Objective. To explore the effect of physician training background on the emergency management of croup.
Methods. Two community hospitals with a general emergency department (ED) staffed by board-certified emergency medicine (EM) practitioners were the setting for this study. At both sites, pediatricians (PED) or pediatric emergency medicine specialists (PEM) managed acute pediatric visits during evening and weekend hours. Retrospective patient cohorts (6 months to 6 years) with a primary discharge diagnosis of croup were identified from a 1-year period. Data abstraction was performed by a registered nurse who was blinded to the study hypothesis.
Results. There were 229, 92, and 209 patients in the PED, PEM, and EM cohorts, respectively, reflecting the practice of 69 physicians (19 PED, 12 PEM, and 38 EM). The groups had similar rates of admission and prescription of steroids at discharge. In regression models that incorporated all recorded clinical variables, EM patients were more likely to have received a chest radiograph (odds ratio [OR]: 6.6; 95% confidence interval [CI]: 3.1–14), racemic epinephrine (OR: 6.5; 95% CI: 3.1–14), albuterol in the ED (OR: 3.0; 95% CI: 1.4–6.4), and parenteral steroids (OR: 3.6; 95% CI: 2.1–6.3) and were less likely to have received oral steroids (OR: 0.41; 95% CI: 0.26–0.64). For the EM cohort, adjusted mean length of ED visit was 40 minutes longer (95% CI: 6.8–72) and mean direct costs were $90 higher (95% CI: $27–$153). Regression models comparing the PEM and PED cohorts revealed no significant management differences.
Conclusion. Compared with physicians with a pediatric background, rates of resource utilization were higher for EM-trained physicians who managed uncomplicated cases of croup.
In a variety of community settings throughout the United States, physicians with backgrounds in general emergency medicine (EM), pediatrics (PED), and pediatric emergency medicine (PEM) all manage acutely ill children. It is important to ask whether differences in training background serve as the basis for substantial variations in management. Identification of consistent differences in practice styles among provider groups might ultimately be linked to differences in health care costs and outcomes.1 At the very least, such differences should prompt a closer look at the educational content of such training experiences.
By examining 3 clinically and demographically similar cohorts of children who presented emergently in the community hospital setting, we wanted to isolate the effect of physician training background on practice patterns. We chose 2 sites at which pediatric patients who had received a diagnosis of uncomplicated croup were treated by practitioners with differing types of pediatric and emergency medicine training. We hypothesized that physician training background would emerge as an independent predictor of resource utilization and overall efficiency.
This study took place in 2 nonacademic community hospitals in a single metropolitan area. Both institutions provided emergency services to suburban, middle-class, and working-class populations. Hospital A had an annual emergency department (ED) volume of 35 000, approximately 20% of which were pediatric visits. Hospital B had an annual volume of 45 000, roughly 20% of which were pediatric visits. At both sites, acutely ill children who presented in the early morning and weekday hours were treated in the general ED by board-certified EM physicians.
At both locations, resources, billing, laboratory, nursing, and ancillary staffing were under the management of the general ED. Pediatric physician staffing was managed by a university-affiliated children’s hospital. Under a partnership agreement, the children’s hospital provided both urgent care pediatricians (full- or part-time contracted/salaried, residency-trained, and board-certified or board-eligible) or PEM-trained specialists (defined as fellowship training) to care for acute pediatric patients who presented on evenings and weekends (at Hospital A: 5 pm–1 am weeknights, noon–1 am Saturdays, and 9 am–1 am Sundays; at Hospital B: 4 pm–2 am weeknights, noon–2 am Saturdays, and 9 am–2 am Sundays). Family practice house staff participated in the care of some patients at Hospital A but had no autonomous decision-making authority regarding management decisions.
At each hospital, a computerized medical record search was performed to identify patients who were between 6 months and 6 years of age and had an International Classification of Disease, Ninth Revision, code for “croup” included as their primary discharge diagnosis over calendar year 1999. Patients who were admitted directly to the hospital or through outpatient clinics were excluded.
Each chart was reviewed by a single registered ED nurse who was from one of the participating institutions and blinded to the study hypothesis. Records including any of the following chronic diagnoses were excluded: history of prematurity (<34 weeks), malignancy, static encephalopathy, structural brain disorder, known structural large airway disease (subglottic stenosis, vascular rings, laryngeal masses, vocal cord paralysis, etc), or history of tracheostomy.
Results of each review were recorded on a standardized data abstraction form. Clinical information included additional acute diagnoses; history of croup episodes; current antibiotic, corticosteroid, or albuterol use; hour of presentation; general appearance; age; gender; vital signs; pulse oximetry; and description of stridor. Demographic information included insurance status and ability to identify a primary care provider. Assessment decisions recorded included the performance of radiographic tests, complete blood counts, blood and urine cultures, and serum electrolytes. Interventions recorded included endotracheal intubation and use of aerosolized racemic epinephrine, corticosteroids, aerosolized albuterol, cool mist, and antibiotics. Patient disposition (admission, discharge, or transfer), total length of ED stay, and the name of the physician who cared for the patient were also recorded.
Direct hospital costs for each visit (not charges) were obtained from a centralized computer database at each hospital. At each site, costs for patients seen in the pediatric areas were assigned in a manner identical to those for the general ED. For accounting purposes, the fixed overhead costs of operating the pediatric areas were not distinguished from those of the general ED, resulting in internally valid comparisons of resource utilization across the cohorts. To isolate better the effects of medical decision making on resource utilization, we did not include in any of these calculations the costs of providing physician services.
Data were entered and analyzed in SPSS for Windows, version 10.0 (SPSS, Inc, Chicago, IL). Unless otherwise stated, statistical significance was set at P < .05. Patient clinical and demographic characteristics were compared using χ2 tests for categorical variables and 2-tailed Student t tests or Kruskal-Wallis tests (depending on data distribution) for continuous variables. Initial analysis of the incidence of assessment and management decisions included χ2 tests. Because of their non-normal distribution, mean length of stay and costs were compared using a Kruskal-Wallis test.
To isolate the effect of physician training background on our dependent variables of interest, we constructed regression models. In addition to the training background of the provider, each model contained as independent variables and/or covariates all of the clinical and demographic variables recorded. All of the recorded variables were presumed to have a priori relevance, and none was eliminated from the models, regardless of strength of association. In our logistic regression models, dependent variables included specific assessment and management decisions (incidence of radiographs, use of racemic epinephrine, etc). Linear regression models were used to analyze the continuous variables of length of visit and costs.
After written assurances that patient information would remain coded and anonymous, this retrospective chart review was exempted from formal institutional review board approval at both participating institutions.
A total of 530 records were identified from the study period. Complete records were obtained for all of the identified patients (100%), and no patient met exclusion criteria. A total of 464 records (88%) contained documentation of all of the study variables (a record of pulse-oximetry was the most frequently missing variable). A total of 179 visits (34%) were identified from Hospital A and 351 (66%) from Hospital B. A total of 229 visits (43%) were managed by a board-certified (or board-eligible) pediatrician (PED group), 209 (39%) were managed by a board-certified EM physician (EM group), and 92 (17%) were managed by a PEM fellow or attending physician (PEM group). The study sample reflected the practice decisions of 19 pediatricians (mean: 12 visits), 38 EM physicians (mean: 5.5 visits), and 12 PEM specialists (mean: 7.7 visits).
Table 1 displays the clinical and demographic characteristics of the 3 groups. When the PED and PEM cohorts were combined and compared with the EM cohort, there were no significant differences on any characteristics, with the exception that the EM cohort was more likely to have presented during the overnight hours (odds ratio [OR]: 5.3; 95% confidence interval [CI]: 3.5–8.0). However, when the PEM cohort was compared separately with the other 2 cohorts, these patients were least likely to have commercial insurance (OR: 0.36; 95% CI: 0.19–0.68), to have identified a primary care provider (OR: 0.44; 95% CI: 0.25–0.81), or to be male (OR: 0.52; 95% CI: 0.32–0.83). They were most likely to be febrile (OR: 1.7; 95% CI: 1.0–2.8) and hypoxic (OR: 3.5; 95% CI: 1.3, 9.9).
Table 2 compares the assessment and management decisions made during each visit. No patients in the study required endotracheal intubation. When the EM cohort was compared with the PED and PEM cohorts in this unadjusted analysis, EM patients were significantly more likely to have received a chest radiograph, lateral neck radiograph, cool mist, nebulized albuterol, racemic epinephrine, and intramuscular or intravenous steroids. There were no differences in the use of steroids or antibiotics at discharge or in admission rate. Mean length of visit and direct hospital costs were highest for the EM cohort.
Regression models were constructed to isolate better the effect of training background on each of the outcomes listed in Table 2. Each model was adjusted for all of the clinical variables listed in Table 1. When the EM cohort was compared with the other 2 cohorts, these patients were significantly more likely to have received a chest radiograph (OR: 6.6; 95% CI: 3.1–14), nebulized racemic epinephrine (OR: 6.5; 95% CI: 3.1–14), nebulized albuterol in the ED (OR: 3.0; 95% CI: 1.4–6.4), cool mist (OR: 2.2;95% CI: 1.4–3.5), and parenteral corticosteroids (OR: 3.6, 95% CI: 2.1–6.3) and were less likely to have received oral steroids (OR: 0.41; 95% CI: 0.26–0.64). Adjusted mean length of ED visit was 40 minutes longer (95% CI: 6.8–72), and mean direct costs were $90 higher (95% CI: $27–$153). No significant differences were noted for any other outcome variables listed in Table 2.
Because the presentation of croup demonstrates diurnal variation and EM physicians saw patients during both extremes of daytime and overnight hours, separate regression models (which included all of the above clinical and demographic variables) were constructed to isolate the effect of diurnal variation on the EM cohort. In these models, which included variables such as hypoxia, stridor, and general appearance, hour of presentation failed as an independent predictor of resource utilization within the EM cohort. There was 1 exception: patients who were seen by EM providers during overnight hours were more likely to have received cool mist (OR: 3.2; 95% CI: 1.5–6.8).
Although our univariate analysis (Table 2) suggested differences between the PED and PEM cohorts for some management decisions (chest radiography, use of racemic epinephrine), when our regression models were used to compare the cohorts with each other, subspecialty status was not significantly associated with any of the dependent variables studied.
Our 3 cohorts of croup patients were clinically and demographically similar. All patients were seen in the setting of 2 community hospitals. These conditions limit the effects of case-mix and referral bias and allow us to isolate the effect of physician training background on management styles. Our findings indicate that patients who were cared for by EM-trained physicians were treated in a more resource-intensive and costly manner (ie, more radiographs, bronchodilators, parenteral steroids, and longer visits). However, the ultimate disposition of these patients (discharge medications and admission rate) did not vary significantly, further supporting the notion that variations in illness severity do not explain these management differences.
Comparative studies of practice variation between EM and pediatric providers have been limited.2–5 Our earlier report on the management of simple febrile seizure uncovered a tendency toward the performance of more expensive evaluations in general ED settings.6 More recently, Isaacman et al7 noted differences in the assessment and management of fever without source. However, both of these studies compared patients who were seen in community EDs with those who presented to pediatric EDs at university-affiliated children hospitals. Our current study reduces the possibility of unmeasured clinical and demographic confounders by limiting our analysis to patients who presented to community EDs.
Because the existing literature strongly supports the assumption that uncomplicated cases of croup have a uniformly good prognosis,8,9 it is unlikely that any of the management differences recorded affected ultimate morbidity in this population. For instance, it is generally agreed that the use of racemic epinephrine provides temporary improvement in croup symptoms but has no effect on the course of illness.10–12 Likewise, both oral and parenteral corticosteroids have been shown to have similar efficacy.13
We chose a diagnosis of “croup” to define our study population because we believed that this common pediatric condition typically offers an unambiguous and straightforward case definition.12 It is interesting that the management of PEM subspecialists did not differ from the other pediatricians. This finding suggests that, for this minor acute illness, a common background in pediatrics had a greater impact on practice style than PEM fellowship training.
Potential explanations for the management differences observed may be inferred. The increased reliance on radiographic studies in the EM cohort suggests less confidence in clinical diagnostic skills as the basis for decision making in these cases. The increased use of albuterol likewise may reflect diagnostic uncertainty. Differences in the use of racemic epinephrine (despite a comparable incidence of reported stridor at rest and hypoxia) may point toward a more cautious approach to children with any degree of respiratory distress. The increased use of injectable steroids (versus oral) may indicate a delay in the dissemination of results of studies reported in the pediatric literature into general EM practice.
Our results emphasize statistically significant differences, yet it is important to note that for all of the decisions studied, the majority of providers in both groups behaved similarly (eg, 83% of patients in the EM cohort did not receive chest radiographs). However, our financial comparisons suggest that even these relatively small differences in management affected overall costs.
This was neither a prospective nor a randomized trial. Logistic considerations, as well as a likely Hawthorne effect, precluded such a design. For most patients studied, cohort assignment was dependent on the hour of presentation (and/or day of the week). It was presumed that a crossover hour of 1 am to 2 am should have modulated such systematic bias, and “postbedtime” presentation (10 pm–8 am) was included as an independent variable in our models. Nevertheless, no patient who presented between 2 am and 8 am was treated by a pediatrician. Therefore, it was important to establish that the other clinical variables recorded (hypoxia, general appearance, stridor, etc.) adequately reflected acuity. After controlling for these variations, our failure to find significant day versus night differences within the EM cohort supports the robustness of our models.
Our study is subject to the limitations that apply to all retrospective reviews.14 We were fortunate that no records were missing, and the data were remarkably complete. We minimized observer bias by using a single, well-qualified data abstractor who was blinded to our study hypothesis. Our review included 69 physicians at 2 different institutions, but generalizability to other settings and groups may be limited. Similarly, a relatively small number of “outlier” practitioners may have magnified or distorted apparent differences between the groups.
The care of a minor, uncomplicated pediatric illness such as croup is indisputably within the scope of general EM practice. However, in our study, such practitioners tended to treat these patients more expensively than their pediatric colleagues. Because it is unlikely that these differences improved final patient outcomes, we suggest that constructive efforts be made to enhance the pediatric training and experiences of EM specialists. These may include additional pediatric exposure during residency training as well as regular continuing medical education covering recent reports in the pediatric literature.
We thank Mary Lou Kelly, RN, for performing the chart reviews for this study. Joan Bothner, MD, and Steven R. Poole, MD, provided helpful reviews of this manuscript.
- ↵Wennberg JE. Understanding geographic variations in health care delivery. N Engl J Med.1999;340 :32– 39
- ↵Scribano PV, Baker MD, Ludwig S. Factors influencing termination of resuscitative efforts in children: a comparison of pediatric emergency medicine and adult emergency medicine physicians. Pediatr Emerg Care.1997;13 :320– 324
- Schweich PJ, Smith KM, Dowd MD, Walkley EI. Pediatric emergency medicine practice patterns: a comparison of pediatric and general emergency physicians. Pediatr Emerg Care.1998;14 :89– 94
- Petrack EM, Christopher NC, Kriwinsky J. Pain management in the emergency department: patterns of analgesic utilization. Pediatrics.1997;99 :711– 714
- ↵Nelson DS, Walsh K, Fleisher GR. Spectrum and frequency of pediatric illness presenting to a general community hospital emergency department. Pediatrics.1992;90 :5– 10
- ↵Hampers LC, Trainor JL, Listernick R, et al. Setting based practice variation in the management of simple febrile seizure. Acad Emerg Med.2000;7 :21– 27
- ↵Isaacman DJ, Kaminer K, Veligeti HV, et al. Comparative practice patterns of emergency medicine physicians and pediatric emergency medicine physicians managing fever in young children. Pediatrics.2001;108 :354– 358
- ↵Denny FW, Murphy TF, Clyde WA Jr, Collier AM, Henderson FW. Croup: an 11 year study in pediatric practice. Pediatrics.1983;71 :871– 876
- ↵McEniery J, Gillis G, Kilham H, Benjamin B. Review of intubation in severe laryngotracheobronchitis. Pediatrics.1991;87 :847– 853
- ↵Ledwith CA, Shea LM, Mauro RD. Safety and efficacy of nebulized racemic epinephrine in conjunction with oral dexamethasone and mist in the outpatient management of croup. Ann Emerg Med.1995;25 :331– 337
- Kunkel N, Baker MD. Use of racemic epinephrine, dexamethasone and mist in the outpatient management of croup. Pediatr Emerg Care.1996;12 :156– 159
- ↵Klassen TP. Croup: a current perspective. Pediatr Clin North Am.1999;46 :1167– 1178
- ↵Rittichier KK, Ledwith CA. Outpatient treatment of moderate croup with dexamethasone: intramuscular vs. oral dosing. Pediatrics.2000;106 :1344– 1348
- ↵Gilbert EH, Lowenstein SR, Kozoil-McLain J, Barta DC, Steiner J. Chart reviews in emergency medicine research: where are the methods? Ann Emerg Med.1996;27 :305– 308
- Copyright © 2002 by the American Academy of Pediatrics