Trends in Regionalization of Emergency Care for Common Pediatric Conditions

Abstract

Background

For children who cannot be discharged from the emergency department (ED), definitive care has become less frequent at most hospitals. It is uncertain whether this is true for common conditions that do not require specialty care. We sought to determine how the likelihood of definitive care has changed for 3 common pediatric conditions: asthma, croup, and gastroenteritis.

Methods

We studied children under 18 years old presenting to Emergency Departments in the United States from 2008–2016 with a primary diagnosis of asthma, croup, or gastroenteritis, excluding critically ill patients. The primary outcome was referral rate: the number of patients transferred among all patients who could not be discharged. Analyses were stratified by quartile of annual pediatric volume. We used logistic regression to determine whether changes over time in demographics or comorbidities could account for referral rate changes.

Results

Referral rates increased for each condition in all volume quartiles. Referral rates were greatest in the lowest pediatric volume quartile. In the that quartile specifically, referral rates increased for asthma (13.6% per year, 95% confidence interval [CI] 5.6–22.2), croup (14.8% per year, 95% CI 2.6–28.3), and gastroenteritis (16.4% per year, 95% CI 3.5–31.0). Changes over time in patient age, sex, comorbidities, weekend presentation, payer mix, urban-rural location of presentation, or area income did not account for these findings.

Conclusions

Increasing referral rates over time suggest decreasing provision of definitive care and regionalization of inpatient care for three common, generally straightforward conditions.

Table of Contents Summary

Through analysis of the NEDS database, this study characterizes trends in regionalization of care for common low-risk emergency conditions based on pediatric volume of the presenting emergency department.

Introduction

Most of the 30 million childhood emergency department (ED) visits annually occur at general community hospitals, but these centers are decreasingly providing definitive pediatric care.^1–4 Pediatric patients are often transferred to tertiary care children’s hospitals for subspecialist care, inpatient admission, or intensive care needs.⁵ For critically ill patients and some specialized conditions, there are clear benefits to transfer⁶, which have informed national recommendations from the American Academy of Pediatrics for implementing regionalized emergency medical services.^7–10

However, it is uncertain which children with lower-acuity conditions benefit from transfer to pediatric tertiary care hospitals. The majority of pediatric patients transferred between EDs are transferred for simple, common conditions, and as many as one third of children transferred are discharged home without requiring further intervention or subspecialty consultation.^11,12 For patients who are admitted after transfer, many are discharged less than 24 hours later.¹³ These findings have raised concerns that a large proportion of inter-facility transfers may be potentially avoidable.¹⁴ Inter-facility transfers have potential downsides including increased cost of care, financial and time burdens to families who are transferred far from home, and crowding of referral EDs.^11,14 In addition, transfers from low-volume to high-volume centers may improve outcomes for some patients but worsen them for others.^15–17 Obtaining a better understanding of how inter-facility transfers have changed over time and the factors associated with avoidable transfers is important for public health planning and informing current recommendations around regionalization of care.

One such factor that influences an ED’s readiness to care for children is the volume of pediatric patients.¹⁸ Prior studies have shown that children transferred from lower-volume EDs are more likely to be discharged from the receiving ED,^18,19 therefore it is possible that more transfers from lower-volume EDs are potentially avoidable and that lower-volume EDs may be more likely to transfer children with all conditions.

Accordingly, our objectives were to assess how the likelihood of definitive care has changed over time for pediatric patients presenting to the ED with three common conditions: asthma, croup, and gastroenteritis. These conditions were chosen for the fact that they are common conditions which are less likely to require specialty or surgical intervention, and for the case of asthma and croup, are relatively specific diagnoses. In addition, we sought to determine how the likelihood of definitive care is influenced by pediatric volume of the presenting ED. We hypothesized that inter-facility transfers are increasing over time and that lower pediatric volume is associated with a higher likelihood of transfer for our chosen conditions.

Methods

Data Source

We used data from 2008–2016 from the Agency for Healthcare Research and Quality (AHRQ) Healthcare Cost and Utilization Project (HCUP) Nationwide Emergency Department Sample (NEDS), the largest all-payer ED database in the United States. NEDS is an annual, cross-sectional, stratified single-stage cluster sample of ED visits across 953 hospitals in 36 states and the District of Columbia.²⁰ Each year, more than 30 million ED visits are sampled, capturing 20% of all ED visits nationally. The sample is stratified by hospital characteristics including census region, trauma center designation, urban or rural location, teaching status, and hospital ownership. The database includes patient and hospital demographics, diagnosis and procedure codes, ED disposition, and charge data. Best practices for utilization of the NEDS database were strictly followed during data extraction and analysis.²¹

Study Sample

We conducted a cross-sectional study of children presenting to an ED for asthma, croup, or gastroenteritis. Visits were included if the patient was less than 18 years old and had a primary diagnosis of asthma (ages 2–17 only), croup, or gastroenteritis. These conditions were prespecified and were chosen as common, canonical pediatric conditions which do not typically require specialty care or surgical intervention. Visits were attributed to asthma if they had a first-listed discharge diagnosis of asthma according to the International Classification of Disease, Ninth Edition (ICD-9) code 493.XX ^22,23 or an ICD-10 code of J45.XX. Visits were attributed to croup if they had a primary ICD-9 diagnosis of 464.4 or ICD-10 diagnosis of J05.0. We used a definition of gastroenteritis adapted from previous literature including ICD-9 codes 1.0, 3.0, 4.0–4.1, 4.3, 4.8, 4.9–5.2, 5.89, 5.9, 8.00–8.04, 8.09, 8.2–8.3, 8.41–8.47, 8.49, 8.5, 8.61–8.64, 8.66–8.67, 8.69, 8.8, 9.0–9.3, and 558.9, and ICD-10 codes A00.0-A00.1, A009, A020, A029-A031, A03.3, A03.8-A05.2, A05.8-A05.9, A08.0, A08.11, A08.19, A08.2, A08.31–8.32, A08.39, A08.4, A08.8, A09.0, and K52.9.^24,25 (supplemental table 1).Visits were excluded if they required critical care as defined by diagnosis or procedure codes for respiratory failure, cardiac or pulmonary arrest, need for mechanical ventilation, arterial or central venous catheterization, chest tube placement, ECMO, IO placement, or dialysis (supplemental table 2). Visits in which the patient died in the ED were also excluded from analysis as they were not eligible for transfer or admission.

Variable Definitions

The primary outcome measure was referral rate, defined as the number of patients divided by all non-discharged patients (Table 1). Referral rate reflects a hospital’s capability to provide definitive post-ED care for a given condition.⁴ Secondary outcomes included absolute transfer and admission counts as wells as transfer and admission rates, defined as the proportion of all ED visits with a disposition of transfer or admission.

Table 1.

Term Definitions

Term	Formula	Interpretation
Referral Rate	T / (T+A)	Proportion of patients transferred among those requiring post-ED care
Transfer Rate	T / All	Proportion of patients transferred among all children
Admission Rate	A / All	Proportion of patients admitted among all children

T = transfers, A = Admissions, All = all ED visits

We assessed the annual pediatric volume of each ED, defined as the number of patient visits for children <18 years at each site. Volume categories were defined by quartile of pediatric volume; hospitals with more than 70% of their total visits among children were called “primarily pediatric hospitals.” Covariates were chosen to reflect location, timing of presentation, patient characteristics, and socioeconomic factors that might impact care delivery and therefore the likelihood of referring hospitals to transfer patients.

Covariates

We analyzed patient age, sex, co-diagnosis of a complex chronic condition (CCC) as defined by Feudtner et al^26,27, weekend vs. weekday presentation, Medicaid vs. non-Medicaid insurance, urban location of presenting ED (defined as large or small metropolitan or micropolitan area), and area income (the quartile of the median income for the patient’s zip code).

Data Analysis

Referral Rates

We first described patient demographics among children with asthma, croup, or gastroenteritis by pediatric volume category. Referral rates were calculated and plotted for each condition and each pediatric volume category on an annual basis. We modeled the odds of referral for each condition using logistic regression to assess for change in referrals over time. The models included year, volume category, and year-volume interaction terms.²⁸ These models were then used to generate lines of best fit for each condition and volume category.

Absolute transfer and admission counts and transfer and admission rates were determined at the start and end of the study period (2008 and 2016) to assess whether changes in referral rates were associated with decreasing numbers of admissions or increasing numbers of transfers.

Impact of Volume on Transfer and Referral Rates

To evaluate the relationship between ED volume and transfer rates, we compared transfer rates in 2016 across volume quartiles using chi square tests. We then assessed whether referral rates increased more quickly in lower- versus higher-volume institutions by examining the year-volume interaction terms in the referral models. These interaction terms evaluate differences in the slopes of referral rates over the study period.

Impact of Patient and ED Characteristics on Referral Rate

To assess the effect of possible patient- and ED-level confounders on referral rates, we repeated the referral models adding age or CCC. We hypothesized that age and CCC might affect referral rate because smaller hospitals may be more likely to transfer younger and more complex patients. We then created a full model with all covariates including the year-volume interaction, age, sex, presence of CCC, weekend vs. weekday presentation, Medicaid vs. non-Medicaid insurance, location of ED (urban vs rural), and area income. All models were determined a priori.

Statistical Analysis

All estimates were calculated using weighted counts, tests, confidence intervals, and models.²⁰ Data were analyzed using R version 3.6.0 (R Foundation, Vienna, Austria) and the R survey package for all NEDS analyses.

The Boston Children’s Hospital IRB declared the study exempt from further review.

Results

We included 13.5 million total weighted visits across the three conditions. Of these, more than 78,000 visits were excluded due to critical care diagnoses or procedures and 224 died in the ED. After exclusions, we analyzed 5.6 million weighted ED visits for children with asthma, 2.8 million with croup, and 5.0 million with gastroenteritis.

The range of quartile cut points for annualized pediatric volume were 1264–1947, 3396–4106, and 6881–8157 pediatric visits per year. Each volume quartile contained an average of 1130–1170 hospitals each year. There were an average of 33 primarily-pediatric EDs each year, all of which were in the highest volume quartile. Overall, patients seen at primarily pediatric EDs were more likely to be Medicaid-insured and have a CCC compared to patients at lower volume centers (Supplemental Table 3).

Referral rates increased in all pediatric volume categories for all conditions between 2008 and 2016, except for primarily pediatric EDs (Figure 1). Referral rates were consistently higher among the lowest-volume EDs compared with the highest volume EDs across all years (p<0.05). Referral odds in the lowest volume quartile increased 13.6 (95% CI 5.6–22.2) per year among children with asthma, 14.8% (95% CI 2.6–28.3) per year among children with croup, and 16.4% (95% CI 3.5–31.0) among children with gastroenteritis (Table 2). The rate of change in referral rate did not differ significantly between hospital volume quartiles. In 2016, the transfer rate for all conditions differed significantly between volume categories (p<0.05).

Figure 1:

Trends in referral rate for asthma, croup, and gastroenteritis

Referral rates are shown annually for each condition and hospital volume category (thin lines). Curves of best fit (thick lines) were determined using logistic regression. Volume was categorized using quartile of pediatric visits to the ED each year.

Table 2.

Unadjusted annual change in referral rates by volume category and condition

	Asthma		Croup		Gastroenteritis
Volume Category	Annual change in referral odds (95% CI)	Interaction term* (95% CI)	Annual change in referral odds (95% CI)	Interaction term* (95% CI)	Annual change in referral odds (95% CI)	Interaction term* (95% CI)
Lowest quartile	13.6 (5.6,22.2)	Ref	14.8 (2.6,28.3)	Ref	16.4 (3.5,31.0)	Ref
Quartile 2	12.2 (6.8,17.8)	1.0 (0.9,1.1)	22.3 (14.3,30.8)	1.1 (0.9,1.2)	20.3 (13.4,27.6)	1.0 (0.9,1.2)
Quartile 3	10.6 (5.8,15.7)	1.0 (0.9,1.1)	15.0 (4.2,26.8)	1.0 (0.9,1.2)	9.0 (0.6,18.0)	0.9 (0.8,1.1)
Highest quartile	9.1 (4.5,13.8)	1.0 (0.9,1.0)	11.9 (7.0,17.1)	1.0 (0.9,1.1)	9.2 (4.2,14.5)	0.9 (0.8,1.1)
Primarily Pediatric	0.8 (−17.7,23.5)	0.9 (0.7,1.1)	7.8 (−9.3,28.1)	1.0 (0.8,1.2)	1.5 (−12.7,18.1)	0.9 (0.7,1.1)

^*Assesses for difference in referral rate slopes compared with the lowest quartile

Transfer rates increased significantly from 2008 to 2016 for patients with asthma (0.3%) presenting to any ED and for patients with asthma (0.5%) or croup (0.3%) presenting to non-pediatric EDs (Table 3). Transfer rates remained unchanged for patients with gastroenteritis. Admission rates decreased for patients presenting to any ED with asthma (−2.5%), croup (−0.9%), or gastroenteritis (−1.5%).

Table 3.

Difference in Transfer and Admission Rates by Diagnosis

		Admissions			Transfers
Location of presentation	Diagnosis	2008 n (%**)	2016 n (%**)	Risk difference in admission rate (95% CI)	2008 n (%***)	2016 n (%***)	Risk difference in transfer rate (95% CI)
All EDs	Asthma	61,780 (11.0)	52,386 (8.5)	−2.5* (−4.2,−0.9)	7,466 (1.3)	10,313 (1.7)	0.3 (−0.0,0.7)
	Croup	7,208 (2.7)	6,178 (1.8)	−0.9* (−1.4,−0.4)	1,506 (0.6)	2,664 (0.8)	0.2* (0.1,0.4)
	Gastroenteritis	32,036 (5.3)	22,094 (3.8)	−1.5* (−2.4,−0.7)	2,106 (0.4)	2,096 (0.4)	0.0 (−0.1,0.1)
Non-pediatric EDs	Asthma	51,114 (10.3)	40,157 (7.8)	−2.6*(−4.3,−0.8)	6,867 (1.4)	10,015 (1.9)	0.5* (0.2,0.9)
	Croup	5,933 (2.6)	4,563 (1.6)	−1.0* (−1.4,−0.6)	1,448 (0.6)	2,513 (0.9)	0.3* (0.1,0.4)
	Gastroenteritis	26,919 (5.1)	16,922 (3.6)	−1.5* (−2.3,−0.6)	2,077 (0.4)	2,065 (0.4)	0.1 (−0.1,0.2)

^*p<0.05

^**admission rate

^***transfer rate

After adjusting for age and CCC, there was no change in referral rate trends compared with the unadjusted models (Figure 2). Trends in referral rates did not differ between the unadjusted model and full covariate models.

Figure 2.

Multivariable modelling of referral rates for asthma, croup, and gastroenteritis

Referral rate odds ratios are shown for each condition and each interaction model: unadjusted model, adjusted for age, adjusted for presence of complex chronic condition, and full multivariable model adjusted for age, sex, presence of CCC, weekend vs. weekday presentation, Medicaid vs. non-Medicaid insurance, location of ED, and area income.

Discussion

From 2008 to 2016, children with asthma, croup, and gastroenteritis who required ongoing medical care had higher referral rates across EDs nationally. Referral rates were higher at EDs with lower pediatric volumes. Temporal trends in age, CCC, and other demographics did not explain this trend.

Referral rates can increase because of decreasing hospitalizations, increasing transfers, or both. Likewise, transfer rates can increase because of decreasing ED volume, increasing transfers, or both. In our study, hospitalizations decreased in all three conditions, in keeping with overall national trends in pediatric hospitalizations.²⁹ Transfer rates increased only for asthma and croup. Taken together, these findings suggest that regionalization of care is occurring to different extents between conditions. The absence of changes in transfer rates for gastroenteritis compared to croup and asthma may represent differences in capability to provide definitive care for particular conditions. Asthma and croup are respiratory illnesses, which have a higher potential to escalate to severe disease requiring positive pressure or intubation. Patients with gastroenteritis may require IV rehydration as their only hospital-level intervention. Decreases in admission rates across all conditions may also reflect an improvement in capability of EDs to discharge these patients or decreasing severity of illness in these patients.

The reason for these changes is likely multi-factorial. First, given that pediatric training of referring clinician is associated with a decreased likelihood of ED discharge after transfer,³⁰ differences in clinician training may be responsible for at least some proportion of transfers. Thus, it is possible that changes in clinical training over time have left non-pediatric trained providers less comfortable with taking care of pediatric patients. The smaller number of pediatric-trained ED physicians and fewer pediatric-specific resources (including respiratory therapists and access to PICU) at lower volume centers may also partially explain the higher rates of transfer from these hospitals.¹⁸ Second, it is possible the increasing referral rate could be due to decreasing availability of inpatient pediatric care at originating EDs. However, there is little data currently available on whether the availability of pediatric inpatient care is changing. Third, it is possible there is increasing societal, medicolegal, parental, or pediatrician pressure on physicians at low pediatric volume centers to transfer pediatric patients, although this hypothesis is unproven. Finally, it is possible that increasing medical complexity of non-discharged pediatric patients is driving transfer to tertiary care centers.³¹

Our findings provide additional evidence of increasing regionalization. Previous work has demonstrated the benefits of regionalization of critical and highly specialized pediatric care. Children with complex conditions are increasingly being cared for at specialty hospitals.³¹ For patients with some chronic conditions including congenital cardiac disease^32,33 or sickle cell disease^34,35, as well as critically ill pediatric or neonatal patients, receiving care at higher-volume centers is associated with improved outcomes.^7,9 However, it is not yet clear to what extent all patients benefit from care at high-volume or primarily pediatric centers.

Regionalization may lead to a number of unwanted effects. If pediatric transfers increase, higher-volume and primarily pediatric EDs may experience increased crowding. Overcrowding increases return visit rates, leads to delays in antibiotic and pain medicine administration, and generally worsens outcomes for all patients.^36–38 Furthermore, with increasing regionalization, community EDs may be less likely to provide pediatric services in the future. This could lead to decreased access to care for patients who live far from tertiary care centers. Additionally, potentially unnecessary transfers are costly to an already overburdened healthcare system and to families.^11,14

Our study has several limitations. The NEDS database does not include clinical information, and thus it is possible that due to diagnosis or procedure mislabeling we unintentionally included some patients requiring critical care. However, given that the overall number of patients with critical disease is relatively low in the pediatric population and would likely disproportionately occur at primarily pediatric hospitals, we do not believe this would have a significant impact on our findings. Additionally, we may have overcounted or undercounted visits for each diagnosis because of errors in diagnosis coding, however we do not have reason to believe those coding issues would affect the likelihood of transfer vs hospitalization. Although we adjusted for presence of complex chronic condition in general, we grouped all conditions together and did not specify which condition, as such it is also possible that more severe complex chronic conditions might have a differential effect on likelihood of transfer compared to less severe chronic conditions.

Conclusion

Increasing numbers of pediatric patients with common conditions who need ongoing post-ED care are being transferred rather than hospitalized at the presenting facility. Importantly, these trends exist for certain conditions including asthma and croup but do not hold for all conditions studied. These findings provide further evidence of pediatric care regionalization occurring even for common conditions that do not routinely require specialty care.

What’s Known on This Subject

Pediatric emergency care delivery is becoming more regionalized, particularly for subspecialty and surgical care. There are improved outcomes with regionalization of certain complex or acute populations. It is not clear to what extent regionalization is occurring for lower-risk pediatric patients.

What This Study Adds

Pediatric patients with common conditions including asthma, croup, and gastroenteritis are increasingly transferred from the emergency room when they require ongoing care. Patients who present to lower pediatric volume emergency departments are more likely to be transferred.

Abbreviations

CCC

Complex Chronic Condition

CI

Confidence Interval

ED

Emergency Department

HCUP

Healthcare Cost and Utilization Project

ICD

International Classification of Disease

NEDS

National Emergency Department Sample