Pediatric malnutrition is “an imbalance between nutrient requirements and intake that results in cumulative deficits of energy, protein, or micronutrients that may negatively affect growth, development, and other relevant outcomes” [1]. Malnutrition in hospitalized pediatric patients is an abnormal condition that may occur due to acute or chronic disease-related factors such as increased energy and nutrient requirements, increased nutrient losses, and poor nutritional status at hospitalization [1,2]. The prevalence ranges from 6 to 41% for acute and 8 to 47% for chronic malnutrition [3].

Anthropometry and average growth charts can follow normal growth and detect nutritional deficiencies. However, they are not suitable and sufficient for the early detection of malnutrition risk developed due to an acute condition [4]. Therefore, it is critical to identify pediatric patients at risk for malnutrition to prevent the deterioration of nutritional status [5]. The European Society for Clinical Nutrition and Metabolism (ESPEN), and the European Society for Pediatric Gastroenterology Hepatology and Nutrition (ESPGHAN) [6,7], recommends simple and rapid nutritional risk screening to identify nutritionally at-risk patients.

Screening tools aim to identify children with average anthropometric measurement results at admission yet at risk of developing malnutrition due to an acute medical condition [8]. Although several pediatric nutritional risk screening tools, such as Screening Tool for Risk on Nutritional Status and Growth (STRONGkids), Screening Tool for the Assessment of Malnutrition in Pediatrics (STAMP), Pediatric Yorkhill Malnutrition Score (PYMS), and Simple Pediatric Nutrition Screening Tool (PNST), have been reported to be effective in identifying children at risk of malnutrition, there is still no consensus on the best nutritional tool for hospitalized children [4,5,9,10].

This study aimed to determine the prevalence of malnutrition risk and compare the anthropometric measurements with the nutritional status by using four nutritional screening tools in hospitalized pediatric patients, and to evaluate which pediatric screening tool is appropriate for Turkish children. This is one of the few studies in Turkey that compared different nutritional screening tools in the pediatric population.

This prospective cross-sectional study was conducted at a tertiary medical center between January 2019 and January 2020. Turkish children aged one month to 17 years, staying in the pediatric wards with an anticipated length of stay >24 h, were included in this study. Patients were excluded if treated in the emergency department and intensive care unit. In addition, children whose anthropometric measurements could not be performed due to neurological problems or limb deficiency, who were of another ethnic origin, and who had missing data were excluded from the study. A total of 753 patients were screened for this study, and 604 of them who met the inclusion criteria were included in the analysis. On the 1st day of admission to the hospital, patient demographic data, including age, sex, anthropometric measurements, the reason for admission, diagnosis, and parents’ education status, were recorded, and four nutritional risk screening tools, including STRONGkids, STAMP, PYMS, and PNST, were applied to appropriate age ranges.

Weight was measured using a baby scale (Model 834, Seca, Birmingham, UK) with a sensitivity of 0.01 kg for under 10 kg and a children scale (Model 769, Seca, Birmingham, UK) with a sensitivity of 0.1 kg for over 10 kg. Height was measured using a Harpenden stadiometer (Holtain Ltd., Crymych, UK) with a sensitivity of 0.1 cm in children >2 years. In children <2 years, recumbent length was measured using a baby stadiometer (Model 210, Seca, Birmingham, UK). Body Mass Index (BMI) was calculated by dividing weight (kg) by height squared (m²). Weight-for-height (WFH), height-for-age (HFA), and BMI-for-age Z-scores were calculated by using the WHO AnthroPlus Software [11].

The diagnosis of malnutrition was based on the recommendations of World Health Organization (WHO) guidelines plotted on the national growth charts as the cut-off point. Moderate malnutrition was defined as <-2 Standard Deviation Score (SDS) of WFH or HFA, and severe malnutrition was <-3 SDS of WFH or HFA. Acute malnutrition was defined as <-2 SDS for WFH, and chronic malnutrition was defined as <-2 SDS for HFA. When WFH Z-score was not available, BMI-for-age Z-score was used. Moreover, BMI-for-age >2 Z-score was considered overweight or obese [12].

The STRONGkids was developed by Hulst et al. [9] to evaluate the nutritional risks of children aged one month - 18 years. The screening tool questions the child’s general condition, whether there is a high-risk disease, food intake and loss, body weight loss, and reduction in weight gain. The risk of malnutrition is evaluated in the 0-5 points range. For malnutrition, 1-3 points represent a medium risk, and 4-5 points is a high-risk.

The STAMP was developed by McCarthy et al. [10] for use by nurses in determining the nutritional risks of hospitalized children aged 2-17 years. It includes three questions evaluating factors affecting nutritional status, food intake, and anthropometric measurements. Each component carries a score of up to 3, and the total score reflects the risk of malnutrition. A score of 2 or 3 indicates medium risk, and ≥4 is high-risk.

The PYMS was developed by Gerasimidis et al. [4] as a quick and easy screening tool to detect malnutrition risk of hospitalized children aged 1-16 years, in line with ESPEN’s recommendations for screening tools. It consists of 4 questions related to current nutritional status, food intake, recent changes in nutritional status, and acute diseases that will adversely affect nutritional status, with a maximum score of 7 points. A score of 0 indicates low-risk, 1 is medium risk, and 2 or above is high-risk.

The PNST was developed by White et al. [5] to determine nutritional risk in pediatric inpatients aged 0-16 years and includes four “yes-no” questions related to unintentional weight loss, insufficient weight gain, less food intake, and the patient’s underweight/overweight status. Participants were evaluated using original [5] and adjusted [13] cutoffs as at risk of malnutrition or not at risk. The original cutoff is at least 2 “yes” answers, while the adjusted cutoff is 1 or more “yes” answers.

Ethics committee approval was obtained from the local ethics committee (approval nº 2018/1544), and informed consent was obtained from the parents of the children.

Power analysis was calculated in the statistical software G*Power (version 3.1), and the sample size of 604 participants provided 99.9% power based on a significance level of 0.05 and a prevalence of malnutrition of 24.2%. In addition, when sample power was estimated using malnutrition identified by WHO guidelines in relation to the malnutrition risk by the STRONGkids, STAMP, PYMS, and PNST tools obtained by Logistic regression, the sample size of 604 participants provided 99.9% power for all parameters at an alpha level of 0.05.

The IBM® SPSS® software (version 22.0) was used for statistical analysis. Categorical variables were summarized as numbers (percentage, %) and compared using the Chi-square test. Continuous variables were presented as median, minimum-maximum, and 25^th - 75^th percentiles. Normality was assessed by the Kolmogorov - Smirnov test. Since continuous variables do not follow a normal distribution, the Mann-Whitney U test was used for two-group comparisons, and the Kruskal-Wallis test was used for more groups.

Diagnostic parameters (sensitivity, specificity, and positive and negative predictive values) of STRONGkids, STAMP, PYMS, and PNST were calculated using the web-based software MedCalc’s Diagnostic Test Evaluation Calculator and expressed as percentages. The 2×2 crosstab tables were constructed to assess the ability of STRONGkids, STAMP, PYMS, and PNST to detect malnutrition risk as compared with WHO diagnostic criteria. Confidence Intervals (CI) for sensitivity and specificity were “exact” Clopper-Pearson CI, while CI for the predictive values were the standard logit confidence intervals given by Mercaldo et al. 2007 [14].

Logistic regression analyses were performed to determine the associations between malnutrition identified by WHO guidelines (as a reference standard) and malnutrition risk by the STRONGkids, STAMP, PYMS and PNST tools, mother’s education level, and age. Odds Ratios (OR) and 95% CI were reported. For all statistical analyses, p<0.05 was considered statistically significant.

This is one of the few studies conducted in Turkey that compared different screening tools to determine malnutrition risk in the Turkish pediatric population. Given the attention to the consequences of malnutrition on hospitalized children, the findings of this study may signify the importance of using a nutritional risk screening tool in pediatric hospitals and help identify the appropriate screening tool for Turkish children.

In this study, the prevalence of acute and chronic malnutrition was found as 15.1% and 11.8%, respectively, in line with the relevant results of the previous studies conducted in Turkey (11.2% and 16.6%, respectively) [8]. However, local studies in the literature also reported a lower rate of chronic malnutrition and a higher rate of acute malnutrition (4.7% and 20.1%, respectively) [15,16]. The differences between the results may be attributed to using different parameters to assess the nutritional status of patients.

It is well known that children, especially the young are more susceptible and vulnerable to malnutrition than adults [17,18]. In this present study, the prevalence of acute and chronic malnutrition was found to be more common in children <5 years and < 2 years than in children aged ≥5 years and ≥2 years (31% vs. 17.3%, p<0.001; 32.4% vs. 21%, p=0.003 respectively). These results are consistent with previous studies [8,9,19,20].

Studies on pediatric screening tools used in identifying children with malnutrition risk have mostly focused on the diagnostic properties of screening tools by assessing the high-risk versus low-risk or high-risk versus moderate and low-risk children [10,15,19,21,22]. The screening tools should be able to distinguish the children at real risk of malnutrition from those exempted from the detailed nutritional assessment because they are not malnourished at the initial assessment. In a study conducted in Turkey, Pars et al. [23] found that PYMS had the highest sensitivity (96.8%), specificity (65.0%) and NPV (99.2%), STRONGkids had the lowest specificity (30.0%), and STAMP had the lowest sensitivity (70.0%). In this study, it was determined that PNST (adjusted cutoffs) had the highest sensitivity (74.5%), PNST (original cutoffs) had the highest specificity in detecting malnutrition (acute and chronic) risk. On the other hand, the PNST (with original cutoff values) screening tool had the lowest sensitivity (41.1%), and the STRONGKids had the lowest specificity (55.0%) among the screening tools investigated within the scope of this study.

In a recent study, Carter et al. [13] demonstrated that PNST was unsuitable based on threshold values for clinical use. They adjusted the threshold values of PNST for nutritional risk using receiver operating characteristics curve analysis. They consequently determined that the PNST tool with adjusted cutoff values had more robust inter-rater reliability and concurrent validity than STRONGkids. Similarly, in this study, PNST with original cutoff values had the lowest sensitivity (41.1%) among all screening tools, while PNST with adjusted cutoff values had the highest sensitivity (74.5%).

In this study, the rates of patients identified as moderate- or high-risk patients by the PYMS, PNST with adjusted cutoffs, STAMP, STRONGkids, and PNST with original cutoffs were 74.5%, 74.0%, 67.4%, 64.4%, and 41.1%, respectively. In comparison, the rates reported in different studies using pediatric screening tools vary greatly [8,13,20,22-25]. Inconsistencies between relevant results can be attributed to differences between target populations and a lack of consensus on the best method to assess nutritional status and the best definition of pediatric malnutrition, therefore, on which gold standard should be used to validate any screening tool [8,13,19,22-25]. In our study, we used the recommendations of WHO guidelines as a reference standard for diagnosing moderate, severe, acute, and chronic malnutrition. However, assessing the nutritional status of children with moderate, acute malnutrition is a challenge as no single indicator can be used alone [17].

Gerasimidis et al. [4] demonstrated that children at high-risk for malnutrition had significantly lower BMI values and that low BMI was associated with being assessed in the high-risk category. Similarly, as shown in Table 4, patients with acute malnutrition (WFL/H or BMI-for-age <-2 SDS) constituted the majority of patients in this study’s high malnutrition risk category, regardless of the screening tool used.

Different authors have argued that screening tools based on anthropometric measurements (e.g., PYMS and STAMP) detect a greater number of children with abnormal anthropometric measurement results compared to screening tools that do not include anthropometric measurements (e.g., STRONGkids and PNST) [26,27]. However, in this study, it was determined that only the PYMS tool was associated with an increased risk of malnutrition (OR: 5.88) compared to the STAMP (OR: 4.89) and STRONGkids (OR: 3.88) tools. These positive associations were more remarkable for acute malnutrition (OR: 6.57 for STRONGkids, 5.84 for STAMP, and 20.35 for PYMS) than chronic malnutrition (OR: 1.27 for STRONGkids, 3.28 for STAMP, and 1.72 for PYMS). In addition, associations with chronic malnutrition were statistically significant only for STAMP and PNST with adjusted cutoff values.

It is well known that children are more susceptible and vulnerable to malnutrition than adults due to their low energy reserves, higher energy requirements per unit of body weight, and higher nutrient requirements [17]. Malnutrition can occur in children of any age, but as the WHO emphasized, younger children are more vulnerable [18]. In line with the results of previous studies [8, 19, 20], the prevalence of acute and chronic malnutrition was found to be more common in children <5 years old and <2 years old compared to children aged ≥5 years and ≥2 years, respectively (31.0% vs. 17.3%, p<0.001; 32.4% vs. 21.0%, p=0.003). Being <2 years old significantly increased the risk of malnutrition 1.8 times, while this risk more than doubled in children <5 years old. Also, having the education level of an illiterate mother was associated with a threefold increased likelihood of malnutrition and acute and chronic malnutrition.

In addition to providing new information on the prevalence and risk of malnutrition in a group of Turkish pediatric inpatients, another major strength of this study is that all eligible children admitted to the tertiary pediatric hospital where this study was conducted were included in the study and studied throughout the study period. Furthermore, to our knowledge, this is the first study in that four nutritional screening tools (STRONGkids, STAMP, PYMS, and PNST) were simultaneously compared to determine malnutrition risk in Turkish pediatric patients. Another strength of this study is that the same researcher conducted all anthropometric measurements. In this way, the possible negative effects of interobserver variability on the study results were avoided. Lastly, considering that most studies on pediatric screening tools focus on the differences between risk categories, screening tools’ sensitivity, specificity, NPV, and PPV values given in this study will likely guide other studies.

The primary limitation of this study was assessing patients’ nutritional status based only on baseline anthropometric measurements (weight and height) without the use of other indicators such as skinfold thickness or body composition, as it might have contributed to the misclassification of some patients as high-risk, particularly in the presence of chronic malnutrition. Secondly, in the initial analysis, the lack of a complete nutritional assessment which includes body composition analysis, biochemical parameters, and food diary records for cross-checking nutritional risk screening tools, may be considered an additional limitation of this study. However, given that this study aimed to determine the adequacy and effectiveness of previously approved screening tools, it is also possible not to consider the lack of a complete nutritional assessment a limitation. Finally, this study is a cross-sectional study without data on the longitudinal analysis of patients’ clinical course and dietary changes over time, including weight loss. Despite the limitations stated, the findings of this study will likely provide guidance for studies to be carried out in the future to determine the risk of malnutrition in hospitalized children in Turkey.

In conclusion, the findings of this study indicated that PNST (with adjusted cutoff values) and PYMS screening tools have higher diagnostic accuracy compared to other screening tools in assessing the nutritional status of hospitalized Turkish children and detecting the hospitalized Turkish children with acute malnutrition in particular. Considering that early detection of malnutrition risk in children admitted to the hospital can improve health outcomes and quality of life, routine use of an easily applicable and appropriate nutritional risk screening tool in hospitalized pediatric patients should be encouraged, and all children who are identified to be at risk of malnutrition should be referred to a dietitian for nutritional intervention.