Preadmission Diet and Zip Code Influences the Pediatric Critical Care Clinical Course for Infants with Severe Respiratory Illness ( N  = 187)

Abstract

We examined preadmission diet and zip code in infants with severe respiratory illness in the pediatric critical care unit. Patients aged 0 to 5 months admitted to the Helen DeVos Children's Hospital from January 2011 to May 2017 ( N  = 187), as exclusively formula, exclusively breastfed or mixed diet were included. Formula-fed infants ( n  = 88; 47%) clustered to zip codes with lower median incomes (<0.005), used public insurance as their payer type ( p  < 0.005), and were prescribed more ranitidine ( p  < 0.05) on admission.

Introduction

Critical illness is often the result of multiple factors that include baseline health (immunity) and environmental factors (diet, sociodemographics, secondary exposures) that contribute to a hospital admission. ^{1 2 3 4 5 6 7 8 9} In viral bronchiolitis, management is often focused exclusively on the severity of illness, so as to determine the appropriate respiratory support for the patient. The causative agent by itself is immaterial as the American Association of Pediatrics (AAP) does not recommend viral testing, which neither leads to prognostication nor direction of therapy. ¹ Considering the use of formula milk has been associated with several negative health consequences, ² it was essential to characterize outcomes and utilization of resources in the pediatric critical care unit (PCCU).

Viral bronchiolitis is a coryzal upper respiratory tract infection involving the lower (bronchiolar) respiratory tract, and progressive in nature, usually viral in origin. It is the leading cause for PCCU admissions in infancy, typically afflicting infants < 6 months of age. ³ Some infants with respiratory viral infections will develop progressive symptoms of the lower respiratory tract such as tachypnea, wheezing, severe cough, breathlessness, and respiratory distress, ³ as common to a variety of viruses causing bronchiolitis including respiratory syncytial virus (RSV). The Pediatric Health Information Systems data identified 25 deaths for every 10,000 admissions from RSV alone, which is a major public health concern worldwide. ⁴ No targeted therapies exist for viral bronchiolitis as the current AAP recommendations are limited to supportive care. There is, however, growing evidence that diet is a modifiable risk factor that can greatly influence respiratory disease development, ⁵ specifically in the first 4 months. ⁶

First, an infant's first food is formula milk, breast milk, or a mix of both. Human breast milk contains components that benefit human health even in the face of adversity: micronutrients, macronutrients, water, immune factors, stem cells, beneficial oligosaccharides, nutritionally essential lipids, and bioactive proteins. ⁷ Human breast milk composition is affected by environmental exposures that change human breast milk cytokine concentrations ^{8 9 10} and has been shown to decrease incidence of respiratory illnesses. ²

Second, sociodemographic factors affect both the initiation and likelihood of infants being breastfed during the first 6 months of life as previously reported. ^{11 12} Specifically, higher household income has been associated with an increased likelihood to breastfeed ¹³ and general feeding practices. ¹⁴ No studies to date have attempted to associate diet as an independent risk factor with PCCU clinical course. We wanted to explore this dietary relationship in a critically ill infant population with respiratory illness to address this knowledge gap. Our hypothesis was that formula-fed infants would have a worse clinical outcome compared with breastfed or mixed fed infants. Preillness diet is an important but poorly understood risk factor in need of further study.

Methods

Design and Sample

A retrospective chart review of admission profiles of patients from a high-volume, urban, tertiary care PCCU facility in Western Michigan was evaluated from the local electronic medical record (EMR) from January 2012 to May 2017. Patients included infants (0–5 months) of the PCCU, those fed enterally, admitted with a primary or secondary respiratory diagnosis based on International Classification of Diseases (ICD)-9/ICD-10 codes ( Table 1 ), together with any symptoms such as wheezing, severe cough, breathlessness, or respiratory distress. Patients were qualified as either exclusively formula fed, exclusively breastfed, or mixed diet. The exact ratio of formula to breast milk was not recorded in the EMR for the mixed population ( Supplementary Table S1 , available in the online version). Patients with the following criteria were excluded: prematurity less than 32 weeks' gestation, respiratory or immune comorbidities, G-tube fed infants, patients with preexisting gut motility disorders, and genetic syndromes. Control variables included: age, gender, race, median household income (from zip code), insurance payer type, level of respiratory support (minimum [room air, nasal cannula]; moderate [continuous positive airway pressure [CPAP], high-flow nasal cannula [HFNC], noninvasive positive pressure ventilation [NIPPV]]; or maximum [mechanical ventilation]), and types of medications prescribed. The three levels of respiratory support are also directly linked to the additional supportive care we provide to these patients in the PCCU. Factors such as the nurse to patient ratio, frequency of pulse oximetry checks, feeding protocol, blood gases, and pulmonary toileting are determined by the three levels of support. G*Power 3.1 was used to calculate the sample size. It was determined that 159 patients would be needed to achieve 0.80 power in a one-way analysis of variance (ANOVA) with three groups (formula fed, breastfed, and mixed diet), an α of 0.05 and medium effect size (0.25). Local Institutional Review Board (IRB) approval was acquired (2017-122-SH/HDVCH), and from an initial cohort of 305 patients, 187 were selected based on inclusion/exclusion criteria and dates of care ( Fig. 1 ). Quality control for data extraction of three independent reviewers was determined based on the interrater reliability (found to be 83% for 1,000 data points [∼5% of the total dataset]).

Table 1. Patient PCCU admission demographics of 0 to 5 months old from 2011 to 2017 ( N  = 187) with severe bronchiolitis .

Variable	Total ( N ) a	Formula ( n  = 88; 47%)	Breast milk ( n  = 55; 29.4%)	Mixed ( n  = 44; 23.5%)	p -Value
Age (mo)	187	88; 2.41 b  ± 1.28	55; 1.75 ± 1.34	44; 1.48 ± 1.20	<0.001
Gender	187
Male	119	56 (29.9)	32 (58.2)	31 (70.5)	0.450
Female	68	32 (17.1)	23 (12.3)	13 (7.0)
Race	187
White	136	59 (67.0)	45 (81.8)	32 (72.7)	0.08 c
Black	19	15 (17.0)	2 (3.6)	2 (4.5)
Hispanic	14	5 (5.7)	5 (9.1)	4 (9.1)
Other	18	9 (10.2)	3 (5.5)	6 (13.6)
Diagnosis	187
Bronchiolitis d	96	47 (53.4)	33 (60.0)	16 (36.4)	0.24 c
RSV e	72	32 (36.4)	27 (49.1)	13 (29.5)
Unspecified f	18	10 (11.4)	5 (9.1)	3 (6.8)
Viral other g	3	3 (3.4)	0 (0.0)	0 (0.0)
Pneumonia h	19	9 (10.2)	3 (5.5)	6 (13.6)
Viral i	8	5 (5.7)	1 (1.8)	2 (4.5)
Bacterial j	5	2 (2.3)	1 (1.8)	2 (4.5)
Viral and bacterial k	6	4 (4.5)	2 (3.6)	0 (0.0)
Respiratory failure l	38	17 (19.3)	10 (18.2)	11 (25.0)
Other m	29	11(12.5)	7 (12.7)	11 (25.0)
Vasopressor use	57	21 (24.4)	20 (37.7)	16 (39.0)	0.13
Respiratory support	180
Minimal	57	24 (28.6)	14 (25.9)	19 (45.2)	0.15
Moderate	49	25 (29.8)	18 (33.3)	6 (14.3)
Maximum	74	35 (41.7)	22 (40.7)	17 (40.5)
Insurance (public)	112	65 (73.9) b	24 (44.4)	23 (54.8)	<0.005
PRISM III	163	81; 3 (5)	47; 2 (7)	35; 4 (9)	0.136 c
Day 1 PELOD	163	81; 11 (1)	47; 10 (1)	35; 11 (2)	0.042 c
Median household income ($)	181	85; 43,031 b (14,250)	53; 49,428 (14,884)	43; 46,451 (21,438)	<0.005 c

Abbreviations: PCCU, pediatric critical care unit; PELOD, PEdiatric Logistic Organ Dysfunction-2 score; PRISM III, Pediatric Risk of Mortality score III.

^a Total may not add up to 187 due to missing data. Respiratory support: minimum (room air, nasal cannula), moderate (continuous positive airway pressure, high-pressure nasal cannula, noninvasive positive pressure ventilation, maximum (mechanical ventilation).

^b Pairwise comparison results denoting statistical significance, using a p -value of 0.017. Data are described as mean ± standard deviation if normally distributed, or median (interquartile range).

^cFisher's exact test was used to correct for unequal variances.

^dIncludes 465.9—Acute upper respiratory infections of unspecified site.

^eICD-9-CM 466.11 (which converts directly to) 2020 ICD-10-CM J21.0 Acute bronchiolitis due to respiratory syncytial virus, Diagnosis Code 079.6 Respiratory syncytial virus.

^f466.19—Acute bronchiolitis due to other infectious organisms; J21.8 Acute bronchiolitis due to other specified organisms; J21.9—Acute bronchiolitis, unspecified.

^gOther viral: J21.1—Acute bronchiolitis due to human metapneumovirus; J12.2—Parainfluenza virus pneumonia; 480.0—Pneumonia due to adenovirus.

^hJ18.9—Pneumonia, unspecified organism.

ⁱ Pneumonia viral/more than one viral: J14—Pneumonia due to Haemophilus influenzae & J21.0—Acute bronchiolitis due to respiratory syncytial virus; J12.2—Parainfluenza virus pneumonia & J14—Pneumonia due to Haemophilus ; 480.1—Pneumonia due to respiratory syncytial virus; 482.2—Pneumonia due to H. Influenzae ; 480.0—Pneumonia due to adenovirus.

^j Pneumonia bacterial: J15.211—Pneumonia due to methicillin-susceptible Staphylococcus aureus ; 481—Pneumococcal Pneumonia ( Streptococcus pneumoniae ); 482.1—Pneumonia due to Pseudomonas (Hcc); 482.9—Bacterial pneumonia, unspecified; 482.83—Pneumonia due to other gram-negative bacteria.

^k Mixed viral/bacterial: J15.211—Pneumonia due to methicillin-susceptible Staphylococcus aureus ; J15.4—Pneumonia due to other streptococci; J21.1—Acute bronchiolitis due to human metapneumovirus; J15.9—Unspecified bacterial pneumonia; J21.0—Acute bronchiolitis due to respiratory syncytial virus.

^lAcute respiratory failure/hypoxemia: J95.821—Acute postprocedural respiratory failure; J96.01—Acute respiratory failure with hypoxia; R09.02—Hypoxemia; 518.81—Acute respiratory failure (Hcc).

^mOther: J90—Pleural effusion, not elsewhere classified; J93.9—Pneumothorax, unspecified; J38.4—Edema of larynx; R06.89—Other abnormalities of breathing; J38.6—Stenosis of larynx; J98.4—Other disorders of lung; 464.4—Croup; 511.9—Unspecified pleural effusion; 519.8—Other diseases of respiratory system, not elsewhere classified; 518.82—Other pulmonary insufficiency, not elsewhere classified.

Fig. 1.

Study flow chart. LOS, length of stay; PCCU, pediatric critical care unit.

The Site

The study and data collection were conducted at Helen DeVos Children's Hospital located in Grand Rapids, Michigan, United States. Helen DeVos Children's Hospital PCCU has more than 1,500 admissions per year with more than 6,000 patient-days. Seventeen board-certified intensivists cover the 24-bed unit with flexibility and capability to care for up to 36 critically ill patients.

Variables

After IRB approval, basic demographic variables were pulled from the EMR for the initial 305 patients based on the dates specified. Dietary history was extracted from the dieticians' EMR, after confirming inclusion criteria were met. All data were populated using Research Electronic Data Capture (REDCap).

The independent variable of diet was listed as a categorical variable (formula fed = 1; breastfed = 2; mixed diet = 3). Additional categorical variables included respiratory support (high-frequency oscillator = 1; mechanical ventilator = 2; noninvasive ventilation = 3; NIPPV = 4; CPAP = 5; HFNC = 6; nasal cannula = 7; none/room air = 8; N/A = 9; grouped as described in design and sample) and prescribed medications as listed by dietician from nutrition reports ( Supplementary Table S2 , available in the online version). Dependent variables and primary end points included hospital length of stay (LOS), PCCU LOS, median household income (from zip code), readmissions, age, mortality, admission severity of illness scores (Pediatric Risk of Mortality score III [PRISM III], and PEdiatric Logistic Organ Dysfunction-2 [PELOD]), and patients who went onto extracorporeal membrane oxygenation (ECMO) were all numerical variables. The PRISM III score is a 17-point variable scoring system measured over the first 24 hours of admission. The PELOD score is an 18-point variable scoring system measured over the course of the first 10 days of admission. Binary variables included vasopressor use (no = 0; yes = 1), insurance type (public: no = 0; yes = 1), and dummy variables were used to measure gender (female = 0; male = 1), and race (white/Caucasian = 1; black/African American = 2; Hispanic = 3; other = 4). The only sociodemographic variables available through the EMR were median household income and insurance type. These serve as a proxy for sociodemographic factors for this patient population. ¹⁵

Analysis

Summary statistics were calculated for demographic and outcome data points. Normally distributed data were expressed as mean ± standard deviation, and nonnormally distributed data were expressed as median (interquartile range). Qualitative data were expressed as a frequency percentage. One-way ANOVA was used to look at differences among the three dietary groups with Welch's p -value correction for unequal variance. If normality assumptions were not met, Kruskal–Wallis' test was used. For significant p -values, pairwise comparisons were completed with Bonferroni correction. Categorical demographics and outcomes were compared using chi-square analysis. When individual cell counts were less than five, Fisher's exact test was used. Significance was assessed at the 0.05 level.

Results

Demographic Variables

Patients who met inclusion criteria were further reviewed as shown in Table 1 . In our population, formula-fed infants made up nearly half of the patients ( n  = 88; 47%), with one-third breastfed ( n  = 55; 29.4%) and mixed diet making up the last quarter of patients ( n  = 44; 23.5%). Gender, race, diagnosis, vasopressor use, and the need for respiratory support did not yield significant differences between dietary groups. Differences in age at admission revealed that formula-fed infants were older compared with both breastfed and mixed diet ( p ≤0.001) (2.4 ± 1.3 vs. 1.8 ± 1.3 and 1.5 ± 1.2 months, respectively). Significant differences were found in median household incomes (from zip code) ( p ≤0.005), whereby formula-fed infants on the average lived in zip codes with lower median household income. Severity of illness scores revealed only slightly significant differences on admission for the PELOD score ( p ≤0.042) versus the PRISM III score, which was not statistically different between groups ( p  = 0.135). The day 1 PELOD score was lower for breastfed infants. Insurance type also revealed significant differences ( p ≤0.005). The proportion of patients on public insurance was highest for formula-fed infants (73.9%), followed by mixed diet (54.8%) and breastfed fed infants (44.4%).

Clinical Outcomes

We examined differences in hospital and PCCU LOS. Hospital LOS (HLOS) (7.73 formula fed vs. 7.69 breastfed vs. 9.53 mixed diet) and PCCU LOS (5.96 formula fed vs. 3.52 breastfed vs. 5.04 mixed fed) were not statistically significant when comparing dietary groups. Additionally, there was no difference in mortality or use of ECMO ( Table 2 ). Readmissions to the PCCU within 4 months of the initial admission were highest in mixed diet infants.

Table 2. Patient outcomes by dietary groups ( N  = 187) .

Variable	Total ( N ) a	Formula ( n  = 88)	Breast milk ( n  = 55)	Mixed ( n  = 44)	p -Value
Hospital LOS (d)	187	7.73 (9.72)	7.69 (10.31)	9.53 (21.45)	0.30 b
Initial PCCU LOS (d)	187	5.72 (7.74)	3.31 (7.18)	4.00 (4.12)	0.50 b
Total PCCU LOS (d)	187	5.96 (8.07)	3.52 (7.18)	5.04 (9.48)	0.30 b
ECMO ( n ; %)	7	2 (2.3)	4 (7.5)	1 (2.4)	0.28 b
Readmits ( n ; %)	32	12 (14.3)	7 (13.5)	13 (30.2)	0.05
Deceased ( n ; %)	7	2 (2.3)	3 (5.5)	2 (4.5)	0.61 b

Abbreviations: ECMO, extracorporeal membrane oxygenation; LOS, length of stay; PCCU, pediatric critical care unit.

^aTotal may not add up to 187 due to missing data.

^bFisher's exact test was used to correct for unequal variances. Data are described as the median (interquartile range).

We examined medications prescribed as noted in the dietary assessment for the three dietary groups, including the seven most frequently listed ( Table 3 ). At admission, differences were found for ranitidine (H2-antagonist), ( p ≤0.05), which is frequently prescribed for patients while on steroids to avoid stomach ulcers. ¹⁶ This medication was more frequently prescribed in the formula-fed infants than breastfed or mixed diet infants (27.3 vs. 10.9 vs. 20.5%). We did note, however, that almost half of the formula-fed infants (10/24; 41.7%) were on ranitidine prior to admission, as compared with none of the breastfed (0/6) and one-third of mixed diet infants (3/9). In total, 22/39 (56%) (15 formula-fed; 1 breastfed; 6 mixed diet) patients were on ranitidine and receiving steroids during their admission. In the absence of steroids, 15/39 (38.5%) (8 formula-fed; 5 breastfed; 2 mixed diet) were still receiving ranitidine, and a small minority of 2/39 (5.1%) (1 formula fed; 1 mixed diet) did not obtain ranitidine despite steroid use (data not shown).

Table 3. Medications prescribed in nutrition evaluation on admission ( N  = 187) .

Medication	Formula ( n  = 88) n (%)	Breast milk ( n  = 55) n (%)	Mixed ( n  = 44) n (%)	p -Value
Fentanyl	30 (34.1)	12 (21.8)	9 (20.5)	0.14
Furosemide	17 (19.3)	8 (14.5)	13 (29.5)	0.17
Midazolam	27 (30.7)	12 (21.8)	9 (20.5)	0.33
Milrinone	9 (10.2)	5 (9.1)	4 (9.1)	0.97
Morphine	5 (5.7)	2 (3.6)	5 (11.4)	0.29 a
Ranitidine	24 (27.3)	6 (10.9)	9 (20.5)	0.05
Vecuronium	11 (12.5)	1 (1.8)	5 (11.4)	0.08

^aFisher's exact test was used to correct for unequal variances.

Discussion

This study suggests that preadmissions characteristics such as diet and median household income may influence PCCU outcomes. We did note differences in day 1 PELOD scores which may be related to mechanical ventilation being 1 of the 18 items considered in generating the score. ¹⁷ This may make PELOD scores a better indicator of severity in these patients compared with PRISM III which is focused more on laboratory values (that the AAP discourages), and thus less likely to be available on admission (such as white blood cell counts and serum chemistries). ¹⁸

Breastfed infants in our study were nearly 3 weeks younger than formula-fed infants. This is in contrast to a previous report whereby exclusively breastfed infants hospitalized for acute respiratory illness were older than infants never breastfed. ¹⁹ Studies have shown that the resting energy expenditure is the same for both formula-fed or breastfed infants; however, the length of time taken to complete a feed for breastfed infants is longer. ²⁰ In younger breastfed infants with a viral illness, this difference in length of time to feed could create challenges and risk for undernourishment and fatigue, resulting in the need for PCCU care sooner. One study showed that breastfeeding disruption is an underappreciated frequent event during admissions for viral bronchiolitis that affects younger infants more often. ²¹ Although speculative, recent evidence has shown that breast milk upregulates immune-related genes such as chemokine genes, perhaps contributing to a proinflammatory phenotype in the gut. ²²

Median household income was ∼$6,000 (15%) less for formula-fed infants than breastfed infants, with mixed diet infants showing $3,000 (6.5%) less than breastfed. This financial difference may be a substantial amount for a family. Feeding practices have previously been associated with higher median income homes, ^{12 23} and our results support these findings. Additional contributing sociodemographic factors such as maternal age, social class, and education, as well as marital and smoking status have previously been shown to play a role, ¹² however, were out of scope for this report, and could be included for future studies.

We found that a higher number of formula-fed infants used public insurance. Parents without private insurance may delay obtaining medical attention and/or there may be other issues around medical access. ²⁴ Additional lifestyle factors know to influence respiratory illness, however, not captured in our dataset, include smokers, ²⁵ or over-crowding in the home. ²⁶

We found one-third of the mixed diet infants were readmitted to the PCCU within 4 months of their initial admission. It was previously believed that any amount of breast milk inferred a protective effect for infants. ¹² This is the first report to describe this finding. Fortification of expressed breast milk with formula may be recommended as a result of low-quality or quantity breast milk, ²⁷ short feeding duration and lack of hindmilk intake, ²⁸ or the infant may be deemed a poor feeder. Together these factors may result in suboptimal oral intake. These readmissions may be due, in part, to preexisting issues and will be described in a separate report. Given the small sample size, little direct correlation can be drawn.

Ranitidine is commonly used for acid reflux and prescribed in the case of steroid use. ^{16 29} There are no previous reports that we are aware of, where ranitidine use is higher in formula-fed infants as admitted to the PCCU for acute respiratory illness. A report from the late 1980s found that in infants who were able to consume breast milk during steroid use (at 90 mL/kg), ranitidine use was unnecessary (given the buffering capacity and protective effects by epidermal growth factors found in breast milk). ³⁰ Ranitidine is used to prevent gastrointestinal bleeding and stress ulcers and is frequently prescribed in ventilated patients who are nil per os or have lower gastric pH. ³¹ In our population irrespective of diet or steroid use, ranitidine was prescribed in 37/39 (95%) of ventilated patients. Its use has recently been reported in the neonatal population and associated with a predisposition to necrotizing enterocolitis ³² ; the subject warrants further exploration.

It has been reported that formula-fed infants are three times more likely to require hospitalization in the first 4 months of life for severe respiratory tract illnesses when compared with breastfed infants. ⁶ In a prospective multicenter study of more than 1,800 infants at 12 months, infants were found to have a higher risk of hospitalization for bronchiolitis in the exclusively formula-fed group (hazard ratio: 1.57; 95% confidence interval [CI]: 1.00–2.48). ²² These results are consistent with a large mother–infant cohort study ( n  = 629) that reported a lower odds ratio (OR) for breastfed infants with lower respiratory tract infections (OR: 0.64; 95% CI: 0.42–0.99). ¹⁹ This may be, in part, because human milk composition varies and is tailored to support the infant's growth and development. The composition changes in response to infections and environmental exposures ³³ and is customized in the event of illness. ²¹ However, as stated earlier, the ability to support exclusive breastfeeding has been linked to women of higher socioeconomic demographics. ^{10 13} In the currently presented study, however, we did not find an association between diet and HLOS or PCCU LOS, which may be perceived as an overall beneficial finding for the patients.

In summary, formula-fed patients were older at PCCU admission, from lower income households, more likely to use public insurance and prescribed more medication for acid reflux. Mixed diet populations showed intermediate profiles, with the exception of greater PCCU readmissions. Diet was not directly linked with LOS or respiratory support in our cohort.

The authors did not have access to additional sociodemographic information, such as parent's education level, which would require the parents' consent. To address the study limitation of having multiple data collectors (which may have introduced bias), we evaluated the interrater reliability for all, which showed good agreement. There was a lack of serologically confirmed viral infections. A detailed dietary recall from mothers was not obtained, and the data presented are from a single study site. Room air was grouped with nasal cannula minimal oxygen delivery; however, this is an overestimation in severity and may have contributed to additional study heterogeneity. Information on the mixed feeding population was not determined a priori, and therefore, limited information was available to qualify exact amounts of each diet (formula vs. breast milk) consumed ( Supplementary Table S1 , available in the online version).

Conclusion

Admission profiles revealed that infants from a lower sociodemographic background, based on median household income, were more likely to be formula fed, require maximum respiratory support, and have public insurance as payer type. This study indicated that preadmission factors such as diet (indirectly) and sociodemographic factors (directly) play a role in the clinical course of infants with severe respiratory diseases such as bronchiolitis. Children of lower socioeconomic strata are known to present to the emergency departments more frequently than the general population, ^{34 35 36} perhaps indicative of the lack of a general practitioner, warranting the implementation of care pathways in local communities. ¹⁵ A large multisite study adequately powered to confirm the link between independent risk factors of preadmission diet, sociodemographics, and hospital course is warranted.

Funding Statement

Funding M.L. and S.R. are recipients of the Spectrum Health Office of Research grant (#RI0027-19), researching dietary influences in the infant population.