We conducted a prospective longitudinal analytical single-centre study between October 1 and December 31, 2017. The setting was a tertiary care hospital with a general paediatrics ward with 38 beds that manages 1750 admissions per year.

We included patients by consecutive sampling during the period under study. We included all patients admitted to the paediatric inpatient ward aged 1 month to 15 years with a length of stay of at least 24h and for whom data on anthropometric measurements taken within the first 24h of the stay were recorded. We excluded patients in other paediatric specialities (cardiology, nephrology, oncology and haematology) and postoperative patients.

These patients underwent weight and height measurements at admission (T0), at 48h after admission (T1) and at discharge (T2). The nursing staff weighted the patients at the time of admission and in subsequent measurements in early morning, with the patient naked or lightly clothed, in the decubitus position for patients aged less than 2 years and standing in older children, using validated SECA GmbH 375® baby scales. The length of patients aged less than 2 years was obtained by measuring the distance from the top of the head to the heel with the patient lying down using an ASMEDIC® T101 stadiometer. The height of children aged more than 2 years was measured with the patient standing with the SECA 797® column scale.

We collected data on clinical and demographic variables for each patient, including anthropometric measurements of weight (in kg), height (in cm), body mass index (BMI) and Waterlow classification.

The Waterlow classification assesses the nutritional status of the patient based on the current weight in relation to the median weight in the reference population corresponding to the same height. Based on this expected weight-for-height percentage, the nutritional status is classified as overweight/obesity (>100%), normal (90–100%), mild undernutrition (80–90%), moderate undernutrition (70–80%) and severe undernutrition (<70%).5

We calculated the BMI as the weight divided by the height square (in kg/m2). We documented the weight, height and BMI based on the growth tables published by Carrascosa in 2010.9

We developed a data collection form specifically for the study, where anthropometric measurements were recorded at different times: T0 (admission), T1 (48h after admission) and T2 (discharge). We also used it to collect other information, such as the nutritional intake during the hospital stay, the reason for admission or the presence of underlying diseases. We classified the severity of disease with the Wood-Downes clinical scoring system (severity ≥8) in patients with bronchiolitis and/or the Pulmonary Score in patients with asthma exacerbations (severity ≥7), both of which have been validated. For all other diseases, we considered a case severe if the patient required admission to the paediatric intensive care unit (PICU). In the subset of patients with severe respiratory illness, we differentiated between those that required high-flow oxygen therapy (HFOT) in the paediatric ward and those that required admission to the PICU (see Fig. 1).

Figure 1.

Methodology: patient recruitment and followup.

(0.46MB).

We assessed the risk of malnutrition in the patients included in the study by means of the Spanish version of the Screening Tool for the Assessment of Malnutrition in Pediatrics (STAMP), validated for use in Spanish children aged 1 month to 18 years the same year the tool was developed.8

The STAMP (http://www.stampscreeningtool.org) allows the identification of patients at risk of malnutrition at admission and provides recommendations for intervention based on the score obtained. It consists of 3 items: diagnosis, which assesses whether the disease causing the admission has nutritional implications; the nutritional intake in the days prior to admission; and the discrepancy in the percentiles for weight and height of the patients based on the number of columns the values are apart (see Fig. 1).31 The scores in each of these items are added to obtain a total score that ranges from 0 to 9 and that establishes the risk of the patient: high (≥4 points), moderate (2–3 points) or low (0–1 points).7,8 (Fig. 1). The recommendations given by the authors are that patients at high nutritional risk be assessed by a paediatric nutritionist, who would be responsible for determining which interventions will be implemented to decrease the risk of malnutrition during hospitalisation.7,8 Patients with a moderate risk should be assessed by the paediatrician in the inpatient ward, with repeated assessment with the STAMP in the next 3 days, while patients with low risk ought to be assessed with the STAMP each week.7,8 Prior to the introduction of this nutritional screening tool, training sessions on its correct use were held to educate the entire nursing and medical staff of the paediatrics ward.

We analysed the collected data with the software SPSS version 22 (IBM SPSS Statistics; Armonk, NY, USA) and Prism version 7 (GraphPad, La Jolla, California, USA). We summarised quantitative variables as mean and standard deviation or as median and interquartile range, based on whether or not the data followed a normal distribution. We have summarised qualitative variables as percentages. To compare categorical variables, we used the chi square or Fisher exact test, and to compare continuous variables the Mann–Whitney U and Kruskal–Wallis tests. We assessed the correlation between quantitative variables with the Spearman rank correlation test. To analyse the association between the level of malnutrition and the median length of stay, we plotted Kaplan–Meier survival curves.

We performed a bivariate analysis and a multivariate analysis. Thus, to assess the factors that could have an impact on the median length of stay, we constructed a linear probability model including as covariates all variables with a statistically significant association in the bivariate analysis and clinically relevant variables such as age, sex, underlying disease, nutritional intake at admission and STAMP score at admission. Subsequently, we built a logistic regression model in which the dependent variable was severity, including as covariates age and the STAMP score at admission. We present the results of linear regression in the form of the β coefficient and t statistic, and of logistic regression as odds ratios (ORs) with their 95% confidence intervals (CIs). We assessed the goodness of fit of the model by means of the R2 or the Nagelkerke pseudo R2. We ruled out colinearity in all models. We defined statistical significance as a P-value of less than 0.05. The study was approved by the Clinical Research Ethics Committee of our hospital (CEIC file 286/17).

In the period under study (October-December 2017) a total of 568 patients were admitted to the general paediatrics ward, of who 200 met the inclusion criteria. In the final sample, 55% of patients were male, and the median age was 15.8 months (IQR, 2.5–42.8). The median length of stay was 3 days (IQR, 1–18 days). Most patients were previously healthy and only 29.5% had underlying disease, of who 7.5% had been born preterm. The most frequent diagnosis at admission was acute respiratory illness (68%), and of these patients, 10% required HFOT. The proportion of patients in this cohort that required admission to the PICU was 7.5%.

In this sample, 1.5% of patients had a Waterlow score of less than 70% at admission (severe undernutrition) and 45.5% had overweight or obesity.

We found a high risk of malnutrition (STAMP ≥4) in 48.3% of patients, a moderate risk (STAMP=2–3) in 48.2% and a low risk (STAMP=0–1) in 3.5%. Table 1 presents the scores obtained in each of the items of the STAMP.

During the hospital stay, nearly 31% of the patients had a dietary intake of less than 50% their normal baseline intake by day 3 of admission, according to the nursing records, and only 2.5% received nutritional supplementation. Table 2 presents the values of the different nutritional parameters at the timepoints under study.

We made a comparative analysis based on the STAMP score, dividing the cohort into 3 groups: STAMP 0–1, STAMP 2–3 and STAMP ≥4 (Table 3). The 3 groups were comparable in terms of age, sex and underlying disease.

We found that patients with higher STAMP scores had longer lengths of stay (Fig. 2), required admission to the PICU more frequently and had more severe disease during their stay (P<.01). The Spearman rho coefficient for the correlation between the STAMP score and the median length of stay was 0.3 (P<.01). We calculated the Spearman correlation coefficient to assess the association between the STAMP score and the BMI (r=–0.14; P=.05) and the STAMP score and the Waterlow classification (r=–0.12; P=.08), and neither correlation was relevant or statistically significant. We also plotted a Kaplan–Meier survival curve that reflected the length of stay of each patient in relation to their STAMP score at admission. The curve revealed that the higher the STAMP score, the longer the median length of stay (Fig. 3). On the other hand, we did not find statistically significant differences in the median length of stay based on the Waterlow classification.

Figure 2.

Comparative analysis of length of stay by risk of malnutrition at admission.

Median and interquartile range of length of stay in days of children with different scores in the STAMP at the time of admission. We used the Kruskal–Wallis test for the analysis. The asterisks mark associations with P<.01.

(0.05MB).

Figure 3.

Kaplan–Meier survival curve for the length of stay in relation to the STAMP score at admission.

Survival represents the median length of stay (Kaplan–Meier method).

(0.08MB).

Children with underlying disease were at higher risk of undernutrition at admission (STAMP score >2 points; 80% vs. 48.9%; P=.02).

In the linear regression model, the median length of stay was the dependent variable, and male sex (t=2.4; P=.01), higher STAMP scores at T0 (t=3.5; P=.01) and greater severity (t=7.8; P<.0001) were the covariates associated with increases in the median length of stay.

The logistic regression model had severity as the dependent variable, and it showed that a higher STAMP score at T0 (odds ratio [OR]=1.48; 95% CI, 1.18–1.86; P=.01) was the covariate most strongly associated with severity. The BMI and the Waterlow weight classification did not seem to be associated with severity (Table 4).

The main limitation of the study is that it was performed in a single centre in a relatively small sample of patients. This sample may not be very representative of the total population of paediatric inpatients in our hospital, as we excluded patients receiving care in other medical specialities. Anthropometric data were only collected in 35% of the patients admitted in the period under study, so our results may not reflect the nutritional status of the entire paediatric population admitted to the hospital. Therefore, our descriptive sample statistics cannot be used to determine the actual prevalence of nutritional risk in paediatric inpatients for the purpose of comparison with previous studies. The low frequency of recording anthropometric data for admitted patients may have been a source of selection bias in our study. The lack of records of weight at different timepoints also limited our ability to assess weight gain or loss during admission. Nevertheless, our study had a prospective design and is the first to confirm the association between the STAMP screening tool score, parameters of disease severity and median length of stay.