Comparing the results of a cholesterol and saturated fat screener when using two different scores Sabine Rohrmann 1 & Gisela Klein 2 1 Division of Clinical Epidemiology, Deutsches Krebsforschungszentrum, Heidelberg; 2 Institut fuer Medizinische Biometrie, Epidemiologie und Informatik, University of Mainz, Germany Accepted in revised form 15 October 2002 Abstract. Screener are useful instruments for cate- gorising individuals according to their nutrient in- take. To accurately classify the individuals it is important to correctly analyse the screener. Therefore we compared a simple and a weighed score of a cholesterol and saturated fat (SF) screener with the cholesterol and SF intake, respectively, calculated from an extensive food frequency questionnaire. The validation results did not change when a weighed score was used instead of the simple one, indicating that a simple score does not introduce a higher amount of misclassification. Key words: Cholesterol, Saturated fat, Score, Screener, Validation Dietary assessment methods are often complicated, expensive, and time-consuming. Additionally, some- times it is not necessary to compute the nutrient in- take in detail. Instead screening instruments that quickly classify subjects into broad categories of low, medium and high intake according to their intake of one or a few specific nutrients can be helpful [1]. Such screener should, per definition, be very simple. As a consequence, some information is going lost, e.g. portion size or less important food items. The chal- lenge is to correctly classify individuals despite the shortness of the instrument. Thus, it was the purpose of this study to compare two different scoring systems of a screening instrument with the cholesterol and SF intake assessed by means of an extensive food fre- quency questionnaire (FFQ) and to investigate the question whether a simple or a weighed score is more appropriate to categorise subjects according to their cholesterol and SF intake. Our cholesterol and SF screener qualitatively as- sesses the dietary cholesterol and SF intake and consists of 16 food items. The cholesterol score in- cludes nine items, the SF score 11 (Table 1). The consumption frequency of the food items was as- sessed by five categories (ranging from ‘once a day or more’ to ‘once a month, less, or never’). All answers in the first category (once a day or more) get 5 points, all answers in the second column (4–6 times a week) get 4 points and so on. A simple score was obtained by adding up the points corresponding to the marked consumption frequency of each item. A high score indicates a high cholesterol and/or SF intake. Addi- tionally, a weighted score was calculated by multi- plying the marked consumption frequency (the points) by the cholesterol and SF content, respec- tively, of a medium portion size. The reference in- strument was a 148-items FFQ, which was applied during the recruitment of the European Prospective Investigation into Cancer and Nutrition (EPIC) in Heidelberg, Germany, between 1994 and 1998 [2, 3], From the FFQ we calculated the cholesterol as well as the SF intake per day. Spearman’s rank correla- tion coefficient was calculated between the intake from the FFQ and the scores of the cholesterol and SF screener. The ability of the screener to correctly classify people according to their intake was mea- sured by cross-classification in quartiles. Additionally, we compared the single with the weighed score for each nutrient. The screener was sent to 300 randomly selected participants of EPIC in Heidelberg, Germany, in October 1999. The response rate was 81.3% (244 of 300), 46.3% of the respondents were men, 53.7% women. The median age of the participants when they filled in the FFQ was 51 years, ranging from 35 to 64 years. For both nutrients, the weighed score correlated slightly better with the FFQ results than the simple score (Table 2). While the results of the quartile’s cross-classification are better for the cho- lesterol than for the SF intake, the results are similar when comparing the scores of each of the nutrients with each other. Our results indicate that weighing the items of the screener with the help of their cholesterol and SF content, respectively, neither improved the correla- tion between the scores and the nutrients intake as- sessed from the FFQ nor the categorisation of participants. Thus, using simple scores did not lead to a higher portion of misclassified subjects in our study and the high agreement between both scores indicates European Journal of Epidemiology 18: 413–415, 2003. Ó 2003 Kluwer Academic Publishers. Printed in the Netherlands.