Use of Internal Validity in the Construct of an Index of Undernutrition

Abstract

An index of undernutrition (IOU) has been developed from a crosssectional study of 200 general surgical patients. The IOU was constructed using a ranking method based upon the relative values of anthropometric measurements (weight loss, mid-arm muscle area, mid-arm fat area) and serum protein concentrations (albumin, transferrin). The IOU was derived without reference to the incidence of adverse clinical events and took the form of a boolean cluster. A subsequent prospective longitudinal study evaluated the IOU in 367 general surgical patients at the time of admission to hospital. The results of this study supported the validity of the IOU. There was no association between gender and IOU scores. However, patients with high scores were more likely to be elderly (p less than 0.02), have cancer (p less than 0.001), stay in hospital longer (p less than 0.001), or experience an adverse clinical event after surgery (p less than 0.001). In accord with other published indices of undernutrition, the association between the extent of undernutrition and the incidence of adverse clinical events was only of moderate diagnostic potential (overall accuracy 68%). This study demonstrates that it is possible to derive an index of protein-energy undernutrition that is independent of the incidence of adverse clinical events.