Development of a nutritional screening or assessment tool using a multivariate technique

The aims of this paper are to consider the important design stages required to develop a nutritional screening or assessment tool and to present an analysis for identifying important risk factors associated with malnutrition. Design issues, such as sample selection and sample size, are addressed in relation to the study protocol. The various steps involved in tool development are outlined. Appropriate univariate and multivariate techniques are discussed with the non-statistician in mind. Emphasis is placed on the linear logistic regression method. Practical suggestions relating to the development of a tool for nutritional risk assessment are proposed, and methodology is illustrated with data from a published study. Guidelines for the standardization of the design, analyses, and publication of results are presented.