This was a prospective comparative cohort pilot study. We collected data on children with symptoms at risk of dehydration, who presented to two Italian pediatric ED in Varese (between August 2015 and March 2017) and in Milano (between December 2016 and March 2017) whenever the physicians in charge of the study were on call. Informed consent was obtained from the guardians of all participants. The study was approved by the Institutional Review Board of Ospedale F. Del Ponte (No. 2015-0807).

Eligibility criteria included the presence of: acute diarrhea (at least three stools in liquid or semi-liquid consistency in the previous 24 hours) and/or vomiting (at least 3 episodes in the previous 24 hours). As exclusion criteria were considered: new born babies (correct age, <1 month), children on tube feeding, or with metabolic, or kidney or type 1 diabetes or mental retardation or severe neurological diseases, incomplete data collection form, poor Italian language comprehension and absence of informed consent.

The study included two sequential parts, performed first by the nurse triage, and later by the attending clinician in the examining room. Eight nurses and four doctors participated in the study. We have included all nurses who were currently working in our accident and emergency (A&E) department as a fixed job and all doctors who were also working in the pediatric gastroenterology outpatient clinic. The decision to limit the number of health care professionals was based on personal willingness to participate to this pilot study and on frequent work overloading in A&E. Besides, we tried to have participants with both high motivation and similar experience to get better results.

Each nurse working in our ED was instructed about the study protocol and the correct way to elicit the RT prior to the study with a training session.

The nurses training on RT and dehydration assessment was based on a short frontal lecture on acute diarrhea and dehydration and explanation of the study performed by an expert physician (S.S.) followed by an interactive theorical and practical session during which doubts were clarified. The method of detecting the RT was first showed using slides, explained as reported in the literature and then practically performed and checked. Particularly it was highlighted that to obtain a correct measured the pressure should gradually be greater on the palmar surface of the distal fingertip, which must be released immediately after the capillary bed becomes pale, after which the elapsed time must be measured until the restoration of the normal color. A time less than 2 seconds means no dehydration; the longer the RT, the greater the dehydration. We then carried out practical tests on children admitted to A&E in the same day to make sure everyone has understood the method. All nurses found the explanation clear and the test simple to perform and the physicians were fully satisfied of the ability acquired by the nurses. We decided to repeat the training at the end of the first part of the study, to assess the nurses' performance and the re-evaluation was again excellent (Fig. 1).

To all nurses the CDS scale appeared more subjective, difficult to score and time consuming than RT, so this scale was included only in the physician assessment. For data collection we used a standardized evaluation form which included items shown in Fig. 2.

At triage the nurses enrolled all children reporting symptoms of acute diarrhea or vomiting and filled in the first part of the form recording the RT. According to the triage code determining the time priority, the doctor reassessed the child performed a detailed history and physical examination, assigned a CDS score and RT, completed the form and started treatment. Whenever the nurse noticed a state of dehydration based on an alteration of the capillary RT, the physician was informed about the general condition of the child and the nurse assessment and authorized the administration of oral rehydration therapy (ORS) even before the visit, unless a suspicion of severe infection, lethargy or acute abdomen was raised.

The partial results we obtained allowed us to create a new form (Fig. 3), which replaced the old one starting from July 2016. In this new form, we added the weight gain after rehydration, measured by a physician after the ORS rehydration therapy, to use it as an indicative weight of a state of good hydration and compare it with the weight measured at triage. At the beginning we not consider the patients' waiting time between nurse's and physician's RT assessment, so in the new form we introduce this measure and the quantity of ORS assumed by the dehydrated child.

The data collected by the triage nurse included basic demographics, presenting symptoms (diarrhea, vomiting), general appearance and state of consciousness (unconscious, lethargic, normal, irritable), RT at the arrival in ED and the providing of ORS (after triage rating). The physician completed the assessment and filled in the second part of the same standardized form, recording the general appearance and state of consciousness; HR and RR, temperature, current weight and the previous weight (if available); symptoms (presence and type of diarrhea or vomiting, timing of onset, number of episodes in the previous 24 hours); RT, skin turgor, urine output reported at home, intake of ORS or other drink at home and CDS scale. RR was measured for 60 seconds by observing chest wall movements with the child quiet and comfortable [20]. We have defined tachycardia and tachypnea when HR and RR were above the 90th centiles, according to different age, as reported by a systematic review by Fleming et al. in 2011 [21].

Capillary RT was assessed at the fingertip by using a standardized technique [20]. Skin turgor was assessed by pinching a small skin fold between the thumb and index finger on the lateral abdominal wall at the level of the umbilicus [22].

Type of rehydration needed and the pop up of episodes of vomiting during oral rehydration were also reported. Finally, we considered the patient evolution in terms of discharge timing or hospitalization. In the new form we considered weight gain after rehydration to overcome the lack of an available previous weight in most of cases.

The primary outcome of our project was reliability of RT; to demonstrate the equivalence between triage nurse and clinician assessments of the RT, by comparing results of the tests and finding out if they match for every child or if they were different. Our final purpose is to introduce this parameter in the triage line up for quantification of dehydration and assignment of the priority code.

The secondary outcome was to demonstrate the validity of RT for the rapid quantification of presence and degree of dehydration. We compared RT with parameters thought to be reflective of dehydration presence and severity; the CDS scale, the number of episodes of vomiting and diarrhea before presentation to the ED, RR, HR, length of hospital staying, percentage of weight loss [17]. Discriminative validity was assessed by the need of intravenous rehydration therapy.

Characteristics of the study sample were described with frequency counts and percentages for categorical variables. The triage nurse evaluation of general aspect and RT were compared with the valuation of attending physician by using chi-squre test of Pearson's and Cohen's kappa to demonstrate interobserver reliability. A p-value less than 0.05 was considered statistically significant. Results were interpreted according to the criteria of Landis and Koch [23], in which κ values of 0.21 to 0.40, 0.41 to 0.60, 0.61 to 0.80, and 0.80 signify fair, moderate, substantial, and almost perfect agreement, respectively. In estimating validity, correlations were quantified by using Pearson correlation coefficient (Pearson's ρ). We sought statistical associations between variables indicative of dehydration (weight loss, cardiac and respiratory frequency, the number of episodes of diarrhea, vomiting, combination of diarrhea and vomiting, and the discharge time) and the RT; we repeated the analysis also for the score obtained on the CDS scale. Finally, discriminative validity was assessed as the area under the receiver operating characteristic (ROC) curve. We calculated the area under the curve (AUC), to estimate the specificity and sensitivity of RT and CDS, using as discriminating positive for the state of dehydration the start of intravenous therapy, then measuring the difference between the two AUC. Sensitivity, specificity, positive likelihood ratio, positive predictive value, and negative predictive value were quantified by using the start of intravenous therapy as parameter reflective of dehydration severity.

The RT evaluated by the triage nurse was <2 seconds in 196/242 children (81.0%) and ≥2 seconds in 46 cases (19.0%). The attending physician found the RT <2 seconds in 208/242 children (86.0%) and, ≥2 seconds in just 34 (14.0%) cases (Table 2).

We compared the RT test results obtained by nurses and physicians. The evaluation was concordant in 207 cases (85.5%), discordant in 35 cases (14.5%): among these, in 26 children (74.3%, or 10.7% of the total) the triage nurse overestimated the dehydration compared to the physician's one. In 9 of these 26 children, the nurse gave the ORS to the children after the triage, so the rehydration therapy was started before the visit; this can explain the relative overrating of dehydration in these children. The correlation between the measurements obtained by the nurse with the physician's evaluations was rated as moderate in accordance with the valuation guidelines Cohen's kappa (κ=0.480; 95% confidence interval [CI], 0.073 to 0.534).

After that we ruled out the children whom a nurse gave ORS before the physician's visit and we calculated the Cohen's kappa with the new data: the result was similar, with a moderate Cohen's kappa value (κ=0.556; 95% CI, 0.411 to 0.701). Pearson's chi-square test showed a moderate contingency between the measurements (contingency coefficient, 0.42).

Results of correlation between RT, CDS scale scores and variables are shown in Table 3.

The highest degree of correlation was found between RT and weight loss of percentage calculated using a previous weight reported by the parents, when available (ρ=−0.27; 95% CI, −0.47 to −0.04; p=0.019); it confirmed the statistical significance of the parameter. The correlation between length of stay and RT was lower but still significant (ρ=0.22; 95% CI, 0.09 to 0.35; p<0.01).

The number of diarrhea associated with vomiting episodes in 24 hours was correlated with the increase in RT (ρ=0.33; 95% CI, 0.12 to 0.51; p<0.01).

The correlation between CDS scale scores and the different parameters was similar: we found the maximum degree of correlation between CDS and percentage of weight loss (ρ=−0.37; 95% CI, −0.55 to −0.15; p<0.01), with statistical significance. A significant correlation was found between CDS and length of stay (ρ=0.15; 95% CI, 0.01 to 0.28; p=0.03) and between CDS and the number of episodes of diarrhea associated with vomiting during the previous 24 hours (ρ=0.37, 95% CI, 0.16 to 0.54; p<0.01).

For the outcome of intravenous rehydration therapy, the AUC was 0.65 for the CDS (95% CI, 0.57 to 0.73; p<0.01). Sensitivity and specificity were found to be respectively 51.2% and 72.5%. AUC for the RT score was 0.65 (95% CI, 0.57 to 0.72; p<0.01). The sensitivity and specificity were evaluated as 33.0% and 97.6%, respectively (Fig. 4).