Comparing the accuracy of the three popular clinical dehydration scales in children with diarrhea

Diarrhea has the highest incidence of any childhood disease in all regions of the world and kills approximately 1.9 million children each year, accounting for 19% of all deaths in children under 5 [1, 2]. The World Health Organization (WHO), the Centers for Disease Control and Prevention (CDC), the American Academy of Pediatrics (AAP), and the European Society of Pediatric Gastroenterology and Nutrition (ESPGAN) all support rehydration with oral rehydration solution (ORS) for children with mild to moderate dehydration, reserving intravenous (IV) fluids for children with severe dehydration [3–6]. Oral rehydration solution is a hypo-osmolar solution composed of salts, sugar, and citrate, while the recommended intravenous fluids include lactated Ringer's or normal saline [4, 6–8].

In order to apply the most appropriate treatment for dehydration in children with gastroenteritis, healthcare providers must first accurately assess the severity of dehydration [5]. Underestimating fluid deficit, and not providing proper rehydration with either ORS or IV therapy can lead to acidosis, electrolyte disturbances, acute kidney injury, or even death. Alternatively, overestimating fluid deficit can lead to unnecessary interventions, longer hospital stays, and increased adverse events in children [9, 10]. Accurate fluid assessment is of utmost importance in low- and middle-income countries, where many patients travel several hours to reach a healthcare facility, and resources such as IV fluids and hospital beds are scarce.

Thus far, only the CDS has been prospectively validated against the accepted gold standard of percent weight change with rehydration at a single pediatric referral center in North America [11]. None of these scales have been validated in low- or middle-income countries where disease patterns may differ from high-income countries, patients often present later in the course of their disease, and healthcare providers often lack specialty training. The purpose of this study was to determine the accuracy of the WHO scale, Gorelick scale, and CDS in a resource-limited setting.

Seventy-three children were enrolled in the study. Two children died prior to discharge, 12 children had evidence of severe malnutrition, and 7 children were missing discharge weights, leaving 52 children for analysis. Of these 52 children, 49 children were between 1 month and 5 years old and could be classified by the Gorelick and WHO scale, while 48 children were between 1 month and 3 years old and could be classified by the CDS.

The mean age was 10.5 months for children 1-36 months and 11.14 months for children 1 month-5 years. The median percent weight change for children between 1 and 36 months and children between 1 and 60 months was 4.8%. Twenty-nine percent of children presented with severe dehydration, classified as > 10% weight change between admission and discharge. The median length of stay was 4 days, with 88% of patients staying at least 3 days for both age groups.

None of the scales had an area under the ROC curve statistically different from the reference line (see Figures 1, 2, 3, 4 and 5). Stratifying by healthcare provider (doctor versus nurse) did not improve the accuracy of any of the ROC curves (data not shown). The WHO scale for moderate dehydration (5-10% percent weight change) had an area under the curve of 0.58 (95% CI = 0.39-0.78); sensitivity was 50% and specificity 61%; LR+ was 1.28 and LR- 0.82. For severe dehydration (> 10% percent body weight change) the WHO scale had an area under the curve of 0.58 (95% CI = 0.41-0.75); sensitivity was 79% and specificity 43%; LR+ was 1.38 and LR- 0.50. For moderate dehydration (between 5-10% percent body weight change) the 4- and 10-point Gorelick scales had an area under the curve of 0.62 (95% CI = 0.45-0.78) and 0.58 (95% CI = 0.42-0.74), respectively; sensitivities were 64% and 21%, and specificities were 69% and 82%, respectively; LR+ was 1.15 and 1.25 and LR- 0.78 and 0.53, respectively. For severe dehydration (≥10% body weight change) the 4- and 10-point Gorelick scale had areas under the curve of 0.62 (95% CI = 0.45-0.79) and 0.60 (95% CI = 0.44-0.77), respectively; sensitivities were 68% and 82%, and specificities were 41% and 35%, respectively; LR+ was 1.09 and 2.04, and LR- 0.52 and 0.88, respectively. For the prediction of moderate dehydration (percent weight change ≥6%), the CDS had an area under the curve of 0.64 (95% CI = 0.44-0.77). The sensitivity of the scale was 68% and the specificity 45%, with a LR+ of 1.24 and LR- of 0.70.

While most experts agree that children with diarrhea should be treated based on the severity of their dehydration, with children receiving intravenous fluids for severe dehydration and oral rehydration solution for mild to moderate dehydration, there is no clear consensus on how best to determine the severity of dehydration, especially in resource-limited settings. The gold standard for dehydration is percent volume loss with diarrhea, which is defined as the difference between pre-illness weight and acute-illness weight divided by pre-illness weight. Because pre-illness weight from a pediatrician's office is often not available, especially in low- and middle-income countries, it is ideally substituted by stable post-rehydration weight or the child's weight after undergoing therapy. In fact, Gorelick et al. validated post-rehydration weight as a surrogate for pre-illness weight by demonstrating near perfect correlation (r = 0.9988) between the two values in a small cohort of children with diarrhea [12]. In our study, we were not able to guarantee that participants had reached a stable post-rehydration weight prior to discharge. However, most children in the study by Gorelick et al. achieved a stable weight after 24 h in the hospital, and nearly all children achieved a stable weight by 72 h. Since all patients in our study spent at least 24 h in the hospital, and 88% spent more than 3 days, it is likely that they had the opportunity to achieve a stable rehydration weight, so we believe that percent weight change with rehydration can be used as a valid gold standard for dehydration in our study.

While percent weight change with rehydration makes an excellent gold standard for the severity of dehydration, it is not a useful tool in practice, since it is not available at the time of presentation when a decision about how best to manage a child with diarrhea must be made. For many years, experts have recommended the use of physical exam signs to predict the severity of dehydration in children with diarrhea. Steiner et al. found in a systematic review that the most useful individual signs for predicting 5% dehydration were abnormal capillary refill time, abnormal respiratory pattern, and abnormal skin turgor, which had positive likelihood ratios spanning from 2.0-4.1 [9]. However, none of these signs had very good negative likelihood ratios, meaning that they were not useful for excluding severe dehydration in children. Steiner et al. concluded that no individual clinical sign had adequate sensitivity and specificity for the prediction of dehydration. Other studies that have looked at laboratory values, such as BUN, anion gap, base deficit, bicarbonate concentration, and urine specific gravity, have generally not found them to be very good predictors of dehydration status, with only bicarbonate greater than 15 or 17 mEq/L useful for reducing the likelihood of dehydration [6, 13–16].

Given the limitations of individual clinical signs, several prior authors have tried to combine physical findings into clinical scales to predict percent dehydration in children. Gorelick created a 4-point and 10-point scale for assessing dehydration in children 1-60 months old presenting to Children's Hospital of Philadelphia, resulting in sensitivities of 79% and 87% and specificities of 82% and 85%, respectively, for predicting ≥5% dehydration. The two scales had sensitivities of 82% and 90% and specificities of 83% and 90%, respectively, for ≥10% dehydration. The CDS was derived at the Toronto Hospital for Sick Children and then prospectively validated at that site by specialized pediatric staff. In the validation study, Parkin et al. demonstrated likelihood ratios for moderate dehydration of 2.2, 1.3, and 5.2 for CDS scores of 0, 1-4, and 5-8, respectively [11, 17].

Neither the Gorelick scale nor the CDS performed as well in our population of children in Rwanda as they did in North America. In fact, both scales had areas under the ROC curves statistically indistinguishable from the reference line, meaning they were no better than chance at predicting moderate or severe dehydration. In addition, the WHO scale, considered the standard of care in most low- and middle-income countries, although it has never been prospectively validated for predicting severe dehydration, also performed poorly in our population of children. To our knowledge, this study is the first to prospectively assess a clinical dehydration scale in a low-income country, where children tend to present with more severe dehydration and be evaluated by personnel with less specialized training than their high-income country counterparts. Our study highlights the need for more research into better methods for detecting the severity of dehydration in children with diarrhea in resource-limited settings and supports a general rule that clinical scales derived in a high-income country setting require validation in resource-limited settings before being recommended for widespread use in these settings.

In this study, we found that the WHO scale, Gorelick scale, and CDS, when performed by general practice physicians and nurses in a resource-limited setting, were not accurate predictors of severe dehydration in children with diarrhea and/or vomiting. Due to the high prevalence and significant morbidity associated with diarrhea in children throughout the world, further research is necessary to develop and validate new clinical scales or other diagnostic tools with greater accuracy for assessing dehydration in children in resource-limited settings.