Nutrition screening tools: Does one size fit all? A systematic review of screening tools for the hospital setting

Summary

Background & aims

Numerous nutrition screening tools for the hospital setting have been developed. The aim of this systematic review is to study construct or criterion validity and predictive validity of nutrition screening tools for the general hospital setting.

Methods

A systematic review of English, French, German, Spanish, Portuguese and Dutch articles identified via MEDLINE, Cinahl and EMBASE (from inception to the 2nd of February 2012). Additional studies were identified by checking reference lists of identified manuscripts. Search terms included key words for malnutrition, screening or assessment instruments, and terms for hospital setting and adults. Data were extracted independently by 2 authors. Only studies expressing the (construct, criterion or predictive) validity of a tool were included.

Results

83 studies (32 screening tools) were identified: 42 studies on construct or criterion validity versus a reference method and 51 studies on predictive validity on outcome (i.e. length of stay, mortality or complications). None of the tools performed consistently well to establish the patients' nutritional status. For the elderly, MNA performed fair to good, for the adults MUST performed fair to good. SGA, NRS-2002 and MUST performed well in predicting outcome in approximately half of the studies reviewed in adults, but not in older patients.

Conclusions

Not one single screening or assessment tool is capable of adequate nutrition screening as well as predicting poor nutrition related outcome. Development of new tools seems redundant and will most probably not lead to new insights. New studies comparing different tools within one patient population are required.

Introduction

Over the last decades numerous nutrition screening tools for use in the hospital setting have been developed, with the purpose to facilitate easy screening or assessment of a patient's nutritional status or to predict poor clinical outcome related to malnutrition. Some of the tools have been endorsed by international nutrition societies; e.g. the European Society for Clinical Nutrition and Metabolism advises the use of MUST,¹ NRS-2002² and the MNA(-SF)3, 4 for the elderly. Other tools are widely used in certain countries but less frequently applied worldwide (e.g. MST for Australia and New Zealand⁵ and SNAQ for the Netherlands⁶). Some tools claim to be valid for all populations, ages and settings, whereas others have been developed for screening a specific target population. In addition, there probably are many unpublished, not validated local tools that we are unaware of.

There is no international consensus on a single ‘best tool’, if there is so such thing. The use of different tools in different studies hinders the comparison between studies and does not allow for the drawing of conclusions on defining the ‘best tool’ for a certain patient population, age group or setting.

The purpose of this study is to systematically review the publications on screening and assessment tools and to study the validity of these tools for the general (adult and older) hospital population. This review will give an overview of the available instruments, and of the ability of tools to assess the patient's nutritional status or to predict the clinical outcome. Finally, the results of our comparisons of tools will be presented.

Most of the available nutrition screening tools have been developed to obtain an indication of a patient's nutritional status. Well-known examples include MST,⁵ SNAQ,⁶ NRI,⁷ MUST,¹ SGA⁸ and MNA.³ Within this range, some of the tools are ‘quick and easy’; not requiring any calculations, blood samples, anthropometric measurements, or clinical examinations (e.g. MST,⁵ SNAQ⁶). A patient indicated to be at high nutritional risk by one of these tools, requires further nutritional assessment by a professional to get a more complete indication of the severity and the nature of the nutritional depletion. Such tools are typically called ‘screening tools’. Other tools are more complex, requiring – for example – calculation of a BMI and/or an indication of disease severity (e.g. MUST¹), clinical assessment (e.g. MNA³), or an extensive questionnaire addressing several aspects of nutritional intake (e.g. NRI⁷). These tools are more time-consuming, but on the other hand they give a better estimation of (the background of) a patient's nutritional status. Some of these tools are still regarded as screening tools (e.g. MUST,¹ NRI⁷), whereas others are qualified as assessment tools because they combine data on nutritional status with clinical observations (e.g. medical examination, evaluation of cognitive function), disease status and/or laboratory values (e.g. SGA,⁸ MNA³).

The terms “screening” and “assessment” are often used interchangeably in both literature and practice. While in this manuscript both terms will also be used, the original purpose (screening or assessment) will be explained where appropriate.

In the absence of a gold standard for malnutrition, most of the screening and assessment tools have been developed with assessment by a professional or a full nutritional assessment as the reference method. Also, the lack of a ‘gold’ reference method has resulted in the use of many of the existing screening and assessment tools as the reference method, where the method considered to be the reference method is always superior to the tool to be validated. Because of the natural superiority of the reference tool this becomes confusing when, for example, in one study the NRS-2002 is validated with the SGA as a reference,⁹ while in another the SGA is validated against the NRS-2002.¹⁰

The NRS-2002² is generally used as a screening tool, while, in fact, it was designed as a tool to identify patients at nutritional risk. NRS-2002 was developed differently from other tools. It was developed from a literature overview including 275 studies reporting on the effectiveness of nutritional intervention and its purpose was to identify malnourished hospitalized patients likely to benefit from nutritional support.

Few tools have been designed specifically with the purpose to predict clinical outcome (morbidity, mortality, (postoperative) complications, or length of hospital stay), for example GNRI.¹¹ MUST¹ and SGA⁸ have been developed both to identify patients at nutritional risk and to predict outcome.

Next to the different purpose of tools, some tools were originally developed for certain subgroups of patients or for certain settings. The MNA³ and GNRI,¹¹ for example, have been developed specifically for the elderly. However, in practice, studies have applied all tools for all purposes; tools designed to assess nutritional status are used to predict outcome and vice versa, and tools for the elderly are also applied to the younger hospitalized patients and the other way around. MNA,³ developed within a group of frail and healthy elderly is most frequently used for hospitalized or nursing home patients. Thus there is a variety of studies at our disposal, applying all kinds of tools in all kinds of populations and all kinds of settings, with different results. This requires a structured approach to rate the validity of the different tools for the different purposes, which we attempt to give in this systematic review.

This systematic review will answer the following research questions:

• How good is the performance of a tool in assessing patients' nutritional status?

• How well can a tool predict clinical outcome?

• Are one or more tools superior to other tools when applied in the same population?

To assess which tool is preferred, the latter research question is the most important one.

The research questions will be answered for both the adult and the elderly hospitalized population.

Studying the validity of a tool is usually done versus a gold standard. In the absence of a perfect gold standard for malnutrition, studies use different reference methods to validate their tools. Roughly the following main reference methods were identified:

• objective assessment by a professional

• nutritional assessment and anthropometry

• another screening or assessment tool

• other reference methods

For this review we decided to consider the following methods ‘valid’ reference methods:

• objective assessment by a professional

• nutritional assessment and anthropometry

• the assessment tools MNA³ and SGA⁸

For these comparisons we use the term criterion validity.

The following methods were therefore considered to be less valid reference methods:

• any of the screening tools (e.g. MUST,¹ NRS-2002,² PG-SGA,¹² because screening tools require a further assessment by a professional), and

• laboratory values like pre-albumin and albumin (as these parameters reflect acute disease more than nutritional status)

For these comparisons we use the term construct validity. Still, many studies have used these less valid methods as a reference. Since an ideal gold standard is missing, and (research) groups may differ in their opinion on the most optimal reference method, we have chosen to include all studies, allowing the readers to decide for themselves how valid they rate a tool.

When validating a new tool versus a reference method, one should keep in mind that the new tool can never be better than the reference method. Thus, there should be convincing reasons to develop a new tool, such as: the old (reference) tool being too invasive, or too time consuming.

The majority of studies assesses the ability of a tool to predict clinical outcome. Studies report on length of stay, mortality, or (postoperative) complications. Some studies focus on only one of these clinical outcomes, whereas others address more (or even all) outcomes, sometimes with conflicting results. It is important to note that these outcomes are influenced by more facts than nutrition alone. Therefore studies in which outcome was adjusted for other factors, such as age and disease severity, are regarded to be of higher quality than those presenting unadjusted data.