Unadjusted annual proportions of children administered ondansetron, administered intravenous (IV) rehydration, and hospitalized, and the mean numbers of emergency department visits per site contributing data during a given year (shown as mean number of patients).
Adjusted change in intravenous (IV) rehydration rate (A), admission rate (B), and 3-day emergency department (ED) revisit rate (C) by site between low and high ondansetron use periods. All models were adjusted for age, sex, race, season, admission time, prior acute gastroenteritis visit, laboratory testing, and diagnostic imaging. Error bars indicate 95% CIs.
eAppendix. Data Quality Issues
eTable 1. Inclusion and Exclusion Criteria for the Acute Gastroenteritis Study Population
eTable 2. Billing Codes Employed in the Definition of Intravenous Rehydration
eTable 3. Specific Alternative Diagnoses Identified by ICD-9 Codes A Priori Identified as Markers for Missed Diagnoses at Index Visit
eTable 4. Demographics of Children With Gastroenteritis at Sites Contributing Emergency Department Data to PHIS Database – Grouped According to Site Eligibility
eTable 5. Comparison of Changes in the Percent of Outcomes Occurring Between Ondansetron Usage Periods (Low vs Medium vs High) Among Children 6 Months to 18 Years of Age (n?=?89?250)
Freedman SB, Hall M, Shah SS, Kharbanda AB, Aronson PL, Florin TA, Mistry RD, Macias CG, Neuman MI. Impact of Increasing Ondansetron Use on Clinical Outcomes in Children With Gastroenteritis. JAMA Pediatr. 2014;168(4):321-329. doi:10.1001/jamapediatrics.2013.4906
Ondansetron hydrochloride use in children with gastroenteritis is increasing rapidly; however, little is known about its impact on outcomes.
To determine whether increasing emergency department ondansetron use has resulted in a reduction in intravenous rehydration rates.
Design, Setting, and Participants
Retrospective observational analysis of eligible visits included in the Pediatric Health Information System administrative database. Eligible institutions included 18 emergency departments geographically distributed across the United States, and participants included 804 000 patients aged 0 to 18 years who were diagnosed as having gastroenteritis in an emergency department at an eligible participating institution between January 1, 2002, and December 31, 2011.
The presence or absence of oral ondansetron administration was identified for each patient through database review. Visits were categorized based on institutional ondansetron use: low (<5% administered ondansetron), medium (5%-25%), or high (>25%).
Main Outcomes and Measures
We conducted hospital-level analyses of the associations between ondansetron use and 3 outcomes: intravenous rehydration (primary), hospitalization, and emergency department revisits within 3 days. Time-series regression models were used, adjusting for demographic characteristics, laboratory testing, diagnostic imaging, and rotavirus infection.
A total of 804 000 eligible patient visits were identified. Oral ondansetron use increased from a median institutional rate of 0.11% (interquartile range, 0.04%-0.44%) of patient visits in 2002 to 42.2% (interquartile range, 37.5%-49.1%) in 2011 (P < .001). Intravenous rehydration was administered to 43 413 of 232 706 children (18.7%) during the low ondansetron period compared with 59 450 of 334 264 (17.8%) during the high ondansetron period (adjusted percentage change = −0.33%; 95% CI, −1.86% to 1.20%). During the transition from low to high ondansetron use, we observed no change in the hospitalization rate (adjusted percentage change = −0.33%; 95% CI, −0.95% to 0.29%), but emergency department revisits decreased (adjusted percentage change = −0.31%; 95% CI, −0.49% to −0.13%). The change in all 3 outcomes varied widely between low and high ondansetron use categories at an institutional level. Oral ondansetron was provided to 13.5% (95% CI, 13.3% to 13.7%) of children administered intravenous rehydration.
Conclusions and Relevance
Although ondansetron use increased during the study period, intravenous rehydration rates were unchanged. Most children administered intravenous fluids did not receive oral ondansetron. Our findings highlight the need to focus efforts to administer ondansetron to children at greatest risk for oral rehydration failure.
Acute gastroenteritis (AGE) remains among the most common illnesses for which children receive emergency department (ED) care, accounting for more than 1.7 million ED visits each year.1 Treatment of AGE is dictated by data supporting the equivalence of oral rehydration therapy (ORT) and intravenous (IV) rehydration in moderately dehydrated children.2,3 Despite evidence supporting the use of ORT, it remains underused, with 45% of physicians preferring IV rehydration over ORT in children with moderate dehydration.4,5
Vomiting, a key factor limiting the successful use of ORT,6 is present in approximately 70% of children with AGE.7 To potentially overcome the impact vomiting has on the effectiveness of ORT, clinical trials have evaluated the ability of ondansetron hydrochloride to optimize ORT success.8- 10 Studies have demonstrated efficacy of ondansetron in children with mild to moderate dehydration who had initially failed ORT9,11 and in those with multiple episodes of vomiting prior to ED presentation.12 Although now administered to more than 50% of children with AGE in US EDs,13 its administration is not recommended in guidelines endorsed by the American Academy of Pediatrics.14 There are concerns that indiscriminate use may result in increased ED revisits, rates of misdiagnosis,15 and health care costs. As ondansetron use has increased rapidly,15 an analysis of its effectiveness and potential for undesirable consequences is warranted.16
Although randomized clinical trials (RCTs) have demonstrated the efficacy (ie, the beneficial effects under optimal conditions) of ondansetron use, its clinical effectiveness (ie, the effects under real-world conditions)17 has only been evaluated in a single-center study where use remained low.10 Thus, we conducted a multicenter retrospective cohort study to determine whether increasing ondansetron administration to children with AGE is associated with a concomitant decline in IV rehydration use.
Data for this time-trend, multicenter, retrospective cohort study were obtained from the Pediatric Health Information System (PHIS). Participating hospitals, which provide resource utilization data, are geographically dispersed across the United States.13 Individual institutions provide deidentified data, along with encrypted medical record numbers enabling the tracking of individual patients across hospital visits. The Children’s Hospital Association, Overland Park, Kansas, and the participating hospitals jointly ensure data quality.18 The Boston Children’s Hospital Institutional Review Board approved this study with a waiver of informed consent given the study’s retrospective design and large sample size.
Children who were younger than 18 years, treated in a participating ED between January 1, 2002, and December 31, 2011, and diagnosed as having AGE based on their International Classification of Diseases, Ninth Revision primary discharge diagnosis code (eTable 1 in Supplement)18,19 were eligible. Children with a primary diagnosis code of dehydration (276.51) were eligible if they had a secondary diagnosis code for AGE. Children transferred to the participating institution and those who had a chronic comorbid condition (eg, malignant neoplasm, sickle cell disease, epilepsy)19 were excluded. Hospital eligibility required evidence of progression through an a priori–defined range of ondansetron use (see Study Definitions and Participant Characteristics). Hospitals that do not contribute ED data and those that failed a standardized data quality evaluation were excluded (eAppendix in Supplement).
The primary outcome was the proportion of children administered IV rehydration. Secondary outcomes included the proportion of children hospitalized and the proportion experiencing an ED revisit within 3 days. We also evaluated the proportion of children assigned a significant alternative diagnosis within 3 days and the changes in costs over time. Analyses focused on differences between low and high ondansetron use categories, with the medium use category serving as an intervention introduction period enabling the conduct of an interrupted time-series analysis.
The primary exposure was the administration of oral ondansetron. Receipt of IV rehydration was determined using billing codes (eTable 2 in Supplement). The exposure and outcomes at each hospital were grouped into 3-month periods to conduct time-series analysis. Because ondansetron use varied by institution, instead of using an aggregated chronological approach, each hospital served as its own control and was evaluated through progression of increasing ondansetron use. Thus, to reduce variability across hospitals and focus on the impact of increasing ondansetron use, hospitals were included only if they progressed through periods of low (<5% of patients with AGE receiving ondansetron) to medium (5%-25%) to high (>25%) ondansetron use. Category cut points were selected based on the following: (1) extremely limited ondansetron use (<5%) prior to 2000; (2) the number needed to treat to prevent 1 child from receiving IV rehydration (number needed to treat = 5; 95% CI, 4-8)8; and (3) knowledge that efficacy studies may overestimate effectiveness by incorporating features maximizing the chance of finding a difference in the shortest possible time at the lowest possible cost. Hence, a significant increase in use may be required to see a clinically significant reduction in IV rehydration rates.20 We determined a priori that a 5-fold increase in ondansetron use would be required to adequately evaluate changes in the outcomes of interest. This would ensure adequate variation in ondansetron use and enable the establishment of a lag period, which facilitates the detection of a change in outcomes in time-series designs.10
Hospitalization was defined as admission to an observation (ie, short stay, 24 hours) or inpatient unit. For revisit outcomes, a 3-day window, generally accepted as an ED quality of care metric,21,22 was used because deterioration caused by persistence or worsening AGE symptoms or that due to an alternative diagnosis is expected to manifest within this period. Alternative diagnoses were specified a priori based on prior research (eTable 3 in Supplement).15
Bivariate analyses were used to explore the relationships between potential covariates and ondansetron use categories (ie, low vs high) with the Kruskal-Wallis test for continuous variables, χ2 test for nominal multicategorical variables, and Cochran-Armitage trend test for binary variables. Costs were adjusted to 2012 US dollars using the medical component of the Consumer Price Index and are presented as index visit costs as well as bundled (index visit ± revisit ± revisit with admission). We constructed piecewise linear regression on the mean number of AGE ED visits per participating site with a break point at 2006 to determine changing trends in volumes. We selected 2006 because this year represented the point at which the overall ondansetron use rate (Figure 1) reached the a priori–defined moderate use range.
Multiple time-series regression models23 were used to assess the associations, at the hospital level, of ondansetron use category with the primary and secondary outcomes adjusting for potential confounding by observed baseline characteristics identified in bivariate analysis (ie, age, sex, race, season, admission time, prior visit for AGE, laboratory testing, diagnostic imaging, and rotavirus diagnosis), including quadratic terms when significant. Models were constructed by hospital and also in aggregate by weighting the proportional contribution of each institution to the total number of encounters.24 We assessed the likelihood, expressed in terms of percentage change, of experiencing the outcomes of interest based on an institution’s ondansetron use category. A multiple time-series approach was used, with adjustments made for observed differences in patient population characteristics, to reduce the potential for bias from unmeasured confounders. Because the transition from low to medium to high ondansetron use category occurred at different times for all institutions (range, October 2003 to October 2010), this approach also adjusts for changes in outcomes over time that may be due to other external unmeasured factors (eg, introduction of rotavirus vaccination). Analyses were repeated excluding children younger than 6 months to remain consistent with the inclusion criteria used in earlier RCTs.9,12
All aggregated analyses were clustered by hospital to account for the decreased variability within hospitals compared with between hospitals and to enable us to better understand the changes in outcomes between ondansetron use categories (ie, low to high) at the institutional level. Statistical analyses were performed using SAS version 9.3 statistical software (SAS Institute, Inc). P < .05 was considered statistically significant; missing values were not imputed (institutions with data quality issues were excluded).
Forty-two freestanding children’s hospitals contributed data to PHIS during the study period. Ten were excluded owing to data quality issues (eAppendix in Supplement) and 14 institutions had an inadequate range of ondansetron use; 4 lacked a low use period, 7 lacked low and medium use periods, and 3 did not have a high use period. The final cohort included 18 institutions and 804 000 patient visits. The demographic characteristics of children at these institutions did not materially differ from excluded sites (eTable 4 in Supplement). Children in the high (>25%) ondansetron use category were older, less likely to have a prior ED visit for AGE during the preceding 7 days, and more likely to have a diagnostic imaging study performed (Table 1).
Although there was a modest increase in the mean number of AGE ED visits per participating site between 2002 and 2006 (P = .05), the number of visits stabilized between 2006 and 2011 (P = .98) (Figure 1). Overall, oral ondansetron use increased from a median institutional rate of 0.11% (interquartile range, 0.04%-0.44%) of patient visits in 2002 to 42.2% (interquartile range, 37.5%-49.1%) in 2011 (P < .001). At an institutional level, use ranged from 1.3% in the low ondansetron use category to 40.5% in the high category, representing a 32-fold increase (P < .001) (Table 2). Overall, IV rehydration was administered to 18.5% (95% CI, 18.4%-18.6%) of children. While only 13.5% (95% CI, 13.3%-13.7%) of those administered IV rehydration received oral ondansetron, 54.1% (95% CI, 53.8%-54.3%) received IV ondansetron.
The IV rehydration rates were lower during the high ondansetron use period compared with the low ondansetron use period (17.8% vs 18.7%, respectively; mean difference = −0.9%; 95% CI, −0.7% to −1.1%) (Table 2). However, patients in the high ondansetron use category had a higher rate of hospitalization compared with children in the low ondansetron use category (percentage change = 0.7%; 95% CI, 0.6% to 0.8%). Adjusted median costs were higher in the high ondansetron use period compared with the low use period (P < .001) but were greatest during the medium use period. Further, children presenting during the high ondansetron use period were less likely to return to the ED within 3 days after discharge (percentage change = −0.7%; 95% CI, −0.6% to −0.8%) and to have a revisit associated with hospitalization (percentage change = −0.4%; 95% CI, −0.4% to −0.5%). Among discharged children, the proportion returning with a significant alternative diagnosis was 0.1% in all periods.
Adjusted time-series analyses revealed that increasing ondansetron use was not associated with reductions in the rates of IV rehydration or hospitalization (Table 3). High ondansetron use was associated with a lower rate of 3-day ED revisits compared with low use (mean difference = −0.4%; 95% CI, −0.6 to −0.2). In absolute numbers, this represents a reduction of 1337 revisits (95% CI, 668 to 2006) in our cohort of 334 264 ED visits. Excluding children younger than 6 months (n = 89 250), high ondansetron use was associated with lower IV rehydration rates compared with low ondansetron use (percentage change = −0.27%; 95% CI, −0.45% to −0.09%) (eTable 5 in Supplement).
The percentage change in the IV insertion, hospitalization, and 3-day ED revisit rates between low and high use categories at an institutional level revealed that the relationship between ondansetron use and each outcome varied by institution (Figure 2).
In this multicenter study, we observed a dramatic increase in the rate of ondansetron administration in children with AGE during a 10-year period. In our adjusted time-series analysis, the increase in ondansetron administration in pediatric EDs was not associated with reductions in the rates of IV rehydration or hospitalization. However, some institutions did experience a significant reduction in IV rehydration use with increasing ondansetron administration. Because the vast majority (87%) of children administered IV rehydration did not receive oral ondansetron, the lack of an identified beneficial association may reflect an ondansetron use pattern that differs from the intended target population (eg, children who are most likely to fail ORT) or the excessive use of our target outcome (IV rehydration).
Our findings must be interpreted in the context of findings from RCTs, which found a 20% absolute reduction in the use of IV rehydration with the use of oral ondansetron (number needed to treat = 5).8 Our study, in contrast, did not identify a reduction. The reasons for our discrepant findings are likely multifactorial and may highlight the distinction between efficacy and effectiveness. To date, ondansetron efficacy trials have used rigorous research designs as well as stringent patient eligibility criteria and ORT protocols. One must weigh the strength of high internal validity associated with an RCT design against the potential weakness in external validity or generalizability. In the real world, where many ED physicians prefer to use IV rehydration in children with moderate dehydration, it is unclear whether ondansetron is being administered with the same intent and to similar patients as those participating in the RCTs. We demonstrate that ondansetron use has become more frequent, and those administered IV rehydration are more often administered IV ondansetron and are not given the opportunity to benefit from oral ondansetron in conjunction with ORT.
Another possible explanation for the lack of an association between increasing ondansetron use and a reduction in IV use is the general trend over time for increased resource utilization by ED physicians. This may include a trend for increasing use of diagnostic testing (eg, laboratory studies or diagnostic imaging)25 and treatments26 over time, which in and of themselves may reflect increasing patient complexity or acuity of illness. In nationally representative samples of ED visits, there was a 60% increase in the use of diagnostic imaging between 2001 and 200525 and an 84% increase in the rate of antibiotic administration to patients with pneumonia from 1993 to 2008.26 Because IV rehydration is often performed in conjunction with and while awaiting the results of serum electrolyte tests in children with AGE, a trend toward the increasing use of such diagnostic tests may offset any potential benefit associated with ondansetron administration. Supporting this notion is the fact that diagnostic imaging increased by 34% in our study population during the 10-year study period. By comparison, although the performance of blood testing increased by 10% during the study period, the unadjusted rate of IV rehydration decreased by 5%. While the increased use of diagnostic imaging continued between medium and high ondansetron use categories, as ondansetron use approached 20%, the IV rehydration rate plateaued. However, ondansetron administration has continued to increase without evidence of further improved outcomes, and excessive use is not without potential consequences in terms of both cost and adverse effects.27
The increase in costs observed in the medium and high ondansetron use periods is not surprising given its cost and the lack of reduction in the use of IV rehydration. This contradicts earlier economic evaluations conducted under the assumption of clinical efficacy, which concluded that ondansetron use was cost-effective.1,23 In light of the limited benefits realized, the increase in diagnostic testing that occurred in the high ondansetron use period and the rising costs of health care in the United States (increase from 14% of national economy in 2000 to 18% in 2010), our results are not unexpected.28 On a positive note, during the high ondansetron use period, there was a reduction in 3-day ED revisits. This finding, which has not been incorporated into our economic model, has been identified in a prior retrospective cohort study10 but not in meta-analyses.8
Our findings must be considered in the context of the population studied. While prior studies demonstrated the efficacy of ondansetron,8 they primarily demonstrated benefit with oral ondansetron in children aged 6 months and older. Similarly, when we analyzed only children aged 6 months and older, while of minimal clinical significance, we did document a small reduction in IV rehydration use associated with increased ondansetron administration. While we identified a high rate of IV ondansetron use, this method of administration has previously been shown to not reduce hospitalization rates29; our findings were in keeping with the results of this RCT.
Because the PHIS database lacks detailed clinical information, we cannot discern whether the intensity of vomiting or the severity of dehydration differed between ondansetron use categories or whether ondansetron was being used with an unmeasured objective in mind (eg, nausea reduction). In addition, because IV fluids are overused in many institutions, the effectiveness of ondansetron administration may be better measured using outcomes that are not available in administrative databases such as the frequency of vomiting, tolerance of ORT, and ED length of stay. Despite a lack of uniform benefit in outcomes, many institutions did experience a decrease in IV rehydration and hospitalization during the study period; some achieved a 10% reduction in IV rehydration rates. These observations serve as a reminder that institutions must evaluate their individual care patterns and outcomes to optimize the care they provide.
Despite our attempts to adjust for severity of illness, residual confounding by indication may exist. Therefore, we adjusted the relationship of an institution’s ondansetron use period to other testing including laboratory studies and diagnostic imaging. If residual confounding by indication were present, we would expect patients who presented during periods in which greater severity of illness or diagnostic uncertainty existed to be more likely to undergo greater testing. Further, there exists the potential for misclassification (ie, inclusion or exclusion) of patient visits by our use of International Classification of Diseases, Ninth Revision codes; however, misclassification is likely nondifferential and not any more likely at one hospital over another or during specific periods. Lastly, our findings may not be generalizable to non–children’s hospital settings, as PHIS data are collected only from freestanding pediatric hospitals.
Ondansetron use in children with AGE has increased dramatically in pediatric centers during the past decade without a concomitant reduction in IV rehydration or hospitalizations. However, more than 85% of patients who received IV rehydration for AGE were not given oral ondansetron, suggesting an opportunity to better target its use to children at greatest risk for failing ORT. Although tangible benefits associated with ondansetron use in our cohort may have been masked by the general trend of increased diagnostic testing and treatment in EDs over time, future studies focusing on other benefits such as patient satisfaction, ongoing vomiting, and ED length of stay, which are currently unavailable in the PHIS database, are needed. Lastly, individual institutions need to assess the care they provide to children with AGE to ensure that the interventions provided are achieving the desired therapeutic aim.
Corresponding Author: Stephen B. Freedman MDCM, MSc, Sections of Emergency Medicine and Gastroenterology, Department of Pediatrics, Alberta Children’s Hospital and Research Institute, University of Calgary, 2888 Shaganappi Trail, Calgary, AB T3B 6A8, Canada (firstname.lastname@example.org).
Accepted for Publication: October 28, 2013.
Published Online: February 24, 2014. doi:10.1001/jamapediatrics.2013.4906.
Author Contributions: Dr Hall had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: All authors.
Acquisition of data: Freedman, Hall, Neuman.
Analysis and interpretation of data: Freedman, Hall, Shah, Aronson, Florin, Macias, Neuman.
Drafting of the manuscript: Freedman, Hall, Neuman.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Freedman, Hall, Neuman.
Administrative, technical, and material support: Freedman, Neuman.
Study supervision: Freedman, Neuman.
Conflict of Interest Disclosures: Dr Freedman acknowledges receiving in-kind study drug/placebo from GlaxoSmithKline, the manufacturer of ondansetron, for the conduct of an unrelated study. No other disclosures were reported.
Disclaimer: Dr Shah is the JAMA Pediatrics Section Editor for Clinical Challenges but was not involved in the review process or the acceptance of the manuscript.
Previous Presentation: This paper was presented at the 2013 Annual Meeting of the Pediatric Academic Societies; May 5, 2013; Washington, DC.