Risk Factors Associated With Escalation of Therapy in Hospitalized Children With Asthma

Abstract

OBJECTIVE:

Children hospitalized for asthma exacerbations may require escalation of therapy (EOT), but risk factors remain poorly defined. We aimed to identify risk factors for EOT to inform interventions and improve care.

PATIENTS AND METHODS:

We conducted a retrospective cohort study using the Pediatric Health Information System of hospitalized children aged 4 to 18 years with a diagnosis of an asthma exacerbation from January 1, 2018, to December 31, 2023, across 45 children’s hospitals. EOT was defined as intensive care unit transfer, advanced respiratory support, or adjunctive asthma therapy (ie, magnesium sulfate, methylprednisolone) after the first hospital day. Mixed-effects regression estimated the association between EOT and risk factors: age, sex, race and ethnicity, insurance, Child Opportunity Index (COI), complex chronic conditions (CCC), and season.

RESULTS:

Among 92 224 encounters, 15 989 (17.34%) required EOT. Adjusted analysis showed lower odds of EOT in male patients (adjusted odds ratio [aOR], 0.91 [CI, 0.88, 0.94]) and spring compared with summer admissions (aOR, 0.93 [0.88, 0.99]). Compared with early childhood (<5 years), the odds of EOT were higher in late childhood (5–12 years) (aOR, 1.19 [1.13, 1.26]), adolescence (13–17 years) (aOR, 1.48 [1.38, 1.59]), and adults (≥18 years) (aOR, 1.43 [1.17, 1.73]). Higher odds were observed in Hispanic (aOR, 1.07 [1.00, 1.14]) and multiracial (aOR, 1.17 [1.02, 1.34]) patients and those with CCC (aOR, 2.07 [1.95, 2.19]). COI and insurance were not significantly associated with EOT.

CONCLUSION:

We identified key risk factors for EOT in children hospitalized for asthma exacerbations. Such findings can help with early risk stratification or serve as proxies for modifiable features of inpatient care. Further research is warranted to explore these associations to improve care.

INTRODUCTION

Asthma is a leading cause of pediatric hospitalizations in the United States and disproportionately impacts Black and Hispanic children and those living in high-poverty neighborhoods.^1–5 Although hospitalization rates have declined across all racial and ethnic groups over time, disparities persist for these populations.^2,6 Such differential outcomes often reflect structural racism and unequal access to resources rather than biological variation. Both individual risk factors, such as age, sex, and comorbidities, as well as broader structural inequities, including limited access to care, and neighborhood disadvantage affect asthma health outcomes.^7–13 Understanding how these intersecting risk factors contribute to the hospital course is essential to identifying children at highest risk for adverse outcomes and developing equitable and timely interventions.

Escalation of therapy (EOT) refers to an increase in intensity or complexity of medical management initiated in response to clinical deterioration or inadequate therapeutic response during a patient’s hospitalization.¹⁴ EOT, including intensive care unit (ICU) transfer and the use of advanced respiratory support or adjunctive therapies (methylprednisolone, theophylline, etc), is an indication of clinical severity, system level processes, and clinician variation.^15,16 Factors such as illness severity, treatment response, failure of early recognition of disease progression, quality of care, and resource availability influence the need for EOT in a hospitalized child.^15,16 Across pediatric populations, ICU transfer or rapid response team (RRT) activations as types of EOT were associated with longer lengths of stay, higher cost, and greater risk of mortality.^17–19 Those who require EOT from the general unit to the ICU have worse outcomes then those admitted directly to the ICU.^17,20,21 Understanding these escalation events in the context of pediatric asthma is critical to avoiding subsequent morbidity and unnecessary health care costs, given the disproportionate hospital utilization associated with severe asthma exacerbations.

ICU transfer alone is an incomplete proxy of EOT in asthma, as it may reflect prehospital illness severity. Additionally, criteria for ICU transfer are variable and dependent on institutional policies and resources, such that many children receive higher levels of care outside of ICU settings.^4,22 Thus, a broader definition of EOT, encompassing ICU transfer and additional therapies (ie, respiratory support, adjunctive medications), regardless of being in an ICU setting, is needed. Such a definition of EOT captures clinical deterioration and treatment intensity during hospitalization, providing a complementary and potentially more sensitive measure of inpatient severity. Examining risk factors for EOT as a composite measure may help identify potential variations in care and identify patients at higher risk of adverse outcomes. This knowledge can guide health systems in developing targeted interventions, reduce unwarranted variation in practice, and ultimately improve outcomes for children with asthma.

To date, studies assessing EOT as a composite measure for hospitalized children with asthma exacerbations examined limited risk factors and were only conducted outside of the United States. For example, a study in New Zealand and Australia found that those with EOT had longer length of stay.⁴ However, this study examined only age, sex, and triage category as individual risk factors and did not include known social and environmental factors that influence asthma outcomes. To address this gap, we examined risk factors for EOT among children hospitalized with acute asthma exacerbation in a large US cohort. We hypothesized that younger age, historically marginalized racial and ethnic groups, those with public insurance, and those living in lower-resource neighborhoods would have higher odds of EOT. By clarifying risk factors of EOT, our findings aim to inform risk stratification and highlight opportunities for more equitable inpatient care.

METHODS

Study Design, Participants, and Setting

This is a multicenter retrospective cohort study using the Pediatric Health Information System (PHIS) database of children aged 4 to 18 years admitted to non-ICU settings for asthma exacerbations from January 1, 2018, to December 31, 2023, across 45 children’s hospitals. PHIS is an administrative database containing deidentified clinical and resource utilization data for patient encounters from tertiary pediatric hospitals in the United States that are part of the Children’s Hospital Association (Lenexa, Kansas).²³ Participating hospitals provide discharge and encounter data, including demographic, diagnosis, procedure codes, and resource utilization (eg, pharmaceuticals, imaging, and laboratory). Data quality and reliability are assured through a joint effort between the Children’s Hospital Association and participating hospitals.

Patients were identified using any visit diagnosis code for asthma exacerbation or status asthmaticus using the International Classification of Diseases, Tenth Revision (ICD-10) (J45. 21, J4522, J4531, J4532, J4541, J4542, J4551, J4552, J45902, J45990).^24,25 We excluded those with All Patient Refined Diagnosis Groups (APR-DRG) codes for cystic fibrosis, pulmonary embolism, and trauma, as well as principal diagnosis codes indicative of disease processes other than asthma (ie, malignant neoplasms, lung transplant, bronchiectasis, acute postprocedure respiratory distress). We excluded encounters without an emergency department charge code, which suggests transfer from another institution. PHIS collects data based on daily billing codes and does not provide timestamps for orders, so the sequence in which care was provided on a given day cannot be determined. Those with ICU charges on the first day of the admission (denoted “day of service 0” or “DOS 0”) were excluded, as we could not distinguish those transferred to the ICU due to EOT from those who were admitted directly to the ICU. Those who received respiratory support or adjunctive therapies (ie, magnesium sulfate, epinephrine, etc) on DOS 0 but were not admitted to the ICU were included in our study population because these therapies may have been used for initial stabilization. This study was deemed non–human participant research by the institutional review board.

Study Outcome and Risk Factors

Our primary outcome measure was EOT, defined as transfer to the ICU, need for respiratory support (high-flow nasal cannula, noninvasive ventilation, continuous or bilevel positive airway pressure, or adjunctive asthma therapy (epinephrine, terbutaline, aminophylline, theophylline, Heliox, magnesium sulfate, or methylprednisolone)^22,26,27 after DOS 0.

Individual and environmental risk factors were selected a priori based on prior literature highlighting their association with asthma control or severity, hospitalizations, or ICU admissions and included the following: age, sex, at least 1 complex chronic condition (CCC), insurance status, season, race and ethnicity, and the Child Opportunity Index (COI) 3.0.^28–31 Age at admission was categorized using American Academy of Pediatrics developmental stages: early childhood (1 to <5 years), late childhood (5 to <13 years), adolescent (13 to <18 years), and adult (≥18 years). CCC were identified using the CCC 2.0 system, which defines medical complexity based on chronic conditions involving multiple or severely affected organ systems requiring specialty care (ie, technology dependence, cystic fibrosis).³² Asthma is not included in this classification. Insurance categories included commercial, government, and other (including other payor and self-pay) (Supplemental Table 1). Seasons were defined based on months (winter: December to February; spring: March to May; summer: June to August; and fall: September to November). Reference groups were determined a priori.

The COI 3.0 is a composite indicator capturing neighborhood-level social and environmental risk factors across 3 domains: education, health and environment, and social and economic.^33,34 The COI composite overall score is categorized in to population-weighted quintiles: very low (1 to 20 percentile), low (21 to 40), moderate (41 to 60), high (61 to 80), and very high (81 to 100).³⁴ A low COI indicates an area of low opportunity for health development and outcomes. The nationally normalized COI 3.0 level is linked within PHIS to the zip code associated with each child per encounter. Those with missing COI scores were excluded from our cohort.

Given that COI may not fully capture the systemic and structural conditions that underlie racial disparities in health outcomes, we included race and ethnicity as a social construct rather than a biological variable to serve as a proxy for interpersonal and structural discrimination at the individual level. Race and ethnicity in PHIS is extracted from electronic health records where assignments are presumed to be based on self-report, although this may vary by institution and attainment is uncertain. Race and ethnicity is categorized as non-Hispanic white, non-Hispanic Black, Asian, American Indian/Alaskan Native, Pacific Islander, Hispanic, multiracial, and other specified. Categories are in alignment with the American Thoracic Society and the National Academy of Sciences, Engineering, and Medicine and the 2023 US Office of Management and Budget standards recommendations for the responsible use of race and ethnicity in research.^35–37

Statistical Analysis

Patient characteristics were summarized using standard statistical summaries using frequency (percent) for categorical variables. Mixed-effects logistic regression models were used to estimate the association between our primary outcome and exposures (age, sex, race and ethnicity, CCC, insurance, COI, and season). We accounted for clustering using unique patient and hospital identifiers as random effects. Variance inflation factors were calculated and all covariates demonstrated values less than 2, indicating low collinearity. Adjusted odds ratios (aORs) and 95% CI were used to summarize the association of each risk factor with EOT and a 2-tailed alpha of 0.05 was used to determine significance. All statistical analyses were performed using Stata 18 version (StataCorp LLC, College Station, TX).

RESULTS

Study Characteristics

We extracted a total of 131 284 encounters. We excluded 20 929 (16%) with no emergency department charge, 118 (0.09%) with a missing COI, 184 (0.14%) with APR-DRG codes for cystic fibrosis, pulmonary embolism, and trauma, 201 (0.15%) with a principal diagnosis not suggestive of an asthma admission, and 17 631 (13.43%) who were admitted to the ICU on DOS 0. Our final cohort consisted of 72 968 patients with 92 224 encounters, of which 15 989 (17.34%) required EOT (Table 1). Most children were aged 5 to 12 years (68.2%) and 58.4% were male. There were 43.9% who identified as non-Hispanic Black, 25.27% non-Hispanic white, and 22.4% Hispanic. Most had no CCC (90.7%), had government insurance (65.2%), lived in very low- or low-COI areas (35.1% and 19.4%, respectively), and were admitted in the fall (34.8%).

TABLE 1.

Demographic Characteristics of Pediatric Patients Admitted With Acute Asthma Exacerbation Requiring Escalation of Therapy in 45 Pediatric Hospitals

Characteristics	Total Sample (N = 92 224)	No Escalation (N = 76 235)	Escalation (N = 15 989)
AAP age categoriesa
Early childhood (<5 y)	16 608 (18.0)	14093 (18.5)	2515 (15.7)
Late childhood (5 y to <13 y)	62 897 (68.2)	52 079 (68.3)	10 818 (67.7)
Adolescence (13 y to <18 y)	11 905 (12.9)	9420 (12.4)	2485 (15.5)
Adult (≥18 y)	814 (0.9)	643 (0.8)	171 (1.1)
Sexb
Male	53 824 (58.4)	44 845 (58.8)	8979 (56.2)
Female	378 395 (41.6)	31 386 (41.2)	7009 (43.8)
Race and ethnicityb
Asian	2352 (2.6)	1984 (2.6)	368 (2.3)
Hispanic	20 382 (22.4)	16 693 (22.2)	3689 (23.4)
Multiracial	1948 (2.1)	1571 (2.1)	377 (2.4)
Non-Hispanic Black	39 936 (43.9)	33 063 (43.9)	6873 (43.6)
Non-Hispanic white	23 012 (25.3)	19 152 (25.4)	3860 (24.5)
American Indian	176 (0.2)	140 (0.2)	36 (0.2)
Pacific Islander	196 (0.2)	168 (0.2)	28 (0.2)
Other specified	3067 (3.4)	2514 (3.3)	551(3.5)
Insuranceb
Commercial	26 314 (28.6)	21 761 (28.6)	4553 (28.6)
Government: Medicaid/other gov	59 890 (65.2)	49 380 (65.0)	10 510 (66.0)
Other (other payor/self-pay)	5671 (6.2)	4800 (6.3)	871 (5.5)
Complex chronic conditions
No complex chronic condition	83 650 (90.7)	70 029 (91.8)	13 621 (85.2)
≥1 complex chronic condition	8574 (9.3)	6206 (8.1)	2368 (14.8)
Child Opportunity Index 3.0
Very low	32 392 (35.1)	26 751 (35.1)	5641 (35.7)
Low	17 881 (19.4)	14 798 (19.4)	3083 (19.2)
Moderate	15 487 (16.8)	12 797 (16.8)	2690 (16.7)
High	14 509 (15.7)	12 045 (15.8)	2646 (15.4)
Very high	11 955 (13.0)	9844 (12.9)	2111 (13.2)
Season
Winter	22 908 (24.8)	19 044 (25.0)	3864 (24.2)
Spring	22 777 (24.7)	19 011 (24.9)	3766 (23.6)
Summer	14 475 (15.7)	1 1 926 (15.6)	2549 (15.9)
Fall	32 064 (34.8)	26 254 (34.4)	5810 (36.3)

Abbreviation: AAP, American Academy of Pediatrics.

Reported as n (%) unless otherwise stated.

^aAge range for the study is 4 to 18 years. The “early childhood” group contains those who are aged 4 years only and the “adult” category contains those who are aged 18 years only.

^bMissing data: race and ethnicity (n = 1157; 1.25%), sex (n = 5, 0.01%), and insurance (n = 349; 0.38%).

Multivariable Regression

Table 2 illustrates results from our multivariable mixed-effects regression model. We observed higher odds of EOT for older children. Compared with children in early childhood, those in late childhood had 19% higher odds (CI, 1,13, 1.26) of EOT, adolescents had 48% higher odds (CI, 1.38, 1.59), and young adults had 43% higher odds (CI, 1.17, 1.73). Male compared with female children had lower odds of EOT (aOR, 0.91 [CI, 0.88, 0.94]). We observed higher odds of EOT for Hispanic (aOR, 1.07 [1.00, 1.14]) or multiracial (aOR, 1.17 [1.02, 1.34]) patients compared with non-Hispanic white and for those with a CCC (aOR, 2.07 [1.95, 2.19]). Compared with summer, spring admissions had lower odds (aOR, 0.93 [0.88, 0.99]) and fall admissions had higher odds (aOR, 1.06 [1.00, 1.12]) of EOT. No significant associations were found between EOT and COI and insurance.

TABLE 2.

Mixed-Effects Logistic Regression Model of Escalation of Therapy in Pediatric Patients Admitted With Acute Asthma Exacerbation (N = 90 722 Encounters; 71 730 Patients)

Variable	Adjusted Odds Ratio	95% CI	P Value
AAP age categoriesa
Early childhood (<5 y)	Ref
Late childhood (5 y to <13 y)	1.19	1.13, 1.26	<.001b
Adolescence (13 y to <18 y)	1.48	1.38, 1.59	<.001b
Young adult (≥18 y)	1.43	1.17, 1.73	<.001b
Sex
Female	Ref
Male	0.91	0.88, 0.94	<.001b
Race and ethnicity
Non-Hispanic white	Ref
Asian	0.92	0.81, 1.05	.22
Hispanic	1.07	1.00, 1.14	.04b
Multiracial	1.17	1.02, 1.34	.02b
Non-Hispanic Black	1.03	0.97, 1.08	.34
American Indian	1.25	0.83, 1.89	.29
Pacific Islander	1.00	0.64, 1.58	.98
Other specified	1.00	0.89, 1.13	.91
Complex chronic conditions
No complex chronic conditions	Ref
≥1 Complex chronic conditions	2.07	1.95, 2.19	<.001b
Insurance
Commercial	Ref
Government insurance	1.03	0.99, 1.07	.22
Other (other payor/self-pay)	1.00	0.91, 1.10	.95
Child Opportunity Index
Very low	1.00	0.94, 1.07	.90
Low	1.03	0.96, 1.11	.37
Moderate	0.99	0.93, 1.07	.91
High	0.97	0.90, 1.05	.44
Very high	Ref
Season
Summer	Ref
Fall	1.06	1.00, 1.12	.04b
Spring	0.93	0.88, 0.99	.02b
Winter	0.97	0.92, 1.04	.40

Abbreviation: AAP, American Academy of Pediatrics.

Unique patient and hospital identifiers used as random effects.

^aAge range for the study is 4 to 18 years. Early childhood category contains those who are aged 4 years and adult category contains those who are aged 18 years only.

^bStatistically significant.

DISCUSSION

In this large retrospective analysis, we identified key risk factors associated with EOT for children hospitalized with asthma exacerbations. Children who were older, were female, were Hispanic or multi-racial, had a CCC, or were admitted in the fall had increased odds of EOT. Interestingly, odds of EOT were lower in spring admissions compared with summer, and there was no association between EOT and COI or insurance status. To our knowledge, this is the first study in the United States to assess individual and environmental risk factors for EOT in pediatric asthma hospitalizations. Our inclusive definition of EOT provides a more sensitive measure of inpatient severity that may better reflect variation in care and opportunities for earlier recognition and intervention.

Our study builds on prior research by identifying important individual factors associated with EOT. To our knowledge, only Craig et al assessed EOT as a comprehensive measure in pediatric patients with asthma and similarly reported an association between EOT and older age.⁴ Their study also demonstrated that the type of EOT varied by age, with older children more likely to receive parenteral treatment.⁴ One plausible explanation is that older children may have higher rates of allergen exposure and sensitization leading to higher illness severity and clinical deterioration.^8,38 By contrast, studies assessing ICU transfer or RRT activations only have not identified significant age associations.^19,39 We also found that female children had higher odds of EOT compared with male children, a finding not observed by Craig et al. Although asthma is more prevalent and results in more hospitalizations in male children,⁴⁰ prevalence and morbidity shift in adolescence, rising in female and declining in male children.⁴¹ Adolescent female individuals also experience more severe asthma, with higher rates of underdiagnosis and undertreatment.^42,43 These findings underscore the importance of age- and sex-specific considerations in inpatient asthma care. Awareness that older children and adolescent female individuals may be at higher risk for EOT can guide earlier risk stratification, targeted monitoring, anticipatory management, and caregiver discussions. Incorporating such insights into decision-making may optimize resource allocation and improve outcomes.

Our study further evaluates additional individual (CCC, race and ethnicity, insurance) and environmental (COI, season) risk factors for EOT among children hospitalized with asthma. Children with CCC have been reported to account for up to 38% of hospitalized children in tertiary hospitals and are growing in number.^44,45 Additionally, children with CCC have longer LOS and higher resource utilization.^46,47 Although children with CCC account for a small percentage of our study population, they had significantly higher odds of EOT. Although not previously reported in studies of comprehensive EOT measures, similar findings were seen in studies assessing acute clinical deterioration events, not limited to asthma.^39,48 Their comorbidities and functional limitations may worsen exacerbations and obscure typical signs of respiratory distress, complicating clinical assessment and increasing odds of EOT. Tools commonly used to evaluate asthma severity during hospitalization have not been validated in children with CCC and may not adequately capture their baseline complexity or acute clinical changes. Further research is needed to clarify asthma progression in children with CCC and to develop more appropriate inpatient assessment and management strategies for this growing population.

We found that Hispanic and multiracial patients had higher odds of EOT compared with non-Hispanic white patients, contributing to the literature on inequity in inpatient asthma care. These findings align with prior work highlighting racial and ethnic disparities with prehospital (ie, higher illness severity) and posthospital (ie, increased health utilization) asthma outcomes.^2,13,49 Racial disparities also exist within inpatient outcomes such as adverse patient events and acute deterioration among non–asthma-specific cohorts.^50,51 Racial disparities observed may reflect the influence of systemic and structural racism.^6,36,49,52 Neighborhood-level social needs indices, such as the COI, capture social inequities in housing, education, and community resources.^10,53–56 Low COI has been associated with adverse outcomes including higher illness severity, ICU admissions, and longer hospital stay in children with asthma.^10,53–58 Surprisingly, however, we found no association with EOT. This aligns with Byrne et al, who found no association between COI and markers of inpatient asthma severity, suggesting that standardized care may mitigate neighborhood-level social differences.¹³ However, several limitations of the COI should be considered. First, the potential for ecologic fallacy exists, as neighborhood-level indices may not accurately reflect individual social needs that influence asthma severity and care.^59–63 Second, although the COI serves as a proxy for structural and neighborhood-level inequities, it may not fully capture the systemic or implicit biases within the hospital system that may contribute to disparities observed among Hispanic and Black children in our cohort. Lastly, inequities during hospitalization related to language barriers or health literacy across racial groups are not captured in the COI or in PHIS. These limitations highlight the need to interpret neighborhood-level indices cautiously while also considering individual-level and structural factors that influence asthma out-comes.^7,28–31 Future studies should conduct subgroup analysis and explore causal pathways to disentangle the intersection of race and social factors. Better characterizing individual social needs of children requiring EOT may reveal systemic biases contributing to racial inequities and inform strategies to improve health outcomes.

Lastly, we observed seasonal variability in odds of EOT, with lower odds in the spring and higher odds in the fall compared with the summer. Seasonal variability of asthma is well known, with higher rates of admission and increased severity in the fall and winter, likely due to increased exposure to viruses and indoor allergens.^64–66 Higher illness severity in the fall may contribute to higher odds of EOT in our cohort. Interestingly, odds of EOT were lower in the spring compared with the summer. To our knowledge, no literature has described this association. One hypothesis is that staffing variations across the year, such as fewer respiratory therapists during a lower summer census or introduction of new house staff who are less experienced in asthma care, may prompt earlier escalation. Further research is needed to clarify seasonal differences and underlying contributors to EOT.

Our study had several limitations. First, PHIS is an administrative database lacking detailed variables or information to determine severity, and it does not capture other measures of EOT such as RRT activations. Second, PHIS cannot distinguish between patients transferred to the ICU on DOS 0 from those directly admitted to the ICU. The threshold for EOT may differ between institutions, and additional features of EOT may not have been captured. However, our measure still represents worsening clinical status requiring a higher level of care. Additionally, ICD-10 codes were used to identify patients with acute asthma exacerbation or status asthmaticus, which may introduce misclassification bias due to coding inaccuracies or variable documentation practices.⁶⁷ Such errors are likely nondifferential, biasing results toward the null. Although the COI 3.0 accounts for health insurance rates and some measures of health care accessibility, some dimensions of health care and environmental context may not be captured. Lastly, some results were of borderline statistical significance and should be considered hypothesis generating rather than definitive.

Most existing studies assess risk factors for ICU transfer, with only 1 examining broader predictors of EOT, leaving gaps in our understanding of EOT across the inpatient continuum. ICU transfer is an incomplete proxy for EOT, overlooking children needing higher levels of care outside of the ICU due to variable institutional policies or resources. Our study highlights associations between EOT and risk factors, including older age, male sex, race and ethnicity, presence of CCC, and seasonality. These factors are important for risk stratification and may serve as proxies for modifiable aspects of care, such as monitoring practices and severity assessment tools. Understanding how these factors influence clinical decision-making can inform targeted strategies to reduce escalation, optimize resource utilization, lower health care cost, and improve outcomes. Future research should prioritize subgroup analyses exploring casual pathways and examine modifiable features of inpatient care to identify actionable opportunities to reduce disparities and optimize care for children with asthma.

ABBREVIATIONS

aOR

adjusted odds ratio

APR-DRG

All Patient Refined Diagnosis Groups

CCC

complex chronic conditions

COI

Child Opportunity Index

DOS

day of service

EOT

escalation of therapy

ICD-10

International Classification of Diseases, Tenth Revision

ICU

intensive care unit

PHIS

Pediatric Health Information System

RRT

rapid response team