A preliminary validation of a nutritional risk measure for the elderly.

We used data from a two-stage random sample of 401 noninstitutionalized elderly individuals residing in 18 census tracts in south-central metropolitan St. Louis to establish the reliability and validity of a 16-item nutritional risk measure. Reliability analysis yielded a Cronbach's alpha of 0.603. Concurrent predictive validity was demonstrated by the prediction of physician, emergency room, and hospital use. Individuals with high scores displayed a consistent pattern of higher rates of known nutrition-related problems and their sequelae than did those with low scores, providing evidence of known groups validity. Statistical correlations with components of an established theoretical model of the elderly's health and illness behavior demonstrate preliminary support for construct validity. The advantages of our nutritional risk measure include its ease of administration and telephone portability.