Page 66 - 15Diarrhoeaandvomiting

Assessing dehydration and shock

Evidence overview

A systematic literature search was undertaken to inform the two questions. Two studies are included
for the first question on the accuracy of clinical signs and symptoms in detecting dehydration,
while for the second question four published guidelines are included. These guidelines had
employed different methods for classifying severity of dehydration.

Clinical detection of dehydration
Two relevant studies were identified, the first a systematic review of diagnostic studies and the
second a cohort study comparing digitally measured capillary refill time (CRT) with conventional
CRT and overall clinical assessment.
74
The systematic review was conducted to review the precision and accuracy of symptoms,
signs and basic laboratory tests for evaluating 5% (or worse) dehydration in young children aged
1 month to 5 years. After a systematic literature search of the MEDLINE database, additional
searches were conducted on the individual symptoms and signs, the Cochrane Library, and the
reference lists of text books and of all included articles. After reviewing hard copies of 110
articles, 26 articles fulfilled the inclusion criteria and underwent a quality assessment, and 13
studies were finally selected for inclusion.
The reference standard used for assessing dehydration was the ‘percentage of volume lost’,
calculated as the difference between the rehydration weight (the post-rehydration weight) and
the acute weight (the weight at presentation) divided by the rehydration weight. Three of the
included studies were based on independent, blind comparison of the test with the reference
standard but the participants were enrolled in a non-consecutive manner. The remaining ten
studies were based on non-independent comparisons of a test with the reference standard and
no selection criteria were defined. Hence overall the quality of the included studies was poor.
Meta-analysis of the accuracy results using the random effects model was conducted only if more
than two studies evaluated a specific diagnostic test. [EL = 3]
Although the authors reported test accuracy results for detecting 5% dehydration, detailed
reviewing of the individual studies included in the review revealed that the results were applicable
for the detection of 5% or worse dehydration.

Precision
Agreement between parental observation of signs and the signs elicited by trained emergency
department nurses was evaluated in a single study. The best level of agreement was reported for
sunken anterior fontanelle (k = 0.73) and presence of cool extremities (k = 0.70), while moderate
agreement was seen for general appearance, presence of sunken eyes, absence of tears, and
presence of dry mouth (k values in the range 0.46–0.57). Three studies reported on agreement
among clinicians but wide variation was seen in the results for the various signs. Prolonged CRT
had k values ranging from 0.01 to 0.65, while absent tears had values from 0.12 to 0.75. For the
rest of the signs, the level of agreement was either slightly better than the chance agreement (k
value in the range 0.50–0.60) or worse than the chance agreement (k < 0.50).
Clinical history including symptoms
Three studies were included and all of them evaluated history of low urine output as a test
for detecting dehydration. A pooled analysis showed that it did not increase the likelihood
of detecting 5% dehydration (+LR 1.3; 95% CI 0.9 to 1.9). However, results from two studies
showed that parental reporting of a normal urine output decreased the likelihood of detecting
dehydration, although the results were statistically not significant in one study (−LR 0.27; 95% CI
0.14 to 0.51 and −LR 0.16; 95% CI 0.01 to 2.53). One study reported LRs for a number of
symptoms, including history of vomiting and diarrhoea (severity), decreased oral intake, and
a previous trial of clear liquids, but none of these were found to be helpful in increasing or
decreasing the likelihood. This study also suggested that children who had not been previously
evaluated by a physician during the illness might be less likely to be dehydrated, but the results
were again statistically not significant (−LR 0.09; 95% CI 0.01 to 1.37).

Signs
The results of the test characteristics of various signs are given in Table 4.1. Three signs showed
evidence of increasing the likelihood of detecting 5% dehydration: prolonged CRT (four studies,

41