Design

This systematic review of published validation studies was registered with PROSPERO (online supplementary file 1). The findings are reported according to the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement27 (see PRISMA checklist; online supplementary file 2).

We focused on answering the following research questions (RQs) (see online supplementary file 3): RQ1: What are the currently available screening and assessment tools for detecting malnutrition in hospitalised paediatric patients? RQ2: What is the validity of the screening and assessment tools versus the reference standard?

Search strategy

To identify all relevant publications, we performed systematic searches in the following bibliographic databases: PubMed, CINAHL and MEDLINE. The searches were carried out on October 20, 2017. The keyword combinations used included the following: premature*, immature*, child*, baby, infant*, newborn*, neonate*, kid*, babies, adolescent*, pediatric*, paediatric*, screen*, assess*, tool*, undernutrition*, undernourish*, malnutrition, malnourish*. The search strategies are outlined in online supplementary file 4, and they were adapted to each database and kept consistent across all searchers. The reference lists of the identified studies were manually searched for potentially relevant studies.

Study selection

All potentially relevant titles and abstracts were blinded for author, journal and year of publication, and then screened for eligibility by two reviewers (PK and PPB), independently. Differences in judgement were resolved through a consensus procedure. All of the studies obtained from the bibliographic databases were entered into EndNote X8, and duplicates were excluded. The inter-rater agreement between the reviewers based on Cohen’s kappa statistic was 0.79 for 576 studies. The full texts of the selected studies—no longer blinded to authors and journals—were obtained for further review by two reviewers, independently (PK and PPB), to judge for eligibility. The Cohen’s kappa coefficient here was 0.93 for 64 studies, which indicated a high level of agreement. In cases of doubt, a decision was made by a third reviewer (NMV).

The studies were reviewed to ensure their focus was aligned with the purpose of the literature review. Those that were clearly inappropriate to answer the RQs and/or did not fit the predefined inclusion criteria were excluded. The flowchart of the complete search and selection process is shown in figure 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram of the search and study selection process.

Inclusion/exclusion criteria

The studies eligible for inclusion were validation studies in English that reported sensitivity (Se), specificity (Sp) and positive/negative predictive values (PPVs, NPVs, respectively) in paediatric populations. All of the predefined inclusion and exclusion criteria were used as outlined in online supplementary file 5. The list of studies (n=38) not meeting the selection criteria after reading the full text is presented in online supplementary file 6.

Quality appraisal

The methodological quality assessment of the studies was performed using Review Manager V.5.3,28 with a revised tool for the Quality Assessment of Diagnostic Accuracy Studies (QUADAS)-2.29 The QUADAS-2 tool uses four key domains to rate the risk of bias and the applicability of primary diagnostic accuracy studies. The key domains are as follows: patient selection (sampling, inclusion/exclusion criteria, sampling bias, adequacy), index tests (the validated tool, correct use and interpretation, possible bias), reference standards (the reference tool, correct use and interpretation, possible bias) and the flow and timing (the sequence, time interval, correct performance of reference standard and index test, possible bias). The results from QUADAS-2 can be expressed as high/unclear/ low risk of bias and as high/unclear/ low applicability concerns.

Data extraction and synthesis

The data were extracted by two reviewers (PK and PPB) and checked by a third reviewer (MP) using the predefined data-extraction criteria, which included the following: authors and country, nutritional screening/assessment tool used, study type, sample size, age of participants and reported clinical performance. To evaluate the clinical performance and diagnostic accuracy of the screening tools, the following criteria were considered: Se, Sp, PPV and NPV. Studies that did not report Se, Sp, PPV and/or NPV, but that provided the data that enabled calculation of these values, were also included in the study. These metrics were subsequently calculated manually by the authors and are indicated as such in the results tables.

The validation of the reproducibility and reliability of the screening/assessment tools was also considered, using data from the agreement analysis between the assessed tools and the chosen reference, as well as the inter-rater agreement shown in the studies.

For reasons of clarity, we have rated the results of each study as good, moderate/fair or poor validity. The kappa values were rated by the classification system proposed by Landis and Koch.30 Although the literature does not provide general cutoffs for Se and Sp, as they greatly depend on the clinical consequences, we have rated the values to maintain transparency and clarity, as in van Bokhorst-de van der Schueren et al.31 All of the cutoff points are described in table 1.

Patient and public involvement

No patients or public were involved in the present study.

Search results

Figure 1 shows the flow diagram of the search and study selection process. We identified 724 studies initially, of which 26 met all of the pre-established eligibility criteria and were included in the critical appraisal.

Paediatric nutritional screening tools identified

During this systematic review, we identified eight validated pediatric nutritional screening tools and three nutritional assessment tools for hospitalised paediatric patients. These can be classified based on their specialties, as follows:

• Medical and Surgical Department: Screening Tool for the Assessment of Malnutrition in Paediatrics (STAMP),32 33 Paediatric Yorkhill Malnutrition Score (PYMS),7 34 Screening Tool for Risk On Nutritional status and Growth (STRONG_kids)2 and Paediatric Nutrition Screening Tool (PNST).9

• Surgical Department: Subjective Global Assessment (SGA)35–38 and Subjective Global Nutritional Assessment (SGNA).13

• Oncology Department: Nutrition Screening tool for childhood Cancer (SCAN),39 which was developed specifically for children with cancer.

• Pulmonology Department: Nutrition screening tool for pediatric patients with cystic fibrosis (CF),40 and Nutritional risk screening tool in CF41 for pediatric patients with CF. 

• Neonatal Intensive Care: Neonatal Nutrition Screening Tool (NNST)42 for infants in the Neonatal Intensive Care Unit.

• Clinical Assessment of Nutritional Status (CANS) score,43 to differentiate malnourished from appropriately nourished babies.

However, six additional screening tools were identified in the eligibility step of PRISMA, although the associated studies were excluded in the final step as the inclusion criteria were not met:

• Medical Department: Nutrition Risk Score (NRS).44

• Medical and Surgical Department: Simple Paediatric Nutrition Risk Score (SPNRS)45 and Pediatric Digital Scaled MAlnutrition Risk screening Tool (PeDiSMART).46 47

• Psychiatric Department: St Andrew’s Nutrition Screening Instrument (SANSI).48

• Screening tool to predict malnutrition among children under 2 years old in Zambia.49

• Nutrition screening for infants and young children with special health care needs: A Look at Your Child’s Nutrition.50

Characteristics of the studies included in the systematic review

The characteristics of the 26 studies included in this systematic review are outlined in table 2. Sample sizes varied from 32 to 14 477 participants. Eleven studies (42.3%) excluded patients with length of hospital stay (LOS) of <24 hours.2 9 14 32 39 51–56 The studies often excluded children <1 year (57.7%),2 6 7 32 38–41 52 53 55–59 intensive care unit (ICU) patients (34.6%)2 14 38 51 52 58 60–62 and patients with unstable/specific conditions, such as oncology patients,53–55 61 conditions that affected hydration,9 39 cardiology, renal and orthopaedic specialties,7 fever, diarrhoea,6 obesity53 and others. Some of the studies included only patients with particular conditions, where a specially designed screening/assessment tool was usually used. However, STAMP was designed for clinical and surgical patients and was validated also on patients with spinal cord injury (SCI)6 14 and with inflammatory bowel disease (IBD).59 Additionally, STAMP was also used in outpatients in two studies.57 59 The SGA assessment tool was originally designed for adults; however, in one of the selected studies, it was tested on children.38

There was relatively high heterogeneity in the choice of the reference standards. Anthropometric measurements were used in 18 of the selected studies (69.2%)2 6 9 38 42 51–55 58–65 and full dietetic/nutritional assessments were used in 5 studies (19.2%).7 14 32 41 57 Three studies (12.5%) used SGNA as the reference standard.9 39 56 The Cystic Fibrosis Foundation (CFF) Consensus Report criteria were used as the reference standard in both of the studies on patients with CF.40 41 Validation with other screening tools along with the reference standards were reported in four studies (16.7%).14 40 59 63 Nine studies (37.5%) reported validation of two or more screening tools with the same reference standard and on the same patients.9 53 54 56 57 59 61 63 65

Risk of bias and applicability concerns

The results of the quality appraisal analysis using the QUADAS-2 tool are presented in online supplementary file 7.

The patient selection was considered as high risk or unclear because of the non-specific description of the patient selection process in three studies.7 41 58 Possible bias from conducting non-blinded index tests with respect to the results of the reference standard and/or vice versa was considered as high risk in seven studies,7 9 38 39 56 59 62 and as a possibility for bias in 15 studies.2 6 14 32 40–42 51 53 58 60–62 64 65 Similarly, non-blinded interpretation of the reference standard with regards to the index test results was considered as a possibility for bias in 17 studies.2 14 32 39–42 51–53 56 58 60–62 64 65 The information about the patient flow and timing was considered to be unclear in 15 studies,2 7 14 32 38 40 42 53 54 56 58–60 62 65 as the intervals between the index tests and the reference standard measurements were not reported. In one study,62 the anthropometric measurements were used as the index test and compared with the reference standard CANS score, which in our opinion introduces possible bias.

The reference standard was the second key domain of concern regarding applicability. Murphy et al39 used the SGNA tool for the reference standard, which is an assessment method rather than a reference standard.7 Similarly, Wonoputri et al56 defined the WHO anthropometric grow chart as the reference standard; however, the results presented only showed the comparisons of PYMS, STAMP and STRONG_kids with SGNA. As mentioned before, Soundarya et al62 used the CANS score as the reference standard; however, in our report we present the results as reported in the study and also as calculated from the reported data in an inverted manner. The inclusion or exclusion criteria were not clearly defined in five studies.7 41 56 58 65

Evaluation of the screening/assessment tools

The diagnostic accuracy of the validated screening/assessment tools for the chosen reference standards in the selected studies is presented in table 3.

Nine studies (34.61%) did not report Se, Sp, PPV and NPV; however, the data reported enabled the calculation of these validation metrics. Additionally, two studies did not report PPV and NPV, which were subsequently calculated by the authors of the present study. The calculated values are highlighted with dagger (†) in table 3. In one study, only the Se was reported, with no data for the calculation of the other validation metrics.55

The sensitivities of the screening/assessment tools ranged from 20%63 to 100%.6 39 52 56 59 SCAN with oncological patients and STAMP and STRONG_kids with clinical patients showed the best results versus SGNA in terms of Se.39 56 Additionally, STAMP performed with 100% Se versus the anthropometric measurements for inpatients and outpatients with SCI.6 STAMP, STRONG_kids, SPNRS and PYMS obtained 100% Se versus International Statistical Classification of Diseases and Related Health Problems (ICD)-10; however, the Sp was poor (0%), except for PYMS (53.5%).59

The specificities ranged from 0% (STAMP, STRONG_kids, SPNRS vs ICD-10)59 to 96.4% (STRONG_kids vs PYMS).39 The combination of Se and Sp was evaluated as good only in four studies7 40 61 63 according to classification outlined in table 1.

Relatively high NPVs were seen for most of the studies, which ranged from 0%58 to 100% (SCAN vs SGNA; STAMP vs SGNA; STRONG_kids vs SGNA; PYMS vs ICD-10; STAMP vs nutritional intervention; STRONG_kids vs nutritional intervention).6 39 52 59 On the contrary, the observed PPVs were a lot lower in most of the studies; these started from 2.5% (STAMP vs TSFT)63 and reached as high as 100% (STAMP, STRONG_kids, SPNRS, PYMS, among each other).59

Agreement between the nutritional screening/assessment tool and the reference standard or other screening tool was verified in 12 studies (46.1%).6 7 14 32 38 40 53 56 57 59 63 65

As presented in table 3, all of the abovementioned validation metrics differed greatly when different cutoff values were used. The studies also included different populations and different sample sizes, and therefore direct comparisons of the results are not possible.

Seven studies (26.9%)7 14 32 41 51 55 61 also reported interobserver agreements, which varied from 0.53 for PYMS completed by two dietitians compared with the nursing staff,7 to 0.9 for STAMP completed by dietitians and nursing staff.32 Only two studies reported intraobserver agreement, where there was substantial agreement with the kappa value of 0.6 for STRONG_kids51 and 0.6 for STAMP.14

RQ1: What are the currently available screening and assessment tools for detecting malnutrition in hospitalised paediatric patients?

Currently, there are 14 nutritional screening tools and 3 nutritional assessment tools. In this systematic review, we identified validation studies of eight nutritional screening tools (SCAN, Nutritional screening tool for paediatric patients with CF, Nutritional risk screening tool in CF, NNST, PYMS, STRONG_kids, STAMP and PNST) and three malnutrition risk assessment tools (SGA, SGNA and CANS score). Six screening tools were not included in this systematic review (NRS, SPNRS, PeDiSMART, SANSI, Screening tool to predict malnutrition among children under 2 years old in Zambia and Nutrition screening for infants and young children with special health care needs: A Look at Your Child’s Nutrition), as the studies identified did not include validation of the screening tool.

It is important to emphasise that the nutritional screening and assessment tools were developed and validated for different populations of children, which were mainly focused on age limits and different exclusion criteria in terms of the admission diagnosis/status or other chronic illnesses.

RQ2: What is the validity of the screening and assessment tools versus the reference standard?

This systematic review evaluated 26 studies in all. The methodological quality of all of these was considered moderate. The main methodological problems were related to the lack of detailed descriptions of the study protocols, and to the non-blinded interpretation of index tests with regards to the reference standards, and/or vice versa. In particular, the study by Soundarya et al62 was inadequate in the reporting of information about the study protocol. There was no information about the study flow, the time frame and the number of people involved in the evaluation process. Additionally, the CANS score was used as the reference standard and not as the index test, as would be expected. Apart from the abovementioned study, four additional studies did not report the time frame of the study.39–41 57

Direct comparisons of the screening tools in terms of Se, Sp, PPV and NPV are not possible, as different cutoff values were used in the different studies. Also, the three malnutrition risk groups (ie, low, medium and high) were not uniquely combined into two groups for comparisons. Some studies used the combination of low to medium risk compared with high risk, while other studies used low compared with medium to high risk.

STAMP and STRONG_kids were the most often validated screening tools in the investigated studies. Anthropometry measurements were used as the reference standard in five studies for validation of STAMP, and in nine studies for validation of STRONG_kids. Clinical and surgical paediatric patients (n=223) were included in a study using BMI measurements as the reference standard, while comparing low and medium risk groups versus the high risk group.61 STAMP was validated with a good rating when performed by experts, but rated only fair with non-experts. Similar results were obtained on a larger sample in a multicenter study (n=2567) when validated with BMI and MUAC, but these were poor when validated with triceps skin fold thickness (TSFT).63 However, for comparisons of low risk versus medium and high risk groups, the validation studies reported poor ratings, with the exception of the study that included only a specifically small group of paediatric patients with SCI (n=45).6 When compared with full dietetic assessment, STAMP obtained fair ratings in three studies7 14 32 and poor ratings in the study on 60 outpatients.57 The modified version of STAMP in the last study here obtained a fair rating. Additionally, when validated with SGNA56 and ICD-10,58 the ratings were poor.

STRONG_kids obtained fair ratings when compared with anthropometric measurements in only two studies.53 58 The validation of STRONG_kids on the same group of patients as STAMP resulted in lower agreement with the BMI measurements regardless of the expertise of the assessor.61 Similar conclusions can be drawn from the larger multicenter study on 2567 patients.63

When validated with SGNA, STRONG_kids obtained the highest Se (100%); however, the Sp was very low (7.7%).56 Unfortunately, the results of the validation with anthropometry were not reported. The authors report only the prevalence of malnutrition based on the WHO criteria. Although the number of patients included was not so small (n=116), the authors reported a high percentage of children with oncological disorders (43.1%) and infectious diseases (14.6%), which resulted in a higher percentage of positive samples (28.4%). In the same study, PYMS compared with SGNA obtained the best results (Se=95.3%, Sp=76.9%, PPV=83.6%, NPV=93.0%). These results deviated quite a lot compared with STAMP and STRONG_kids in the same patient group. However, no conclusions can be drawn here as the results were reported only when compared with other screening/assessment tools, and not using an anthropometry measurement as the reference standard.56 Similar results were found in the study with only 46 patients with IBD.59

On the contrary, the CANS score was validated using anthropometry measurements (Roher’s Ponderal Index) on the largest group of newborns (n=14 477), and it gave a Sp of >90% and a Se of <50%.64 The CANS score showed a better performance compared with Roher’s Ponderal Index and BMI on a much smaller sample of 300 newborns.62 The authors of both studies concluded that the CANS score is useful for the identification of fetal malnutrition in newborns; however, the results from the larger sample did not confirm these statements. The second malnutrition screening tool for newborns, NNST, was also validated on a relatively large group of patients (n=909) using anthropometry measurements, where the UK growth charts obtained fair results (Se=89.6%, Sp=75.1%).42 However, the NNST was designed for a specific group of patients: neonatal patients in ICU who were >2 weeks old. The results showed good performance, although the low PPV indicates that two-thirds of the patients were unnecessarily wrongly predicted as at risk of malnutrition.

The PNST was validated in only one study with anthropometric measures (BMI ≤−2SD; Se=89.3%, Sp=66.2%).9 However, the SGNA was also validated with anthropometric measures (BMI≤−2SD) in the same study on the same group of patients. These results showed even better agreement (Se=96.5%, Sp=72.5%) although the very low PPVs indicated high overprediction of positive cases. The SGNA was also used as the reference standard for validation of PNST, which obtained a fair performance.

The PYMS screening tool was validated with anthropometric measurements in three studies,54 63 65 and it obtained a good rating compared with BMI in a multicountry study that included 2567 paediatric patients,63 with a poor rating in the other two smaller studies (n=300, n=119). When compared with STAMP and STRONG_kids, PYMS obtained the best results in the validation with BMI, for Se and Sp. Very low PPVs can be a source of concern, which was indeed noted by the authors of the tool in their first validation study using full dietetic assessments as the reference standard.7

When the PYMS, STAMP, STRONG_kids and PNRS were validated with ICD-10 as the reference standard, they showed poor agreement.59 This study included a small (n=46) and very heterogeneous group of children with IBD, and it did not reach any conclusions about the reasonability of the use of these tools for children with IBD.

The PYMS was also validated with the SGNA, which obtained a fair rating regarding the Se and Sp.56 High PPVs and NPVs also indicated that overprediction is not a cause for concern. As reported by the authors, PYMS obtained the best agreement with SGNA for the detection of acute malnutrition (k=0.3).

Three malnutrition screening tools specially designed for children with specific chronic illnesses were identified in this study39–41 (SCAN, Nutrition Risk screening tool in CF, Nutrition screening tool for paediatric patients with CF). None of these tools were validated with anthropometry measurements. SCAN was validated with SGNA,39 and obtained a poor rating due to weak identification of negative cases (Sp=39%); however, identification of positive cases was 100%. The PPV showed that >40% of the identified children were actually not exposed to the risk. The Nutrition Risk screening tool in CF was validated using full dietetic/nutritional assessment and the CFF consensus report criteria as the reference standard, which obtained a fair rating with high PPVs and NPVs, thus indicating a good performance.41 The second tool for patients with CF (ie, Nutrition screening tool for paediatric patients with cystic fibrosis) was also validated with the CFF consensus report criteria and obtained fair agreement, but with lower PPVs and NPVs, which thus indicated more problems with the overprediction of positive cases (>40%).40 Good agreement was reported when this tool was validated with the Nutrition Risk screening tool in CF (k=0.8).

Comparisons with similar systematic reviews

Similar reviews of validation studies on paediatric malnutrition screening tools can be found in the literature.8 66 67 However, the present systematic review includes the greatest number of paediatric malnutrition screening/assessment tools identified, and also in terms of the validation studies. The systematic reviews by Moeeni and Day68 and Hartman et al69 only focused on the description of the available paediatric malnutrition screening tools (six and five tools, respectively). Six paediatric malnutrition screening tools were described in the study by Joosten and Hulst,8 with eight validation studies included. The authors defined two tools as the most practical and reliable: STRONG_kids and PYMS. They proposed that STRONG_kids is the most reliable for assessing nutritional risk, and PYMS for assessing nutritional risk and actual nutritional status. A systematic review of studies that validated malnutrition screening tools for hospitalised children and included a meta-analysis was published by Huysentruyt et al.66 The systematic review included four malnutrition screening tools (PYMS, STRONG_kids, STAMP and PNRS) and 15 validation studies. As also observed in the present study, the authors were confronted with several problems when comparing the validation results from several studies. This was in particular due to the heterogeneity of the reference standards, the different cutoff points used and the small sample sizes (only one study had more than 100 participants). Their conclusions demonstrated that at the time there was insufficient evidence to choose one screening tool over another. The most recent systematic review was conducted by Teixeira and Viana,67 and this included five malnutritional screening/assessment tools (PYMS, STRONG_kids, STAMP, PNST and SGNA). The authors concluded that STRONG_kids and STAMP showed the best clinical performances in the studies included.

The results of the present systematic review are not consistent with the conclusions of Josten and Hulst8 and Teixseira and Viana,67 who recommended STRONG_kids as the most reliable screening tool; here, PYMS showed better performance (table 3).

Strengths and limitations

There are some key limitations to the present study that have to be emphasised. The most important limitation comes from the lack of a gold standard for evaluation of the malnutrition risk of hospitalised children. Consequently, the studies used different reference standards, most often as anthropometric measures or dietetic assessments. As reported in a number of studies,70–72 the accuracy of the anthropometric measurements was often poor, which resulted in questionable uniformity of the validation of these screening/assessment tools. Full dietetic/nutritional assessment also varied due to the different methods used and the different educational standards for dietetics in different countries.3 Some studies used another screening/assessment tool for the reference standard, which introduces a certain source of bias. As observed in the present study, the PPVs in the validation studies that compared one screening tool to another were higher than for the same screening tool compared with the anthropometry measures. However, this can be expected if both of the screening tools overpredict positive cases when validated with the same anthropometry measurements as reference standard. The choice of the reference standard, therefore, represents a source of bias to the original studies, and consequently also to the present systematic review.

Another limitation comes from the different inclusion and exclusion criteria used in the studies, and the lower power of the studies with small sample sizes. The protocols of the studies were not uniquely defined, so the studies differed in the type and number of assessors using the nutritional screening tool studied. Similarly, different types and numbers of assessors were involved in evaluation of the malnutrition risk, according to the study-defined reference standard. Also, almost one-third of the studies evaluated (30.8%)6 14 39–41 52 57 59 had a sample size of <100 patients. These small numbers of paediatric patients involved mandate caution when generalising the results to the full population.

However, one of the main strengths of the present systematic review is that we focused only on studies that included validation of the screening/assessment tools used based on the chosen reference standard. We are confident that all of the currently available nutritional screening/assessment tools for paediatric patients are included in this review. This study thus presents a complete review of the success of these tools in the prediction of malnutrition risk.

Recommendations

The results of this study show the need for the definition of a gold standard. We propose that an expert group is formed to discuss and define which reference standard should be used worldwide for the evaluation of screening tools. Based on the deficiencies identified in validation studies during the present systematic review, we recommend that the reference standard should never be another screening/assessment tool. We propose that the anthropometric measurements defined in the WHO growth charts/Anthro/ AnthroPlus software or the CDC 2000 growth charts/Epi Info 7 should be considered as the basic reference standard for the purpose of fair comparisons. The American Society for Parenteral and Enteral Nutrition (ASPEN)73 recommendations included the recording of weight, height, BMI and MUAC, and considered the TSFT and mid-arm muscle circumference on admission, with reference to the appropriate growth chart. Head circumference must also be obtained in infants younger than <2 years. The CDC74 and ASPEN73 recommend using the WHO charts for children up to 2 years of age. For children and adolescents from 2 to 20 years of age, they recommend using the CDC 2000 charts. The newest versions of growth charts should also be used. Anthropometric measurements should be performed using calibrated equipment, according to the examination protocols described by the CDC.75 As validation of malnutrition screening tools using anthropometry measurements as the reference standard tend to produce a lot of false-positive results, our recommendation is to use full dietetic/nutritional assessment in the second stage only for the positively identified cases. However, the dietetic/nutritional assessment process and evaluation should also be standardised first. Furthermore, specific disease conditions can cause energy and/or protein imbalances, and therefore these should be considered in the final evaluation of the nutritional status.73 Validation studies should also test several malnutrition screening tools on the same population, to avoid bias due to different patient populations, disease backgrounds or age groups.31

It is important that healthcare professionals who perform nutritional screening are appropriately educated and trained in the measurement of the anthropometric parameters, and they should use the appropriate growth charts or computer software, and the chosen screening tool.

The study protocol should be carefully designed and followed, with particular attention paid to the following:

• Patient selection: The sample of patients included in studies should be as large as possible, with consecutive or random sampling used, and inappropriate exclusions avoided. The inclusion and exclusion criteria should be carefully defined. All of the patients included in a study should be included in the validation procedures.

• Flow and timing: To obtain an evaluation of the nutritional status that is as objective as possible, the evaluator using the screening test and the reference standard should not be the same person. Additionally, at the time of performing the evaluation of the nutritional status, the evaluators should not be acquainted with the results of other evaluations. The exact sequence of the performing of the screening tests and the reference standards has to be defined in the study protocol, while taking into consideration that both the screening test and the reference standard should be evaluated on admission73 and on the same day (or as close as possible). This will avoid possible changes in the patient health condition in between these evaluations. The exact flow and timing of these should also be reported.

• Reporting results: All of the traditional evaluation metrics should be reported, such as Se, Sp, PPV and NPV. When the malnutrition risk is evaluated using three categories (ie, low/moderate/high), the moderate and high risks should both be treated as ‘at risk’. A table showing the cross-classification of the malnutrition risk on the screening test compared with the reference standard should also be reported. When several evaluators use the screening test or the reference standard, the inter-rater or intra-rater agreement should also be reported.