BACKGROUND AND OBJECTIVES:

Croup is a clinical diagnosis, and the available evidence suggests that, except in rare cases, ancillary testing, such as radiologic imaging, is not helpful. Given the paucity of inpatient-specific evidence for croup care, we hypothesized that there would be marked variability in the use of not routinely indicated resources (NRIRs). Our primary study objective was to describe the variation and predictors of variation in the use of NRIRs.

METHODS:

This was a retrospective cohort study that used the Pediatric Health Information System database of generally healthy inpatients with croup aged 6 months to 15 years who were admitted between January 1, 2012 and September 30, 2014. We measured variability in the use of NRIRs: chest and lateral neck radiographs, viral testing, parenteral steroids, and antibiotics. Risk-adjusted analysis was used to compare resource utilization adjusted for hospital-specific effects and average case mix.

RESULTS:

The cohort included 26 hospitals and 6236 patients with a median age of 18 months. Nine percent of patients required intensive care services, and 3% had a 30-day readmission for croup. We found marked variability in adjusted and unadjusted utilization across hospitals for all resources. In the risk-adjusted analysis, hospital-specific effects rather than patient characteristics were the main predictor of variability in the use of NRIRs.

CONCLUSIONS:

We observed an up to fivefold difference in NRIR utilization attributable to hospital-level practice variability in inpatient croup care. This study highlights a need for inpatient-specific evidence and quality-improvement interventions to reduce unnecessary utilization and to improve patient outcomes.

What’s Known on This Subject:

Croup is a clinical diagnosis, and the available evidence suggests that, except in rare cases where alternative diagnoses are being considered, resources such as radiologic imaging and viral testing are not routinely indicated.

What This Study Adds:

There is marked variation across free-standing children’s hospitals in the use of not routinely indicated resources in croup that is not explained by average patient case mix. This study highlights a need for quality-improvement interventions to reduce unnecessary utilization.

Croup affects >1.4 million children under 6 years of age in the United States annually.1,2 In most patients, it is a mild, self-limited disease characterized by rhinorrhea, a hoarse voice, a barky cough, and stridor. The majority of patients with croup are cared for in the outpatient setting, but croup can be a life-threatening disease. Overall, 1.5% to 6% of patients with croup are hospitalized at an estimated cost of $56 million annually in the United States.1,3,5 Despite the potential severity and cost associated with hospitalization, there is a paucity of inpatient-specific research on the diagnosis, optimal treatment, and outcomes for inpatients with croup. As such, many hospitals use evidence from outpatient studies for the inpatient management of croup.

Croup is a clinical diagnosis, and the available evidence suggests that, except in rare patients with atypical presentations in whom alternative diagnoses such as bacterial tracheitis or retropharyngeal abscess are being considered, ancillary testing such as radiologic imaging and viral testing is not helpful.6,7 Systemic corticosteroids are the main treatment in croup, and there is strong evidence to support a single dose of dexamethasone (0.15–0.6 mg/kg) in all patients with croup regardless of the care setting.6,8 Given equal efficacy, except in severe cases or when oral intake is not tolerated, the oral route is preferred to the parenteral route for corticosteroid administration.9,11 Because croup is a viral illness, antibiotics are not indicated except in cases of concomitant bacterial infection.

This study aimed to analyze variation in the use of not routinely indicated resources (NRIRs) in the inpatient management of croup, including the following: the use of chest and neck radiographs, viral testing, antibiotics, and parenteral rather than enteral corticosteroids. The study had 3 primary objectives: (1) to describe unadjusted hospital-level variation in inpatient management, patient outcomes, and utilization of NRIRs, (2) to identify patient factors associated with the utilization of NRIRs, and (3) to determine the predictors of NRIR utilization. We hypothesized that, given the lack of inpatient-specific evidence and practice guidelines, there would be significant variation in the evaluation and management of generally healthy patients hospitalized with croup. We further hypothesized that the main predictor of variation in the use of NRIRs would be the hospital of admission. To test our hypothesis and compare utilization across hospitals that may admit different patient populations, we used risk-adjustment methods for our analysis.

Our secondary aim was to describe the relationship between resource utilization and patient outcomes including the following: readmission for croup within 30 days, return to the emergency department (ED) for croup within 7 days, ICU admission, and intubation. We hypothesized that patient outcomes would not vary significantly between hospitals or by the number of NRIRs used.

For this retrospective cohort study, we used the Pediatric Health Information System (PHIS) database (Children’s Hospital Association, Overland Park, KS). PHIS contains de-identified administrative data on demographic characteristics, diagnoses, procedures, imaging, medications (including route of administration), and readmissions from 47 children’s hospitals in the United States. The data include International Classification of Diseases, Ninth Revision, Clinical Modification, codes and Clinical Transaction Classification codes for each procedure and clinical services, for each patient by hospital day. This study was approved as exempt by the Colorado Multiple Institutional Review Board because it did not involve human subjects.

Patients aged >6 months and <14 years were eligible for inclusion if they were discharged from an inpatient unit or observation status between January 1, 2012, and September 30, 2014 with an International Classification of Diseases, Ninth Revision (ICD-9), code representing viral croup (see Appendix 1). We excluded patients with complex chronic conditions as defined by Feudtner et al12 because these patients may warrant nonstandard evaluation and management. We also excluded patients with diagnoses for which parenteral steroids, viral testing, radiographs, or antibiotics may be indicated, such as asthma, pneumonia, and otitis media. In addition, we excluded patients with diagnoses for which the croup diagnosis had a high likelihood of being secondary, such as congenital anomalies of the larynx/trachea, burns, foreign body ingestion/aspiration, trauma, a surgical diagnosis, and motor vehicle accidents.

Study variables were defined a priori, and complete data were available for 26 PHIS- participating hospitals representing all regions of the country. Patient-level characteristics included the following: demographic characteristics (sex, age in months, race/ethnicity, and public versus private insurance status), season of discharge, All Patient Refined Diagnosis Related Group (APR-DRG) severity classification (3M APR-DRG Classification System), length of stay (LOS) in days, number of hospital days patient received inhaled racemic epinephrine treatments, and number of hospital days patient received steroids (oral or parenteral). Steroid use was measured in days, because individual doses could not be differentiated in PHIS data. Hospital-level variables included number of admissions for croup per year, hospital region (Midwest, Northeast, South, West), mean annual patient census, and admission rate for croup from the ED, which was calculated as the number of patients admitted through the ED with croup over the total number of patients seen in the ED with croup.

For our primary outcomes, we examined the following resources utilization variables during the index croup admission by using associated billing codes: parenteral steroids, viral studies, chest radiographs (CXRs), lateral neck radiographs, and antibiotics. For our secondary outcomes, we examined data on readmission for croup within 30 days, return to the ED within 7 days, ICU admission, and intubation. Although most readmissions for croup occur within 7 days, we examined 30-day readmissions because a common rationale for ordering NRIRs is to avoid a missed diagnosis that may result in a delayed readmission.

Descriptive statistics were used to describe the cohort. Variance across hospitals was analyzed with medians and ranges. Univariate associations between independent variables and NRIR utilization were determined. The final multivariable logistic regression model assessing the outcome of 0 to 1 versus ≥2 NRIRs was adjusted for all patient-level variables (age, sex, race, insurance status, season of discharge, APR-DRG severity, and LOS) and controlled for hospital variability through random effects.

In a secondary analysis, we compared outcomes adjusted for (1) patient effects based on average case mix, defined here as “expected rates” of utilization, and (2) hospital-specific effects, defined here as “predicted rates” of utilization. First, for each NRIR we estimated the expected rates of utilization by hospital on the basis of their average case mix by using multivariable logistic regression with patient factors as fixed effects.13 Patient factors in the model included the following: age, sex, race, insurance status, season of discharge, APR-DRG severity, and LOS. We then calculated predicted rates of utilization of each NRIR by hospital by using hierarchical multivariable logistic regression with hospital as a random effect and patient characteristics as fixed effects. We calculated predicted/expected (p/e) ratios for each hospital for each NRIR. Finally, for each NRIR we multiplied the p/e ratio for each hospital by the average observed rates of utilization across all hospitals to calculate risk standardized utilization outcomes. To allow for comparison of what we observed to what would be expected given each hospital’s average case mix, observed, expected, and predicted outcomes rates for the use of NRIRs were depicted graphically.

To analyze potential patterns of resource use across hospitals, for each risk-standardized outcome, hospitals were ranked from the lowest quartile for utilization to the highest quartile for utilization. We also analyzed hospital-level characteristics of the highest-quartile utilizers and lowest-quartile utilizers.

Finally, to explore associations between resource utilization and patient outcomes, we depicted the proportion of patients at each hospital with each outcome (LOS >1 day, intubation, ICU admission, 30-day readmission, and return to ED within 7 days) stratified by 3 categories of utilization at the hospital level: (1) lowest-quartile rank sum utilization for all resources, (2) hospitals in quartiles 2 and 3, and (3) hospitals with highest-quartile rank sum utilization. For all hospital-level comparisons, hospitals were labeled A through Z, with consistent labeling across Figs 24.

Twenty-six hospitals with complete study data matching our study definitions were selected before data extraction from PHIS. Only the first croup admission for each patient during the study period was included in our cohort. Subsequent hospitalizations within 30 days counted as readmissions. There were 11 210 unique patients aged 6 months to 14 years with a primary or secondary diagnosis of croup (see Appendix 2). We excluded patients with complex chronic conditions, diagnoses for which utilization of measured resources may be indicated, and diagnoses for which croup was likely secondary as described in the Methods section (Fig 1).

FIGURE 1

Variation in inpatient croup management and outcomes.

FIGURE 1

Variation in inpatient croup management and outcomes.

Our final cohort included 6236 unique patients with a median age of 18 months (Table 1). The majority of patients were admitted through the ED (80%; n = 5010) and were classified as an APR-DRG of minor severity. Across the 26 hospitals in our cohort, the median (SD) croup admissions per year was 69 (53) and the median admission rate for croup from the ED was 9.1% (4.6%).

TABLE 1

Patient and Hospital Characteristics

Value
Patients (N = 6236)  
 Median (IQR) age, mo 18 (12–29) 
 Male sex, % (n68 (4263) 
 Race, % (n 
  White 66 (4088) 
  Black 11 (699) 
  Asian/other/missing 23 (1449) 
  Hispanic ethnicity 24 (1474) 
 Public insurance, % (n45 (2801) 
 Season at discharge, % (n 
  Winter 24 (1505) 
  Spring 17 (1079) 
  Summer 22 (1351) 
  Fall 37 (2301) 
 Severity, % (n 
  Extreme/major/moderate 30 (1874) 
  Minor 70 (4362) 
Hospitals (N = 26), median (SD)  
 Croup admissions/year 69 (53) 
 Annual patient census 200 (77) 
 Admission rate for croup from ED, % 9.1 (4.6) 
Value
Patients (N = 6236)  
 Median (IQR) age, mo 18 (12–29) 
 Male sex, % (n68 (4263) 
 Race, % (n 
  White 66 (4088) 
  Black 11 (699) 
  Asian/other/missing 23 (1449) 
  Hispanic ethnicity 24 (1474) 
 Public insurance, % (n45 (2801) 
 Season at discharge, % (n 
  Winter 24 (1505) 
  Spring 17 (1079) 
  Summer 22 (1351) 
  Fall 37 (2301) 
 Severity, % (n 
  Extreme/major/moderate 30 (1874) 
  Minor 70 (4362) 
Hospitals (N = 26), median (SD)  
 Croup admissions/year 69 (53) 
 Annual patient census 200 (77) 
 Admission rate for croup from ED, % 9.1 (4.6) 

IQR, interquartile range.

Table 2 shows unadjusted patient- and hospital-level data on management and outcomes. Depending on the hospital of admission, 10% to 58% of hospitalized patients received ≥2 days of corticosteroids. The majority of patients received 1 day of steroids (51%; n = 3195) and 1 day of racemic epinephrine (55%; n = 3443). However, 19% (n = 1197) of patients did not receive any steroids and 31% (n = 1941) did not receive any racemic epinephrine. Although 30% of patients received ≥2 days of steroids, only 14% received ≥2 days of racemic epinephrine. On the final day of hospitalization, 24% of patients received steroids and 8% received racemic epinephrine.

TABLE 2

Patient- and Hospital-Level Variations in Management and Outcomes

Patients (N = 6236), % (n)% Patients by Hospital, Median (Range)
Days of steroids   
 None 19 (1197) 17 (4–45) 
 1 51 (3195) 50 (33–66) 
 ≥2 30 (1870) 27 (10–58) 
Days of racemic epinephrine   
 None 31 (1941) 24 (7–79) 
 1 55 (3443) 62 (15–72) 
 ≥2 14 (873) 14 (0–35) 
ICU admission 9 (567) 7 (1–27) 
Intubation 3 (165) 3 (0–6) 
ICU without intubation 7 (413) 5 (0–24) 
LOS, in days   
 1 82 (5096) 81 (66–92) 
 2 12 (736) 12 (6–24) 
 ≥3 6 (404) 8 (2–11) 
Readmission rate within 30 days for croup 3 (166) 2 (1–5) 
Return to ED with 7 days (croup only) 2 (105) 1 (0–3) 
Patients (N = 6236), % (n)% Patients by Hospital, Median (Range)
Days of steroids   
 None 19 (1197) 17 (4–45) 
 1 51 (3195) 50 (33–66) 
 ≥2 30 (1870) 27 (10–58) 
Days of racemic epinephrine   
 None 31 (1941) 24 (7–79) 
 1 55 (3443) 62 (15–72) 
 ≥2 14 (873) 14 (0–35) 
ICU admission 9 (567) 7 (1–27) 
Intubation 3 (165) 3 (0–6) 
ICU without intubation 7 (413) 5 (0–24) 
LOS, in days   
 1 82 (5096) 81 (66–92) 
 2 12 (736) 12 (6–24) 
 ≥3 6 (404) 8 (2–11) 
Readmission rate within 30 days for croup 3 (166) 2 (1–5) 
Return to ED with 7 days (croup only) 2 (105) 1 (0–3) 

Although the majority of patients had an LOS of 1 day, at the hospital level the proportion of patients staying 2 days ranged from 6% to 24%. Overall, 9% (n = 567) of patients required intensive care services. At the hospital level, the proportion of patients admitted to the ICU but not intubated ranged from 0% to 24%. Three percent (n = 166) of patients were readmitted within 30 days for croup. Mortality was rare (1 in 6236; 0.02%).

Table 3 shows the basic distribution of NRIR utilization across hospitals. The widest variation was seen in the use of parenteral rather than oral steroids and viral studies.

TABLE 3

Hospital-Level Unadjusted Resource Utilization

RangeMedianInterquartile Range
CXR 9–44 24 19–27 
Lateral neck radiograph 8–51 21 17–26 
Viral studies 1–40 10 4–16 
Parenteral steroid use 16–88 41 28–63 
Antibiotic use 4–15 7–12 
Use of ≥2 of above resources 16–61 29 23–39 
RangeMedianInterquartile Range
CXR 9–44 24 19–27 
Lateral neck radiograph 8–51 21 17–26 
Viral studies 1–40 10 4–16 
Parenteral steroid use 16–88 41 28–63 
Antibiotic use 4–15 7–12 
Use of ≥2 of above resources 16–61 29 23–39 

N = 6236. Data are presented as ranges, medians, and interquartile ranges for the percentage of patients across all hospitals who received the resources in column 1.

Table 4 shows patient factors that were associated with the use of ≥2 NRIRs in the adjusted multivariable logistic regression. The majority of the cohort (71%) received 0 to 1 NRIR. Patient characteristics associated with a higher odds of receiving ≥2 NRIRs included the following: age ≥3 years, black race, discharge outside of the fall season, nonminor APR-DRG severity, and LOS >1 day.

TABLE 4

Associations With Utilization of ≥2 NRIRs

0–1 NRIR (71%), %≥2 NRIRs (29%), %Unadjusted OR (95% CI)Adjusted OR (95% CI)
Patients     
 Age category     
  1 year 22 26 1.19 (1.03–1.37) 1.07 (0.91–1.25) 
  1 to <2 years 42 40 Ref Ref 
  2 to <3 years 17 14 0.92 (0.78–1.09) 0.96 (0.80–1.16) 
  ≥3 years 18 20 1.26 (1.08–1.48) 1.33 (1.13–1.58) 
 Sex     
  Male 69 67 Ref Ref 
  Female 31 33 1.09 (0.96–1.22) 1.10 (0.97–1.26) 
 Race     
  White 66 66 Ref Ref 
  Black 10 14 1.41 (1.17–1.69) 1.26 (1.03–1.54) 
  Asian/other/missing 24 20 0.97 (0.83–1.12) 0.93 (0.79–1.10) 
Insurance     
 Commercial or other 56 52 Ref Ref 
 Public 44 48 1.06 (0.94–1.19) 0.98 (0.95–1.01) 
Season at discharge     
 Winter 23 26 1.38 (1.19–1.60) 1.20 (1.02–1.41) 
 Spring 17 19 1.31 (1.11–1.54) 1.27 (1.06–1.52) 
 Summer 21 23 1.24 (1.06–1.45) 1.31 (1.11–1.54) 
 Fall 39 33 Ref Ref 
Severity     
 Extreme/major/moderate 25 42 2.33 (2.05–2.65) 1.53 (1.33–1.77) 
 Minor 75 58 Ref Ref 
LOS     
 1 day 90 62 Ref Ref 
 2 days 20 3.87 (3.27–4.57) 3.69 (3.11–4.38) 
 ≥3 days 18 15.93 (12.31–20.61) 13.48 (10.33–17.59) 
0–1 NRIR (71%), %≥2 NRIRs (29%), %Unadjusted OR (95% CI)Adjusted OR (95% CI)
Patients     
 Age category     
  1 year 22 26 1.19 (1.03–1.37) 1.07 (0.91–1.25) 
  1 to <2 years 42 40 Ref Ref 
  2 to <3 years 17 14 0.92 (0.78–1.09) 0.96 (0.80–1.16) 
  ≥3 years 18 20 1.26 (1.08–1.48) 1.33 (1.13–1.58) 
 Sex     
  Male 69 67 Ref Ref 
  Female 31 33 1.09 (0.96–1.22) 1.10 (0.97–1.26) 
 Race     
  White 66 66 Ref Ref 
  Black 10 14 1.41 (1.17–1.69) 1.26 (1.03–1.54) 
  Asian/other/missing 24 20 0.97 (0.83–1.12) 0.93 (0.79–1.10) 
Insurance     
 Commercial or other 56 52 Ref Ref 
 Public 44 48 1.06 (0.94–1.19) 0.98 (0.95–1.01) 
Season at discharge     
 Winter 23 26 1.38 (1.19–1.60) 1.20 (1.02–1.41) 
 Spring 17 19 1.31 (1.11–1.54) 1.27 (1.06–1.52) 
 Summer 21 23 1.24 (1.06–1.45) 1.31 (1.11–1.54) 
 Fall 39 33 Ref Ref 
Severity     
 Extreme/major/moderate 25 42 2.33 (2.05–2.65) 1.53 (1.33–1.77) 
 Minor 75 58 Ref Ref 
LOS     
 1 day 90 62 Ref Ref 
 2 days 20 3.87 (3.27–4.57) 3.69 (3.11–4.38) 
 ≥3 days 18 15.93 (12.31–20.61) 13.48 (10.33–17.59) 

Multivariable logistic regression model for the outcome of 0 to 1 compared with ≥2 resources. Resources include intravenous/intramuscular steroids, antibiotics, CXRs, lateral neck films, and viral testing. The model was adjusted for all patient-level variables (age, sex, race, insurance status, season of discharge, APR-DRG severity, and LOS) and controlled for hospital random effect. CI, confidence interval; OR, odds ratio; Ref, reference.

In the risk-adjusted analysis, hospital-specific effects were the main predictor of variability in the use of parenteral steroids, CXRs, neck radiographs, and viral studies. By contrast, the main predictor of variability in antibiotic use was unmeasured patient characteristics (Fig 2). For parenteral steroids, observed and predicted rates were very similar across all hospitals and hospitals with the highest p/e ratios corresponded to hospitals with the highest observed proportion of parenteral steroids. For example, patients admitted to hospital A were less likely to receive parenteral steroids than would be expected on the basis of average case mix, whereas patients admitted to hospital Z were more likely to receive parenteral steroids than would be expected on the basis of average case mix. Graphs for CXRs, lateral neck films, and viral studies had patterns very similar to parenteral steroids (see Supplemental Fig 5). In contrast, in graphs depicting antibiotic use, observed and predicted rates were different. Compared with the other outcomes, the hospital of admission had less influence on whether patients received antibiotics, indicating the variation was likely due to unmeasured patient characteristics rather than the hospital of admission.

FIGURE 2

Risk-adjusted comparison of the use of parenteral steroids (A) and antibiotics (B). Observed rates (+ = observed) of utilization of parenteral steroids and antibiotics for each hospital A through Z were compared with risk-adjusted rates. The analysis adjusted for hospital-specific effects (o = predicted] and average case mix (x = expected). Hospital-specific effects were the main determinant of variability in the use of parenteral steroids. The main determinants of variability in antibiotic use were likely unmeasured patient characteristics.

FIGURE 2

Risk-adjusted comparison of the use of parenteral steroids (A) and antibiotics (B). Observed rates (+ = observed) of utilization of parenteral steroids and antibiotics for each hospital A through Z were compared with risk-adjusted rates. The analysis adjusted for hospital-specific effects (o = predicted] and average case mix (x = expected). Hospital-specific effects were the main determinant of variability in the use of parenteral steroids. The main determinants of variability in antibiotic use were likely unmeasured patient characteristics.

Figure 3 depicts hospital-level comparisons of risk-standardized utilization for each resource across the cohort. Hospitals are ordered top to bottom from lowest to highest risk-standardized rank sum for all outcomes. No hospitals were ranked uniformly in the highest quartiles or lowest quartiles of utilization for all outcomes, but, with the exception of antibiotics, hospitals tended to uniformly use a fewer or greater number of NRIRs.

FIGURE 3

Risk-standardized resource use by hospital. With the use of risk-standardized rates of utilization, each hospital (A–Z) was given a quartile rank of 1 through 4 for each resource on row 1. Hospitals were ordered by their rank sum for all resources. Shading corresponds to rank for each outcome. White shading represents the lowest quartile of utilization and black represents the highest quartile of utilization. Hospitals with the lowest-quartile risk-standardized rates of utilization are underlined in bold in column 1. IM, intramuscular; IV, intravenous.

FIGURE 3

Risk-standardized resource use by hospital. With the use of risk-standardized rates of utilization, each hospital (A–Z) was given a quartile rank of 1 through 4 for each resource on row 1. Hospitals were ordered by their rank sum for all resources. Shading corresponds to rank for each outcome. White shading represents the lowest quartile of utilization and black represents the highest quartile of utilization. Hospitals with the lowest-quartile risk-standardized rates of utilization are underlined in bold in column 1. IM, intramuscular; IV, intravenous.

Characteristics of the highest-quartile utilizers and the lowest-quartile utilizers are shown in Table 5. Compared with the other hospitals in our cohort, the lowest-quartile utilizers and the highest-quartile utilizers were not different from other hospitals with respect to measured characteristics.

TABLE 5

Characteristics of Highest- and Lowest-Quartile–Ranked Hospitals for Overall Risk-Standardized Utilization

All Other Hospitals (n = 20)Lowest-Quartile Utilizers (n = 6)P
Region of country, % (n   
 Midwest 20 (4) 17 (1)  
 Northeast 15 (3) 0 (0)  
 South 45 (9) 17 (1)  
 West 20 (4) 67 (4) .21a 
Census, median (25%–75%) 200 (163–224) 199 (139–237) .82b 
Croup cases per year, median (25%–75%) 64 (48–102) 83 (66–154) .32b 
Admission rate, median (25%–75%) 8.8 (4.8–12.1) 13.1 (4.4–16.6) .26a 
  Highest-quartile utilizers (n = 6)  
Region of country, % (n   
 Midwest 20 (4) 17 (1)  
 Northeast 10 (2) 17 (1)  
 South 40 (8) 33 (2)  
 West 30 (6) 33 (2) .99a 
Census, median (25%–75%) 199 (160–238) 201 (163–218) .97b 
Croup cases per year, median (25%–75%) 89 (52–123) 62 (48–66) .24b 
Admission rate, median (25%–75%) 9.1 (4.9–12.1) 10.8 (4.0–13.4) .93b 
All Other Hospitals (n = 20)Lowest-Quartile Utilizers (n = 6)P
Region of country, % (n   
 Midwest 20 (4) 17 (1)  
 Northeast 15 (3) 0 (0)  
 South 45 (9) 17 (1)  
 West 20 (4) 67 (4) .21a 
Census, median (25%–75%) 200 (163–224) 199 (139–237) .82b 
Croup cases per year, median (25%–75%) 64 (48–102) 83 (66–154) .32b 
Admission rate, median (25%–75%) 8.8 (4.8–12.1) 13.1 (4.4–16.6) .26a 
  Highest-quartile utilizers (n = 6)  
Region of country, % (n   
 Midwest 20 (4) 17 (1)  
 Northeast 10 (2) 17 (1)  
 South 40 (8) 33 (2)  
 West 30 (6) 33 (2) .99a 
Census, median (25%–75%) 199 (160–238) 201 (163–218) .97b 
Croup cases per year, median (25%–75%) 89 (52–123) 62 (48–66) .24b 
Admission rate, median (25%–75%) 9.1 (4.9–12.1) 10.8 (4.0–13.4) .93b 

The lowest-quartile utilizers are hospitals in the lowest quartile for the rank sum of risk-adjusted NRIR utilization, and the highest-quartile utilizers are the hospitals in the highest quartile for the rank sum of risk-adjusted NRIR utilization.

a

Fisher’s exact test.

b

Wilcoxon test.

Although we found wide variation in the risk-adjusted use of NRIRs across hospitals, the variability in patient outcomes, such as intubation (3%; range: 0%–6%), return to ED within 7 days for croup (1%; range: 0%–3%), and 30-day readmission (2%; range: 0%–3%), was narrow by comparison. Figure 4 displays the variability across hospitals for patient outcomes. No clear associations between utilization and outcomes emerged. For example, although hospital Z has the second-highest proportion of higher-severity patients and is in the highest quartile for risk-adjusted utilization, it has ICU admission and intubation rates below the hospital median. Hospital C has the highest proportion of higher-severity patients, is in the lowest quartile for risk-adjusted utilization, and has the highest proportion of 30-day readmissions for croup.

FIGURE 4

Range across hospitals for patient outcomes. From top to bottom, the bar graphs labeled I through VII show the proportion of patients at each hospital with each outcome. Hospitals are ordered A through Z across the x axis. I, proportion of patients at each hospital with nonminor severity; II, LOS >1 day; III, intubation; IV, ICU admission; V, ICU admission without intubation; VI, readmission within 30 days for croup; VII, return to ED within 7 days for croup.

FIGURE 4

Range across hospitals for patient outcomes. From top to bottom, the bar graphs labeled I through VII show the proportion of patients at each hospital with each outcome. Hospitals are ordered A through Z across the x axis. I, proportion of patients at each hospital with nonminor severity; II, LOS >1 day; III, intubation; IV, ICU admission; V, ICU admission without intubation; VI, readmission within 30 days for croup; VII, return to ED within 7 days for croup.

To our knowledge, this is the first study to examine hospital-level variability in the inpatient management of croup. We observed marked variation in the use of NRIRs in the management of croup, including the following: parenteral rather than oral steroids, CXRs, lateral neck films, viral testing, and antibiotics. With the notable exception of antibiotic use, in the risk-adjusted analysis, this wide variation did not reflect differences in average patient case mix but rather hospital-level practice variability. For example, in the case of parenteral steroids, we observed a difference of more than fivefold in utilization attributable to hospital-level practice variability, with some hospitals overutilizing and others underutilizing resources compared with what was expected on the basis of patient case mix. Conversely, we did not find significant variability in patient outcomes.

This level of variability in practice may reflect the paucity of inpatient-specific evidence for the management of croup. Despite the significant cost and health care resources associated with this common childhood illness, and the large number of children hospitalized for croup, there are essentially no data on the optimal management of inpatients with croup. Where strong evidence to guide care is limited, local differences in provider practice may become more prominent.14 

At the hospital level, we also found striking variation in the use of corticosteroids, with 10% to 58% of hospitalized patients receiving ≥2 days of corticosteroids depending on the hospital. Given the high rate of ICU admission, future studies need to determine whether inpatients may benefit from multiple doses rather than a single dose of dexamethasone.6,7 In addition, more research is needed to understand why, unlike other outcomes, the hospital of admission did not influence antibiotic use. We hypothesize that this finding is due to patient-level variables that were not captured in our analysis, but other explanations may also exist.

Similar studies have shown that increased utilization is not associated with improved patient outcomes.15,18 We found that nearly 1 in 10 patients hospitalized with croup required intensive care services and 3% of patients required intubation. However, ICU admission rates varied considerably by hospital, ranging from 1% to 27%. Furthermore, the 2 hospitals with the highest ICU admission rates did not have parallel rates of intubation, suggesting a possible overutilization of intensive care resources. This practice variation at the hospital level that was unexplained by differences in patient case mix signifies that there is significant opportunity to improve care for patients hospitalized with croup, and a potential for cost savings.

Health care reform and value-based health care models are driving practice change nationwide.19 Describing variation is only the first step to improving care. Quality-improvement interventions are needed to reduce unnecessary utilization and to improve care for inpatients with croup. One strategy, developing clinical care guidelines, has reduced unnecessary utilization in other pediatric illnesses.20,23 However, croup studies to date have focused on outpatient management, and inpatient-specific evidence is needed to inform clinical practice guidelines and protocols in the inpatient setting. In addition, the future success of guideline implementation in croup necessitates an understanding of the factors associated with medical providers’ decisions to use resources and how utilization is linked to patient-centered outcomes such as health-related quality of life.

Our study has several limitations. This retrospective cohort study included administrative data from 26 tertiary care children’s hospitals that had complete ED, observation, and inpatient data for all study variables in PHIS during the study period. Without clinical data, our ability to control for severity was limited to APR-DRG severity classifications. We used ICD-9 codes to exclude patients with complex chronic conditions and other comorbidities, but ICD-9 codes may not accurately capture all comorbidities. Because our data are only from free-standing children’s hospitals, our results may not be generalizable to other settings. Our cohort may represent a more severe patient population than the general population, resulting in higher resource utilization. Alternatively, providers at children’s hospitals may have more experience with croup management or may have institutional clinical care guidelines to guide management and treatment decisions, in which case our data may underestimate utilization. Finally, during the index hospitalization, 19% (n = 1197) of our cohort did not have record of receiving any steroids. Some of these patients may have received steroids (or NRIRs) as an outpatient before presenting to a PHIS hospital for admission, and these data were not captured in our study.

We observed an up to fivefold difference in utilization attributable to hospital-level practice variability in the inpatient management of croup. This marked variation raises concerns about potential over- or underutilization of resources not routinely indicated in the management of croup. There is a critical need for inpatient-specific croup research to define the best care for inpatients with croup and for quality-improvement interventions to reduce unwarranted resource utilization to deliver high-value care and to improve outcomes for inpatients with croup.

ICD-9 codes representing viral croup in our study included the following: 464.4 (croup), 464.20 (acute laryngotracheitis without mention of obstruction), 464.21 (acute laryngotracheitis with obstruction), 464.50 (unspecified supraglottitis without mention of obstruction), 464.51 (unspecified supraglottitis with obstruction), and 786.10 (stridor).

Nearly all patients, or 99.5% (n = 6208), had a primary or secondary diagnosis code of 464.4 (croup). Two percent (n = 145) had a primary or secondary diagnosis code of 464.10–464.11 (acute tracheitis with and without mention of obstruction), 1.1% (n = 70) had a primary or secondary diagnosis code of 464.20–464.21 (acute laryngotracheitis with and without mention of obstruction), and 0.1% (n = 5) had a primary or secondary diagnosis of 464.50 (unspecified supraglottitis without mention of obstruction). The majority of patients were admitted through the ED (80%; n = 5010), classified as APR-DRG minor severity, and had an LOS of 1 day. Nine percent of patients required intensive care services, 3% of patients were readmitted within 30 days for croup, and 2% returned to the ED within 7 days of hospital discharge for croup. Mortality was rare (1 in 6236; 0.02%).

     
  • APR-DRG

    All Patient Refined Diagnosis Related Group

  •  
  • CXR

    chest radiograph

  •  
  • ED

    emergency department

  •  
  • ICD-9

    International Classification of Diseases, Ninth Revision

  •  
  • LOS

    length of stay

  •  
  • NRIR

    not routinely indicated resource

  •  
  • p/e

    predicted/expected

  •  
  • PHIS

    Pediatric Health Information System

Dr Tyler conceptualized and designed the study, participated in data collection, assisted in data analysis, drafted the initial manuscript, and critically reviewed and revised the manuscript; Drs McLeod, Kempe, and Juarez-Colunga conceptualized and designed the study, participated in data analysis, and critically reviewed and revised the manuscript; Ms Beaty conceptualized and designed the study, participated in data collection, analyzed the data, and critically reviewed and revised the manuscript; Ms Birkholz and Dr Todd conceptualized and designed the study, participated in data collection, and critically reviewed and revised the manuscript; Dr Hyman conceptualized and designed the study and critically reviewed and revised the manuscript; Dr Dempsey conceptualized and designed the study, participated in data collection, participated in data analysis, and critically reviewed and revised the manuscript; and all authors approved the final manuscript as submitted.

The contents are the authors’ sole responsibility and do not necessarily represent the official views of the National Institutes of Health.

FUNDING: Supported by National Institutes of Health/NCATS (National Center for Advancing Translational Sciences) Colorado CTSI (Clinical & Translational Sciences Institute) grant UL1 TR001082. Funded by the National Institutes of Health (NIH).

1
Denny
FW
,
Murphy
TF
,
Clyde
WA
 Jr
,
Collier
AM
,
Henderson
FW
.
Croup: an 11-year study in a pediatric practice.
Pediatrics
.
1983
;
71
(
6
):
871
876
[PubMed]
2
Federal Interagency Forum on Child and Family Statistics
. America’s Children: Key National Indicators of Well-Being, 2015.
Washington, DC
:
US Government Printing Office
;
2015
3
Klassen
TP
.
Croup: a current perspective.
Pediatr Clin North Am
.
1999
;
46
(
6
):
1167
1178
[PubMed]
4
Marx
A
,
Török
TJ
,
Holman
RC
,
Clarke
MJ
,
Anderson
LJ
.
Pediatric hospitalizations for croup (laryngotracheobronchitis): biennial increases associated with human parainfluenza virus 1 epidemics.
J Infect Dis
.
1997
;
176
(
6
):
1423
1427
[PubMed]
5
Klassen
TP
.
Recent advances in the treatment of bronchiolitis and laryngitis.
Pediatr Clin North Am
.
1997
;
44
(
1
):
249
261
[PubMed]
6
Petrocheilou
A
,
Tanou
K
,
Kalampouka
E
,
Malakasioti
G
,
Giannios
C
,
Kaditis
AG
.
Viral croup: diagnosis and a treatment algorithm.
Pediatr Pulmonol
.
2014
;
49
(
5
):
421
429
[PubMed]
7
Bjornson
CL
,
Johnson
DW
.
Croup in children.
CMAJ
2013
;
185
(
15
):
1317
1323
8
Russell
KF
,
Liang
Y
,
O’Gorman
K
,
Johnson
DW
,
Klassen
TP
.
Glucocorticoids for croup.
Cochrane Database Syst Rev
.
2011
;
1
:
CD001955
[PubMed]
9
Amir
L
,
Hubermann
H
,
Halevi
A
,
Mor
M
,
Mimouni
M
,
Waisman
Y
.
Oral betamethasone versus intramuscular dexamethasone for the treatment of mild to moderate viral croup: a prospective, randomized trial.
Pediatr Emerg Care
.
2006
;
22
(
8
):
541
544
[PubMed]
10
Donaldson
D
,
Poleski
D
,
Knipple
E
, et al
.
Intramuscular versus oral dexamethasone for the treatment of moderate-to-severe croup: a randomized, double-blind trial.
Acad Emerg Med
.
2003
;
10
(
1
):
16
21
[PubMed]
11
Rittichier
KK
,
Ledwith
CA
.
Outpatient treatment of moderate croup with dexamethasone: intramuscular versus oral dosing.
Pediatrics
.
2000
;
106
(
6
):
1344
1348
[PubMed]
12
Feudtner
C
,
Feinstein
JA
,
Zhong
W
,
Hall
M
,
Dai
D
.
Pediatric complex chronic conditions classification system version 2: updated for ICD-10 and complex medical technology dependence and transplantation.
BMC Pediatr
.
2014
;
14
:
199
[PubMed]
13
Krumholz
HM
,
Wang
Y
,
Mattera
JA
, et al
.
An administrative claims model suitable for profiling hospital performance based on 30-day mortality rates among patients with heart failure.
Circulation
.
2006
;
113
(
13
):
1693
1701
[PubMed]
14
In
H
,
Neville
BA
,
Lipsitz
SR
,
Corso
KA
,
Weeks
JC
,
Greenberg
CC
.
The role of National Cancer Institute-designated cancer center status: observed variation in surgical care depends on the level of evidence.
Ann Surg
.
2012
;
255
(
5
):
890
895
[PubMed]
15
Aronson
PL
,
Thurm
C
,
Alpern
ER
, et al
.
Variation in care of the febrile young infant <90 days in US pediatric emergency departments.
Pediatrics
.
2014
;
134
(
4
):
667
677
16
Wennberg
JE
.
Unwarranted variations in healthcare delivery: implications for academic medical centres.
BMJ
.
2002
;
325
(
7370
):
961
964
[PubMed]
17
Wennberg
JE
.
Practice variation: implications for our health care system.
Manag Care
.
2004
;
13
(
9
suppl
):
3
7
[PubMed]
18
Florin
TA
,
French
B
,
Zorc
JJ
,
Alpern
ER
,
Shah
SS
.
Variation in emergency department diagnostic testing and disposition outcomes in pneumonia.
Pediatrics
.
2013
;
132
(
2
):
237
244
[PubMed]
19
Berwick
DM
,
Hackbarth
AD
.
Eliminating waste in US health care.
JAMA
.
2012
;
307
(
14
):
1513
1516
[PubMed]
20
Conway
PH
,
Keren
R
.
Factors associated with variability in outcomes for children hospitalized with urinary tract infection.
J Pediatr
.
2009
;
154
(
6
):
789
796
[PubMed]
21
Neuman
MI
,
Hall
M
,
Hersh
AL
, et al
.
Influence of hospital guidelines on management of children hospitalized with pneumonia.
Pediatrics
.
2012
;
130
(
5
). Available at: www.pediatrics.org/cgi/content/full/130/5/e823
[PubMed]
22
Todd
J
,
Bertoch
D
,
Dolan
S
.
Use of a large national database for comparative evaluation of the effect of a bronchiolitis/viral pneumonia clinical care guideline on patient outcome and resource utilization.
Arch Pediatr Adolesc Med
.
2002
;
156
(
11
):
1086
1090
[PubMed]
23
Parikh
K
,
Hall
M
,
Teach
SJ
.
Bronchiolitis management before and after the AAP guidelines.
Pediatrics
.
2014
;
133
(
1
). Available at: www.pediatrics.org/cgi/content/full/133/1/e1
[PubMed]

Competing Interests

POTENTIAL CONFLICT OF INTEREST: Dr Dempsey serves on advisory boards for Merck and Pfizer. She does not receive any research funding from these companies. The other authors have indicated they have no potential conflicts of interest to disclose.

FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.

Supplementary data