Prospective Assessment of Practice Pattern Variations in the Treatment of Pediatric Gastroenteritis
OBJECTIVES: We aimed to determine whether significant variations in the use of intravenous rehydration existed among institutions, controlling for clinical variables, and to assess variations in the use of ancillary therapeutic and diagnostic modalities.
METHODS: We conducted a prospective cohort study of children 3 to 48 months of age who presented to 11 emergency departments with acute gastroenteritis, using surveys, medical record reviews, and telephone follow-up evaluations.
RESULTS: A total of 647 eligible children were enrolled and underwent chart review; 69% (446 of 647 children) participated in the survey, and 89% of survey participants (398 of 446 children) had complete follow-up data. Twenty-three percent (149 of 647 children) received intravenous rehydration (range: 6%–66%; P < .001) and 13% (81 of 647 children) received ondansetron (range: 0%–38%; P < .001). Children who received intravenous rehydration had lower Canadian Triage Acuity Scale scores at presentation (3.1 ± 0.5 vs 3.5 ± 0.5; P < .0001). Regression analysis revealed that the greatest predictor of intravenous rehydration was institution location (odds ratio: 3.0 [95% confidence interval: 1.8–5.0]). Children who received intravenous rehydration at the index visit were more likely to have an unscheduled follow-up health care provider visit (29% vs 19%; P = .05) and to revisit an emergency department (20% vs 9%; P = .002).
CONCLUSIONS: In this cohort, intravenous rehydration and ondansetron use varied dramatically. Use of intravenous rehydration at the index visit was significantly associated with the institution providing care and was not associated with a reduction in the need for follow-up care.
WHAT'S KNOWN ON THIS SUBJECT:
Although gastroenteritis guidelines describe the need to perform oral rehydration, it remains underused, resulting in excessive use of intravenous rehydration. Other interventions, such as antiemetic administration, vary according to location, often resulting in differences in cost and outcomes.
WHAT THIS STUDY ADDS:
In this nationwide cohort, intravenous rehydration use varied dramatically. Use was associated with the institution providing care and an increase in the need for future health care provider visits. Use of ondansetron also varied significantly across Canada.
Gastroenteritis accounts for >1.5 million outpatient visits per year among children in the United States1 and 6% to 7%2,3 of all pediatric emergency department (ED) visits. Although guidelines describe the need for oral rehydration therapy (ORT) for children with mild to moderate dehydration,1,4,5 there continue to be significant variations in ORT use6 and poor acceptance of the guidelines.7 Despite a 96% ORT success rate in clinical trials conducted in developed countries,8 the majority of pediatricians think that moderate dehydration and vomiting preclude the use of ORT.9 The discrepancy between guidelines and practice was evident in a single-center Canadian study that reported moderate-to-severe dehydration was present in only 2% of 205 children with gastroenteritis but 30% received intravenous rehydration.10
Furthermore, other interventions have been demonstrated to vary according to practice location. Rates of antiemetic use, which have increased in the past decade, ranged from 2% to 23% in developed European and North American countries in 2005.11 Three percent of Canadian children with gastroenteritis received dimenhydrinate, whereas >20% of US children received promethazine.11 Similar variations in hospitalization rates for children with gastroenteritis have been reported. In the United States, 16-fold variations in hospitalization rates between centers have been reported,12 which are not accounted for by differences in disease prevalence.13 Similarly, 14-fold variations in hospitalization rates have been documented in Ontario, Canada,14 with rates being positively associated with bed availability.
Because significant practice variations might be linked to differences in costs and outcomes,15 patient-level data are necessary to evaluate which factors affect outcomes for children with gastroenteritis. Therefore, we conducted a prospective, multicenter, cohort study to evaluate the treatment of pediatric gastroenteritis in Canadian EDs. Our primary objective was to determine whether significant variations in the use of intravenous rehydration exist among institutions. Secondary objectives included assessment of variations in the use of ancillary therapeutic and diagnostic modalities.
Study Design, Setting, and Population
We conducted a prospective, 11-center, ED cohort study of consecutive children who presented with acute gastroenteritis during a 7- to 10-day period in spring 2009. The study hospitals reflect a cross-country distribution and include 9 pediatric EDs in pediatric institutions, 1 pediatric ED in a general hospital, and 1 general ED. All study hospitals are members of Pediatric Emergency Research Canada, are staffed by attending physicians 24 hours per day, and provide training to residents and/or fellows. The research ethics boards of all participating institutions approved this study.
All patients presenting to the ED who were 3 to 48 months of age and had acute gastroenteritis, defined on the basis of a history of ≥3 watery stools in a 24-hour period within 72 hours before the ED visit,4,16 were eligible. The total duration of all symptoms must have been ≤7 days. We excluded children who had been enrolled previously and families that were unavailable or unable to complete telephone follow-up evaluations.
All ED staff members except for the site investigators and research assistants remained blinded to the primary outcome of intravenous rehydration use, to minimize the impact of that knowledge.17 The success of the blinding strategy was not evaluated. Trained research assistants recruited patients in the ED, administered a standardized survey to caregivers, provided caregivers with a diary, conducted chart reviews, and completed telephone follow-up evaluations. Research assistant study coverage varied according to site and ranged from 10 to 24 hours per day. Informed consent was obtained from caregivers after the completion of nursing triage. The research ethics board of the participating institutions waived the need for informed consent for the chart review portion of the study for patients who were not approached for participation in the survey.
Children who met all eligibility criteria and had a final diagnosis consistent with gastroenteritis were included in the analysis. Examples of ineligible diagnoses are presented in Fig 1. Telephone follow-up evaluations with the children's caregivers were conducted 2 weeks later. A daily review of ED records identified eligible patients who were missed. After the ED visit, chart reviews were performed for all eligible patients, to obtain information on investigations, treatments, and outcomes. A final hardcopy/electronic chart review was completed at each study site 2 weeks after the ED visit, to confirm parental reports of subsequent ED visits and hospitalizations.
The primary outcome was the proportion of children treated with intravenous rehydration at the participating institutions. Specifically, we sought to determine whether significant variations in the use of intravenous rehydration existed among institutions when patient-level data were used to control for clinical variables. Secondary outcomes included variations in the use of other therapeutic and diagnostic modalities, including investigations, antiemetic treatment, hospitalization, and prescribed medications, and future health care provider visits.
Sample Size Determination
The estimated proportion of children who would receive intravenous rehydration was 14%, on the basis of the 2005 unweighted annual mean from 3 participating sites (S.B.F., J. Fernandes, BA, and K. Farion, MD, personal written communications, 2008). Significant variation was defined as a difference of >5% from the mean (ie, 9%–19%). For regression analyses with our anticipated number of outcomes and the ability to include site as a variable, the 11 institutions were grouped according to geographic location for preliminary analyses. This approach was on the basis of previous evidence of significant variations in the appropriateness of the use of specific interventions for other disease processes between provinces.18 A sample size of 131 subjects per group was needed to detect a difference between the 2 regions with 80% power at the 5% level of significance. Possible confounding was adjusted for by increasing the sample size by 30%, which represents the estimated correlation between the predictor variable (ie, site) and the other variables in the model.19 Therefore, the final necessary sample size was 290 subjects.
Descriptive statistics were used to describe patient characteristics, clinical symptoms, treatment and management practices, and primary and secondary outcomes. Dichotomous variables are expressed as proportions, and continuous variables are expressed as means and SDs. Baseline characteristics were compared by using the χ2 test or Fisher's exact test for proportions and the t test or analysis of variance for continuous variables, as appropriate. A logistic regression, multiple-predictor model was constructed to identify predictors of the primary (dependent variable) outcome of intravenous rehydration. Independent variables included in the model were those with P values of ≤.2 and biological plausibility.20 Candidate variables included age, weight, diarrhea and vomiting frequency, duration of fever, triage temperature, heart rate, respiratory rate, Canadian Triage Acuity Scale score, presence of decreased oral intake, energy or urine output, previous physician assessments, or referral to the ED. Additional candidate variables included caregiver reports of the presence of hematemesis (“When your child vomits, is there blood or small black flecks that look like coffee grinds?”), hematochezia (“Has your child had blood in their stools [poo]?”), or bilious emesis (“When your child vomits, is it dark green?”). Independent variables that were neither statistically associated with the primary outcome nor confounders were assessed for interaction effects, which were corrected for before exclusion from the model. After all appropriate, statistically significant, independent variables (geographic location, child's age, number of vomiting and diarrhea episodes in the preceding 24 hours, triage heart rate, previous physician visit, bilious or bloody vomitus, and decreased oral intake) had been identified, relevant adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were calculated. Because of limitations in the numbers of observed events, the secondary outcomes of antiemetic treatment, probiotic treatment, hospitalizations, investigations performed, and health care use after discharge did not have logistic models constructed. SAS 9.1 (SAS Institute, Cary, NC) was used for statistical analyses. Significance was set at the .05 level.
To verify that the model obtained fit the data well, a cross-validation procedure was conducted by using two-thirds of the data as the training set and the remaining one-third as the test group. This procedure is one of several possible approaches to estimate how well a model performs with unseen data. The training set provided regression parameters, which then were used to predict the outcome of intravenous rehydration in the test group. The sensitivity and specificity of the prediction based on the regression model were calculated for both the training and test groups, to assess how well the derived model fit unseen data.
A total of 656 children participated in the study and underwent chart review. Nine (2%) ultimately did not meet inclusion criteria, which left 647 eligible participants. Sixty-nine percent of subjects (446 of 647 subjects) were enrolled prospectively and completed the survey; 201 children (31%) participated only in the retrospective chart review (Fig 1). Eighty-nine percent of eligible participants (398 of 446 participants) were successfully contacted for follow-up evaluations. For the remaining 11% of eligible participants (48 of 446 participants), only chart reviews were completed. The mean age of participants was 21 months (Table 1). Of subjects who had seen a physician before presenting to the ED, 12% (24 of 197 subjects) were given an antibiotic, 9% (17 of 197 subjects) dimenhydrinate, and 0.5% (1 of 197 subjects) a probiotic.
Emergency Department Investigations and Management
Twenty-three percent of study children (149 of 647 children) received intravenous rehydration (range for hospitals: 6%–66%). In bivariate analyses, children who received intravenous rehydration had experienced more episodes of vomiting and diarrhea in the preceding 24 hours and had a greater duration of fever, higher temperature, and higher heart rate at triage, as well as lower Canadian Triage Acuity Scale scores (ie, higher acuity) (Table 2). Logistic regression analyses revealed that the following variables predicted the outcome of intravenous rehydration: institution location (OR: 3.0 [95% CI: 1.8–5.0]), vomiting bile or blood (OR: 2.6 [95% CI: 1.2–5.5]), previous physician visit (OR: 1.7 [95% CI: 1.0–2.7]), and number of vomiting episodes in the preceding 24 hours (per additional vomiting episode, OR: 1.1 [95% CI: 1.0–1.1]). The proportions of children who received intravenous rehydration differed among institutions (P < .001) (Fig 2). In the final regression model, with analysis according to individual institution, site remained significant. The site with the median rate (14%) of intravenous rehydration served as the base case for the calculation of ORs. Subsequent analyses revealed that individual sites predicted the use of intravenous rehydration, with ORs ranging from 0.09 (95% CI: 0.02–0.43) to 6.0 (95% CI: 1.7–20.7). In our cross-validation analysis, we found the sensitivity and specificity of the model to be similar for the learning (sensitivity: 60%; specificity: 75%) and test (sensitivity: 61%; specificity: 72%) groups.
Ondansetron was administered to 13% of children (81 of 647 children), and its use also varied significantly among institutions (range: 0%–38%; P < .001) (Fig 2). It was administered more frequently by physicians who had completed a pediatric emergency medicine fellowship (29 [16%] of 184 physicians), compared with those who had not (17 [9%] of 189 physicians; P = .05). The volume of intravenous fluids administered before disposition determination among children given intravenous rehydration (n = 149) also varied according to site (P = .003), with a range of 15 mL/kg (site 9) to 87 mL/kg (site 11). There was a direct correlation between the absolute number of children given intravenous rehydration at each site and the volume administered (r = 0.60; P = .05) (Fig 3). Five percent of children (30 of 647 children) were hospitalized; this outcome varied among institutions (range: 0%–10%; P = .04) (Table 3) and was associated with the frequency of vomiting in the 24 hours before ED presentation.
There was significant variation among institutions in the performance of complete blood counts and electrolyte measurements (P < .001) (Table 4). The rates of stool culturing (3% overall; P = .50), stool virologic testing (4% overall; P = .31), and Clostridium difficile toxin testing (1% overall; P = .66) did not differ significantly between sites, whereas rates of testing for ova and parasites did (range: 0%–13%; P = .01).
Discharge medications varied according to site, with significant outliers for many options. A single institution discharged 42% of children (13 of 31 children) with instructions for administration of ondansetron; no other institution used this strategy (P < .001). Probiotics were recommended at 4 sites, and rates of use ranged from 0% to 17% (P < .001). Physicians did not recommend the use of herbal medications or Pepto-Bismol (Procter & Gamble, Cincinnati, OH); dimenhydrinate was recommended for 4 children (1%) and Kaopectate (Chattem, Inc, Chattanooga, TN) for 1 child (0.2%).
Rates of repeat health care provider visits varied between institutions (range: 5%–38%; P = .008). Of the children who did require a repeat ED visit, 28% (13 of 47 children) received intravenous rehydration. The rate of intravenous rehydration at the repeat visit was not significantly higher than the rate at the index visit (P = .29). In addition, children who received intravenous rehydration at the index visit were more likely to have an unscheduled health care provider visit in the subsequent 2 weeks (27 [29%] of 94 children vs 62 [19%] of 321 children; P = .05) and to revisit an ED (19 [20%] of 94 children vs 28 [9%] of 321 children; P = .002).
To the best of our knowledge, this is the first nationwide prospective study to collect patient-level data describing the treatments administered to children with gastroenteritis. Our results demonstrate that intravenous rehydration use remains common in Canada and varies between EDs, even with adjustment for clinical variables. The most significant sources of variation seem to be the institution where the child is treated, the presence of bile or blood in the vomitus, a previous physician visit, and the number of vomiting episodes in the preceding 24 hours. Significant variation also exists in other key parameters, including the use of ondansetron, laboratory investigations, and discharge medications.
Although still not in line with published guidelines,1,4 the rate of intravenous rehydration use in our cohort was lower than the rate reported (43%) in a prospective, multicenter, European cohort21 and is in keeping with practices at 2 academic institutions within the United States.22 Although those studies detected site-to-site variations, they did not include prospectively collected clinical severity data.
In gastroenteritis, the observed variations seem to be related to a gap between evidence-based guidelines and clinical practice.23,24 In fact, 75% of pediatric emergency medicine fellowship directors reported that ORT requires additional time, compared with intravenous rehydration.25 Other barriers cited include a lack of convenience, support staff preference for intravenous rehydration, and the need to train support staff members regarding ORT administration.9
Guidelines can succeed when they are developed locally by the people who will use them and are introduced as part of a specific educational intervention.26 Greater familiarity with guidelines is associated with improved physician behavior, which implies that knowledge translation may improve compliance.23,27 The implementation of ORT protocols can reduce admissions (from 23% to 5%) and hospital stay (from 7.9 hours to 2.3 hours), as evidenced at an urban ED in Australia.28 A Medicaid network using a process that included educational sessions, patient materials, performance feedback, conference calls, and toolkits reduced the gastroenteritis admission rate by >40%.29 This multimodal approach overcame a major barrier to the use of ORT, namely, parental preference. The latter is significant because, when given the opportunity to decide, 62% of caregivers chose intravenous rehydration over ORT.30 However, the availability of an effective, orally administered antiemetic might make them choose ORT instead.30
Dehydration has been identified as the most common diagnosis resulting in ED revisits, as well as revisits requiring hospitalization.31 However, this fact is underrecognized by health care providers.31 We speculate that, when health care providers treat children whom they assess as being at risk for a revisit, they are more likely to administer intravenous rehydration in an attempt to “fill the tank,” a process thought, albeit without supporting evidence, to prevent future health care provider visits. However, we found the opposite; children who received intravenous rehydration at the index visit were more likely to have an unscheduled health care provider visit and were twice as likely to revisit an ED. Although no other studies addressed this question directly, a trend toward fewer ED revisits among children who received ORT (17%), compared with those who received intravenous rehydration (24%), was reported previously.32 Those findings are comparable to the results obtained in this study.
Although at the time of disposition children who received intravenous rehydration achieved rehydration status comparable to that of their counterparts who received ORT, their baseline severity levels might have been different. Our participants who received intravenous rehydration reported significantly greater frequency of vomiting, diarrhea, hematochezia, and bilious or bloody vomitus. They also were determined through triage to be more unwell and to have higher temperatures, heart rate, and Canadian Triage Acuity Scale scores. Their caregivers more-frequently reported decreased intake, urine output, and energy. Whether this scenario persists after discharge is not known presently. The alternative explanation is that the increased revisit rate for children who received intravenous rehydration may reflect a psychological response by the caregivers. Because of the invasive intervention received, these parents may perceive their children as sicker and the initial ED visit as more necessary, compared with the caregivers of patients who received ORT. They also might have less confidence in the likely success of ORT.
Although we achieved complete capture of all eligible subjects, the short sampling time frame is a potential limitation. Enrollment during a consistent 1-week time frame in the spring was chosen purposefully, to correspond to recent evidence demonstrating that the rotavirus peak has shifted toward the middle of April33 and to minimize the interaction between season and practice patterns at any individual ED. Routine stool testing to rule out the possibility of an unequal distribution of etiologic agents across sites was not performed. Although we included clinical factors in our models, we cannot exclude the possibility that differences in etiologic agents might have contributed to our findings. Although the use of site-specific interviewers and data abstractors was necessary and they used standardized data collection forms, we cannot rule out the possibility that this introduced an element of bias. Lastly, interobserver reliability in data extraction was not assessed.
Our findings indicate that there are significant practice variations in the treatment of acute gastroenteritis in Canadian EDs. The use of intravenous rehydration, laboratory investigations, ondansetron, and probiotics varied significantly according to site. Although children who received intravenous rehydration at the index visit were more unwell than those who received ORT, our finding of increased numbers of subsequent health care and ED visits among children who received intravenous rehydration remains concerning. Future research into both the role of probiotics and the association between intravenous rehydration and subsequent health care use is warranted.
The participating study institutions included the following: British Columbia Children's and Women's Hospital (Vancouver, Canada), Alberta Children's Hospital (Calgary, Canada), Children's Hospital of Western Ontario (London, Canada), Hospital for Sick Children (Toronto, Canada), Hotel Dieu Hospital (Kingston, Canada), Children's Hospital of Eastern Ontario (Ottawa, Canada), Montreal Children's Hospital (Montreal, Canada), University Hospital Center St Justine (Montreal, Canada), Centre Hospitalier Universitaire de Québec (Quebec City, Canada), Izaak Walton Killam Health Centre (Halifax, Canada), and Janeway Children's Health and Rehabilitation Centre (St John's, Canada).
This study could not have been accomplished without the help of research coordinators at all of the participating sites: Marika Anthony-Shaw (Montreal Children's Hospital), Deborah Atchison (Hospital for Sick Children), Lynell Aucoin (Alberta Children's Hospital), Sarah Bocking (Montreal Children's Hospital), Adrianna Breen (Hotel Dieu Hospital), Chantelle Clarkin (Children's Hospital of Eastern Ontario), Annie-Claude Danjou (Centre Hospitalier Universitaire de Québec), Eleanor Fitzpatrick (Izaak Walton Killam Health Centre), Debbie Harnum (Memorial University of Newfoundland), Syed Rafat Ali Jaffri (Hospital for Sick Children), Laura Keating (Hospital for Sick Children), Anne-Marie Krancevic (Hospital for Sick Children), Cindy Langford (London Health Sciences Centre), Nathalie Morin (University Hospital Center St Justine), Jackie Fiander (Memorial University of Newfoundland), Ann Marie White (University of British Columbia), and Janie Williamson (Alberta Children's Hospital). We also thank all of the volunteers who assisted with patient identification and recruitment at all of the participating sites.
- Accepted November 8, 2010.
- Address correspondence to Stephen B. Freedman, MDCM, MSc, Hospital for Sick Children, Division of Pediatric Emergency Medicine, 555 University Ave, Toronto, ON, Canada, M5G 1X8. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
- ED =
- emergency department •
- OR =
- odds ratio •
- ORT =
- oral rehydration therapy •
- CI =
- confidence interval
- Guarino A,
- Albano F,
- Ashkenazi S,
- et al
- 5.↵Canadian Paediatric Society. Oral rehydration therapy and early refeeding in the management of childhood gastroenteritis. Paediatr Child Health. 2006;11(8):527–531
- Kleinman RE
- Gavin N,
- Merrick N,
- Davidson B
- Reis EC,
- Goepp JG,
- Katz S,
- Santosham M
- Goldman RD,
- Friedman JN,
- Parkin PC
- 16.↵World Health Organization. The Treatment of Diarrhoea: A Manual for Physicians and Other Senior Health Workers. 4th rev.Geneva, Switzerland: WHO Press; 2005
- Kennedy J,
- Quan H,
- Ghali WA,
- Feasby TE
- Chow SC,
- Shao J,
- Wang H
- Hosmer DW,
- Lemeshow S
- Forster J,
- Guarino A,
- Parez N,
- et al
- Yee EL,
- Staat MA,
- Azimi P,
- et al
- Ozuah PO,
- Avner JR,
- Stein RE
- Boyd R,
- Busuttil M,
- Stuart P
- Zolotor AJ,
- Randolph GD,
- Johnson JK,
- et al
- Freedman SB,
- Powell E,
- Seshadri R
- Tate JE,
- Panozzo CA,
- Payne DC,
- et al
- Copyright © 2011 by the American Academy of Pediatrics