those cultured (N=135)	0 (0)	0 (0)	3 (4)	0 (0)	0.43
Rapid RSV test obtained	57 (62)	60 (41)	253 (72)	47 (57)	<0.001	0.001	2.68 (1.37–5.27)
Positives among those tested (N=428)	30 (51)	37 (56)	171 (67)	25 (51)	0.023
Rapid influenza test obtained	17 (18)	11 (7)	71 (20)	20 (24)	0.002	0.01	2.51 (1.10–5.73)
Positives among those tested (N=143)	3 (14)	3 (16)	9 (11)	4 (17)	0.86

^*The Pearson Chi-square test was used to compare proportions

^§Odds ratios for the effect of child race/ethnicity on healthcare resource utilization, with African-American as referent, were calculated using a multivariable logistic regression models that adjusted for infant age, gender, bronchiolitis severity score, and insurance type.

Among Latino infants, although the number of subjects was small, a higher proportion of those with Spanish-speaking mothers received chest x-rays, cultures of blood, urine, and CSF, and rapid RSV and influenza antigen testing, compared with Latino infants with English-speaking mothers. These differences were only significant for RSV testing (76% of Latino infants with Spanish-speaking mothers versus 51% of Latino infants with English-speaking mothers, p=0.016); urine cultures were of borderline statistical significance (53% of Latino infants with Spanish-speaking mothers versus 33% of Latino infants with English-speaking mothers, p=0.059).

In multivariable regression analyses that adjusted for infant age, gender, bronchiolitis severity score, and insurance type (as a measure of socioeconomic status), compared with African-American infants as referent, Latino infants had increased odds of receipt of antibiotics, and increased odds of providers ordering chest x-rays, cultures of blood, cerebrospinal fluid, and urine, and rapid antigen testing for RSV and influenza (Table 2). Because fever may drive empiric antibiotic use, particularly in the younger infant, we additionally explored adding adjustment for fever to the multivariable model for receipt of antibiotics, and Latinos still had increased odds of receipt of antibiotics versus African-Americans (OR=2.56, 95%CI: 1.41, 4.66).

Discussion

Latino children encounter many socioeconomic disparities that reduce access to outpatient healthcare,^1–3 whereas limited data suggest that acute care resources are utilized at higher rates than for other groups, in order to compensate for language barriers.^4–6 This is one of few studies to examine differences by race/ethnicity in utilization of acute care resources for Latinos versus other groups, and one of few studies to focus on care of infants or young children, where the history is obtained entirely from the infant’s caregiver. We demonstrated that healthcare providers ordered more interventions for Latinos infants with acute respiratory illness than for other groups, including African-Americans.

The comparison of Latino infants with African-American infants is compelling, because children from both groups similarly encounter socioeconomic disparities and disparities in access to healthcare,¹¹ and in our study cohort, they have very similar disease severity. In spite of this, treating clinicians ordered more diagnostic tests (chest x-rays, body fluid cultures, rapid RSV and rapid influenza testing) and antibiotics for Latino infants than for African-Americans. Although chest x-rays and testing for respiratory viruses are germane to the care of infant acute respiratory illness (and are most commonly ordered for whites, who have greater respiratory disease severity than Latinos), cultures of body fluids are not. About 40% of the Latino infants had blood and urine cultures sent, and 28% underwent lumbar punctures, in a cohort that is largely out of the age range for empiric workup of fever without source.¹² The high proportion of antibiotic use in Latinos is not explained by the presence of fever prior to presentation, which did not differ by race/ethnicity, or by the rate at presentation of acute otitis media, which was lower in Latinos than other groups and is by far the most frequent comorbid bacterial infection seen with bronchiolitis.¹³ Furthermore, there was no difference by race/ethnicity in the proportion of body fluid cultures that were positive to support the appropriateness of increased antibiotic use or increased testing in Latinos. Given the similar clinical and socioeconomic features of the Latino and African-American infants (apart from infant age, and utilization differences persisted after adjusting for this), the language barrier is a plausible, if not likely explanation for why providers would order more medications and testing intended to treat or diagnose other conditions.

Among admitted infants, Latinos had a lower bronchiolitis severity score than other groups, including after adjusting for covariates (including infant age, which is an important factor in admission decisions). Although this cohort was a convenience sample and not population-based, and therefore does not allow comparison of admission rates by race/ethnicity, the lower severity score among admitted Latino infants compared with other admitted infants suggests a lower threshold for admission for Latino infants. Since many of the interventions that were more common in Latinos are generally obtained in the inpatient setting, a lower threshold for admission among Latinos, likely at least partially related to the language barrier, may explain some of the increased level of interventions being ordered for Latino infants.

For infants, being from a Spanish-speaking family is linked with several other demographic covariates (decreased income, lower rate of insurance and greater reliance on government insurance programs) that impair access to healthcare, with the result that Latino children from Spanish speaking families are less likely to have access to a usual source of care and after-hours advice, which may increase reliance on acute care.^1,4,14 Although this may result in an overall less ill population of infants brought in for care for acute respiratory illness, adjusting for disease severity does not attenuate the observed differences in utilization of healthcare resources. Once infants are receiving care, it is possible that observed differences could be explained by issues apart from language, such as different family expectations by race/ethnicity regarding amount of testing undertaken or need for admission, greater deference among Latinos versus other groups to physician recommendations for testing or other interventions,¹⁵ or a concern among providers that appropriate follow up is not available, leading to a lower threshold for admitting and/or ordering testing in the acute setting.

There are several limitations of this study that should be considered. First, it is not completely possible to exclude an alternate interpretation of the utilization data, that utilization for Latinos was appropriate, whereas it was low for African-Americans. However, the culture positivity data argues against this (essentially no children had positive body fluid cultures, regardless of race/ethnicity), and for RSV and influenza rapid antigen testing, Latinos and African-Americans had similar proportions of positive tests in spite of different frequencies of testing. If there were a higher rate of positivity in African-Americans, this would argue that the testing threshold was too low (i.e. only patients with a high likelihood of a positive test were selected, and some true positives were missed by not testing as many infants with lower anterior probability of a positive test). However, higher rates of test positivity were not seen in African-Americans, suggesting their rate of testing was appropriate.

A second limitation is that this cohort is not principally designed to address differences by language. Although there was a trend toward increased interventions among Latino infants with Spanish-speaking mothers compared with Latino infants with English-speaking mothers, the number of Latinos in the cohort was not large enough for an effective subgroup analysis. In addition to studying a larger group of Latinos, additional information that would strengthen the case that language affected our observed differences could include showing that a gradation in English fluency (e.g. no English versus limited English versus working in a job that requires speaking English) has a “dose response” effect on utilization. In general, however, Latinos in middle Tennessee are recent immigrants, and mothers rarely work outside the home, where they would regularly encounter spoken English.^16–17 Because our cohort recruited mother-infant dyads, mothers were the caregiver who provided information about all infants in our cohort, and so establishing a dose response for the effect of English fluency on healthcare utilization may prove difficult in this group. Although only 37% of mothers of Latino infants reported their principal language to be Spanish, the remainder likely had only partial fluency in English, and may have had difficulty with the technical language common to healthcare encounters. Related to this is the finding that Latino mothers had lower educational attainment compared with other groups in this study, which is linked with recent immigrant status and non-English speaking status. We performed analysis with further adjustment for language, and although this did attenuate the differences in utilization seen by race, there is a high correlation between language and race/ethnicity (data not shown).

In conclusion, treating clinicians were more likely to order hospitalization, diagnostic tests, and interventions for Latino infants with acute respiratory illness, particularly interventions intended to treat or diagnose other conditions. The differences in utilization are especially pronounced in comparison with African-Americans, who have similar disease severity and are likely to be socioeconomically similar apart from the language barrier, which may be an explanation for the observed differences. Future work should clarify the effect of language, perhaps by studying the effect of increasing levels of English fluency among caregivers, or the availability of language-concordant healthcare providers or professional interpreters, on healthcare utilization and costs.

Abbreviations

BSS

bronchiolitis severity score

CI

confidence interval

ED

emergency department

IQR

interquartile range

OR

odds ratio

SD

standard deviation

URI

upper respiratory infection

RSV

respiratory syncytial virus