 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Division of Emergency Medicine, Children's Hospital and Harvard Medical School, Boston, Massachusetts
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
METHODS. Our prospective sample included patients presenting to the emergency department with a chief complaint of vomiting and/or diarrhea. The association between end-tidal carbon dioxides and serum bicarbonate concentrations was determined with simple linear-regression analysis. Receiver operating characteristic curves were computed to determine the predictive ability of the end-tidal carbon dioxide to detect metabolic acidosis.
RESULTS. One hundred thirty of 146 subjects who were approached were included in the final analysis. For those for whom laboratory studies were performed, the mean serum bicarbonate concentration was 17.3 ± 4.3 mmol/L and the mean end-tidal carbon dioxide level was 34.2 ± 5.2 mm Hg. End-tidal carbon dioxide levels and serum bicarbonate concentrations were correlated linearly in bivariate analysis. Receiver operating characteristic curves were calculated for end-tidal carbon dioxide as a predictor of serum bicarbonate concentrations of 
13, 15, and 17 mmol/L, with areas under the curves of 0.94, 0.95, and 0.90, respectively. The relationship between end-tidal carbon dioxide levels and serum bicarbonate concentrations was independent of other potential predictors of acidosis in multivariable analysis. The mean end-tidal carbon dioxide level for patients who required an unanticipated return visit (33.0 ± 4.0 mm Hg) was lower than the level for those who did not seek reevaluation (36.6 ± 3.6 mm Hg).
CONCLUSIONS. End-tidal carbon dioxide levels were correlated with serum bicarbonate concentrations among children with vomiting and diarrhea, independent of other clinical parameters. Capnography offers an objective noninvasive measure of the severity of acidosis among patients with gastroenteritis.
Key Words: end-tidal carbon dioxide • acidosis • gastroenteritis
Abbreviations: EtCO2—end-tidal carbon dioxide • ROC—receiver operating characteristic • CI—confidence interval • HCO3—bicarbonate • ED—emergency department
Children commonly present to the emergency department (ED) with vomiting and diarrhea. Between 2 and 4 million outpatient visits per year and 10% of hospital admissions for children <5 years of age are attributed to gastroenteritis.1 Published treatment guidelines for the management of gastroenteritis are based on the degree of illness.2,3 Although clinical assessment forms the basis for therapeutic decision-making, laboratory data can provide additional valuable information. Serum bicarbonate (HCO3) concentration, as a measure of metabolic acidosis, has been shown to improve physician assessment of the degree of dehydration,4 to help predict which patients are more likely to tolerate oral rehydration therapy,5 and to influence patient disposition.6 The difficulty with the use of laboratory data such as HCO3 levels are that measurement necessitates venipuncture and results are not readily available at the bedside, where clinical decision-making often occurs. A rapid noninvasive measure of the degree of metabolic acidosis would be a useful adjunct for clinical assessment.
The normal physiologic response to the metabolic acidosis that is seen often among children with gastroenteritis is a compensatory respiratory alkalosis. By increasing minute ventilation, patients are able to decrease PaCO2, to help correct any underlying acidemia. Among patients with normal lung function, PaCO2 has been shown to correlate with the partial pressure of carbon dioxide during exhalation, known as end-tidal carbon dioxide (EtCO2).7 Given the relationship between PaCO2 and HCO3 levels, EtCO2 would be expected also to be correlated with HCO3 levels and potentially to provide noninvasive assessment of metabolic acidosis. This correlation between EtCO2 and HCO3 has been demonstrated with metabolic acidosis secondary to diabetic ketoacidosis.8,9
We aimed to determine the correlation between EtCO2 and serum HCO3 among children with gastroenteritis and to determine the predictive capability of capnography to detect acidosis among those patients. In addition, we compared EtCO2 with other clinical factors from the history and physical examination that might also be associated with the degree of acidosis. Finally, we explored the relationship between EtCO2 and return visits.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
During a 1-year period (August 2003 to August 2004), we enrolled a convenience sample of patients who presented to a large, urban, children's hospital ED with a volume of 
50000 visits per year. Children >1 month of age who presented with a chief complaint of vomiting and/or diarrhea were eligible for enrollment. Subjects were enrolled when either of the 2 primary investigators (J.N. and B.K.) was available. Exclusion criteria included any medical condition that can affect baseline EtCO2 or HCO3 values, including acute or chronic respiratory illness, diabetes mellitus or other metabolic disorders, and known renal or cardiac disease.
A focused history and physical examination was performed by 1 of the 2 primary investigators, to assess signs and symptoms of dehydration, including those identified on the basis of World Health Organization guidelines.3 Information was recorded on a standardized data collection sheet that included the following: demographic data, symptoms (duration of vomiting/diarrhea, number of episodes in the past 24 hours, history of fever, urine output, tears, and level of activity), and physical examination data (vital signs, weight, skin turgor, appearance of lips, oral mucosa, eyes, and fontanelle, respiratory pattern, and capillary refill). After the initial assessment, patients were connected to a portable handheld capnograph (Microcap; Oridion, Needham, MA) with an oral/nasal cannula. A 60-second stabilization period was used before initiation of EtCO2 recording, to allow subjects to adjust to breathing with the cannula in place. After stabilization, a 30- to 60-second EtCO2 measurement was obtained. To allow for minor fluctuations in EtCO2 corresponding to normal respiratory variation, the recorded value was the EtCO2 reading displayed most frequently during the recording period. In cases in which variability precluded identification of a single numeric value, the median of the range obtained during the recording period was used. For patients for whom laboratory data were obtained, all EtCO2 values were recorded before the HCO3 concentrations were available. Management decisions, including oral versus intravenous hydration, whether laboratory studies were performed, length of therapy, and final disposition, were made by the treating attending physicians, who were blinded with respect to EtCO2 values.
We used serum HCO3 concentration as our primary outcome measure in determining the correlation with EtCO2. We used return visits for additional care within 72 hours after discharge as a secondary outcome measure.
Relationship Between EtCO2 and HCO3
The correlation between EtCO2 and HCO3 was measured with Pearson's correlation coefficient and simple linear- regression analysis. We created a residual versus predicted value plot to assess measurement bias. In addition, we converted the HCO3 variable into a dichotomous categorical variable by using cutoff points of 13, 15, and 17 mmol/L. With these concentrations, we computed receiver operating characteristic (ROC) curves to determine the diagnostic accuracy of EtCO2 in predicting serum HCO3 concentrations.
Relationship Between EtCO2 and Other Clinical Variables
To be a useful adjunct, EtCO2 would have to predict HCO3 concentrations beyond other routinely available clinical information. Therefore, we used multivariate linear-regression analysis, with variables including demographic information, patient symptoms, and physical signs, to determine whether EtCO2 was an independent predictor of HCO3 levels. First, univariate distributions of continuous variables were examined to determine departures from normality. To determine the covariates for the final model, associations between demographic or clinical parameters and HCO3 levels were explored with the use of bivariate scatter plots for continuous variables and cross-tabulations for discrete variables. Those clinical variables that were found to be the best predictors of metabolic acidosis (P < .1) were then entered into a multivariate linear-regression model with EtCO2, with HCO3 levels as the dependent variable.
We used standard regression diagnostics to look for evidence of influence or colinearity. We generated new models excluding each of the significant predictors identified as above (respiratory rate, days with diarrhea, and days with vomiting) and compared the R2 of the full model with those of the models missing each covariate. Cook's D and dfbeta were used to measure influence, and the variance inflation factor was used to test for colinearity.
Relationship Between EtCO2 and Return Visits
We also explored the secondary question of whether EtCO2 was related to unanticipated return for evaluation within 72 hours. For patients discharged to home, a follow-up telephone call was made by one of the study investigators (J.N.) within 2 weeks, to evaluate unanticipated return visits to the ED or the primary care physician's office because of persistent or worsening symptoms. For families we were unable to contact, medical records were reviewed to determine whether patients were treated again in our ED. We used Student's unpaired t test to compare mean EtCO2 values between patients with unanticipated return visits and those who did not seek reevaluation. Stata for Windows statistical software (version 8.0; Stata Corp, College Station, TX) was used for all statistical calculations.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
|
15 mmol/L. For all enrolled patients, the mean EtCO2 was 34.2 ± 5.2 mm Hg (range: 20–49 mm Hg). The mean EtCO2 for patients treated only with oral rehydration was 37.3 ± 2.2 mm Hg (95% confidence interval [CI]: 35.8–38.7 mm Hg) and was significantly higher than the value for those who received intravenous hydration (34.0 ± 5.3 mm Hg; 95% CI: 33.0–34.9 mm Hg; P < .05).
For patients for whom laboratory studies were performed, paired data points for EtCO2 and HCO3 levels are shown as a scatterplot, with a fitted line representing the regression of EtCO2 versus HCO3 values, in Fig 2. Bivariate analysis showed the linear correlation of these variables, with a variance (R2 value) of 0.64 and a Pearson's correlation coefficient (r value) of 0.80 (P < .0001).
|
|
13, 15, and 17 mmol/L HCO3. The areas under the curves were 0.94 (95% CI: 0.92–0.99), 0.95 (95% CI: 0.92–98), and 0.90 (95% CI: 0.88–0.97), respectively, demonstrating excellent discriminatory performance. Figure 4 shows the fitted curve for EtCO2 prediction of HCO3 levels of 15 mmol/L. Visual inspection of the ROC curve showed that maximal sensitivity occurred at EtCO2 values of 34 mm Hg (sensitivity: 100%; 95% CI: 89%–100%; specificity: 60%; 95% CI: 49%–69%) and the best specificity without compromise of sensitivity occurred at EtCO2 values of 31 mm Hg (sensitivity: 76%; 95% CI: 63%–91%; specificity: 96%; 95% CI: 88%–98%).
|
Relationship Between EtCO2 and Return Visits
We were able to contact families of 65 (86%) of 76 patients who were discharged from the hospital. Eleven of these patients returned to the ED because of persistent or worsening symptoms. The mean EtCO2 in this group was 33.0 ± 4.0 mm Hg, compared with 36.6 ± 3.6 mm Hg for those who did not return (P = .01). Nine (81%) of the 11 patients received intravenous hydration therapy, and 8 (73%) of the 11 were admitted to the hospital at the subsequent visit. Of the 11 families we were unable to contact, none was seen again in our ED; however, we cannot be certain that they did not return to their primary care provider or another ED for repeat evaluation.
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Several studies support the clinical utility of HCO3 values as an adjunct in the assessment and treatment of patients with gastroenteritis. Two studies demonstrated a relationship between serum HCO3 concentrations and severity of dehydration, as measured by changes from baseline weight.4,10 Vega and Avner4 also showed improvement in the sensitivity of predicting moderate and severe dehydration by adding HCO3 levels of <17 mEq/L to a clinical assessment system. Yilmaz et al10 demonstrated that patients with HCO3 levels of >15 mmol/L were unlikely to be >10% dehydrated (positive predictive value: 90%), which suggests a potential role in ruling out severe disease. Reid and Bonadio5 showed that children with HCO3 levels of 13 mEq/L were less likely to tolerate an oral challenge than were those with higher HCO3. Wathen et al6 found that a low serum HCO3 concentration was one of the factors that most frequently prompted clinicians to change their treatment of patients, generally with prolonged intravenous hydration.
Relationship Between EtCO2 and HCO3 Levels
Our study demonstrates a strong correlation between EtCO2 and HCO3 levels among patients with gastroenteritis, with a Pearson's coefficient of 0.80. This reflects the expected physiologic response to metabolic acidosis through respiratory compensation. This link between respiratory pattern and severity of disease in gastroenteritis has been reported previously. The World Health Organization clinical assessment scale uses rapidity and depth of breathing as a parameter to categorize the degree of dehydration.3 Similarly, Mackenzie et al11 described a significant association between "deep (acidotic) breathing" and the degree of dehydration among children with gastroenteritis. Our study provides a quantitative assessment of this association. Specifically, EtCO2, as measured with capnography, provides a means for assessing a patient's ventilatory status as a proxy for the degree of metabolic acidosis.
To demonstrate a potential role in clinical decision-making, we investigated EtCO2 performance as a predictive tool in determining the likelihood of acidosis among patients with gastroenteritis. The cutoff points of 13, 15, and 17 mmol/L were chosen on the basis of defining values used in previous studies and to accommodate different practice patterns.4,5,10 In predicting HCO3 levels of 15 mmol/L, the area under the ROC curve was 0.95, which suggests excellent predictive value (Fig 4). To demonstrate how this might translate into practice, use of an EtCO2 value of 34 mm Hg provided 100% sensitivity. This means that EtCO2 values of >34 mm Hg effectively ruled out HCO3 levels of 15 mmol/L for our study population. Use of an EtCO2 value of 31 mm Hg increased the specificity to 96%, which means that patients with EtCO2 values in this range were very likely to be acidotic.
Relationship Between EtCO2 and Other Clinical Variables
The clinical utility of EtCO2 would be limited if it predicted HCO3 levels with no independent discriminatory power, compared with other routinely available information, including demographic factors, historical features, vital signs, and other clinical parameters. To address this, we conducted a multiple linear-regression analysis. We found that the strength of the association between EtCO2 and HCO3 levels was not affected by controlling for any clinical covariates, which suggests that EtCO2 is independent of these other signs and symptoms. Therefore, EtCO2 seems to provide information regarding the likelihood of acidosis that can serve as a useful adjunct to information obtained through routine clinical assessment.
Relationship Between EtCO2 and Return Visits
We explored whether EtCO2 would be a useful tool in predicting the likelihood of unanticipated return visits. We chose not to use admission as an outcome measure because, at our institution, many practitioners use low HCO3 levels as an indication for hospitalization. Therefore, demonstrating lower EtCO2 values for patients who were admitted might be misleading. For patients who were discharged from the hospital from the ED, we found a statistically significant difference in EtCO2 measurements between the 11 patients who returned for unanticipated reevaluation and those who did not. This suggests that there may be an adjunctive role for EtCO2 in guiding disposition for patients with gastroenteritis. However, given the small sample size and the absence of investigation into which factors influenced disposition decisions in this study, this information needs to be investigated further in a larger study.
Potential Clinical Applications
Given the demonstrated predictive abilities of EtCO2 in detecting metabolic acidosis, there are a number of potential clinical applications for EtCO2 among patients with gastroenteritis. As a triage tool, EtCO2 could help identify patients who are likely to have clinically significant acidosis. For example, an EtCO2 value of 31 mm Hg gives a positive likelihood ratio of 20.4 in detecting HCO3 of 15 mmol/L (positive likelihood ratio of 14.1 for HCO3 of 13 mmol/L), which means that these EtCO2 values are 20 times more likely to occur for an acidotic patient than for a patient with an HCO3 of >15 mmol/L. This suggests that a low EtCO2 can be helpful in "ruling in" acidosis. Information such as this might help prioritize patients for earlier therapy, which, on the basis of practice patterns, might include earlier intravenous catheter placement and rehydration therapy. In addition, detecting patients with the most severe acidosis could help identify a subgroup at higher risk for an underlying metabolic disorder (eg, organic acidemias, fatty acid oxidation defects, or glycogen storage diseases). The availability of this information before intravenous catheter placement could allow critical metabolic laboratory studies to be obtained before initiation of hydration therapy.
Use of EtCO2 values to determine that a patient is unlikely to have significant metabolic acidosis also provides valuable information. Data from the study by Yilmaz et al10 suggested that patients without metabolic acidosis had a low probability of being severely dehydrated. Use of an EtCO2 value of 
34 mm Hg had a negative likelihood ratio of 0.04 for detecting HCO3 levels of 15 mmol/L (negative likelihood ratio of 0.05 for HCO3 levels of 13 mmol/L), which means that these higher EtCO2 values can help "rule out" metabolic acidosis among these patients. On the basis of previous studies,5 this information could be used to guide clinical decision-making in triage, for example, by helping to determine early in the course of the ED visit which patients might be more likely to tolerate oral rehydration therapy successfully.
Limitations
Our study has several potential limitations. First, we enrolled a convenience sample of patients. Therefore, our subjects might not be representative of the broader population of patients with gastroenteritis presenting to pediatric EDs. In our study, this sampling strategy resulted in a group of subjects with relatively high acuity (intravenous hydration: 89%; admission: 42%). Within this sample, however, the correlation between EtCO2 and HCO3 seemed to hold even for patients with no or mild acidosis. In addition, the known biological association between PCO2 and HCO3 suggests that the linear relationship between EtCO2 and HCO3 should not be found solely in higher-acuity populations. Despite these reassurances, we recognize that, on the basis of this study alone, these results may not be generalizable to patient populations with greater distributions toward oral rehydration therapy. Unfortunately, demonstrating this correlation definitively in such populations would require unnecessary venipuncture for patients for whom oral rehydration is deemed appropriate; therefore, such a demonstration might not be feasible.
Second, recording EtCO2 values, like oxygen saturation levels, requires some interpretation and therefore may be subject to measurement bias. In our study, EtCO2 values were recorded before knowledge of serum HCO3 concentrations; therefore, any measurement error would have been nondifferential with respect to HCO3 measurements. Blinded assessment of EtCO2 would bias our results toward the null hypothesis, and the association between EtCO2 and HCO3 levels in our study population might actually be stronger than the reported results. In addition, as discussed above, there are previously established relationships between PCO2 and HCO3, which are known to be linear.12 Our study merely confirms that the EtCO2 device can measure this relationship noninvasively among patients with gastroenteritis. We think that this biological foundation, coupled with our study design (blinded assessment of EtCO2 in relation to an objective outcome measure), supports the internal validity of our study findings.
Third, laboratory evaluations for patients with gastroenteritis might reveal information about other electrolyte imbalances or hypoglycemia, which would not be detected with capnography alone. In our study, 2 patients (1.5%) had hyponatremia or hypernatremia. Both of those patients also had metabolic acidosis and EtCO2 values of <28 mm Hg, which would likely have reaffirmed the decision to perform intravenous hydration for these patients. Twelve patients had serum glucose levels of <60 mg/dL. Because it is routine practice for patients with gastroenteritis at our institution to receive either intravenous dextrose administration or sugar-containing oral rehydration before discharge, these patients would have been treated appropriately even if serum electrolyte testing had not been performed. However, the question of whether capnography or other point-of-care testing might eventually replace serum electrolyte evaluation in the treatment of patients with gastroenteritis requires additional investigation.
Finally, our study shows that EtCO2 values correlate well with acidosis. Although previous studies demonstrated that serum HCO3 levels could add to the assessment and treatment of patients with gastroenteritis, we did not directly measure a correlation between EtCO2 and the degree of dehydration.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
FOOTNOTES |
|---|
Address correspondence to Joshua Nagler, MD, Division of Emergency Medicine, Children's Hospital, 300 Longwood Ave, Boston, MA 02115. E-mail: joshua.nagler{at}childrens.harvard.edu
Financial Disclosure: Dr Krauss is a consultant for Oridion Medical, a capnography company, and holds 2 patents in the area of capnography.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2006 American Academy of Pediatrics. All rights reserved. |
||
|
||
