本研究目的為發展適用國小高年級以上學童自填的中式飲食頻率問卷(Chinese food frequency questionnaire: CFFQ)，期望未來能大量收集學童長期的飲食攝取資料。研究前進行國小學童之飲食評估質性觀察，作為CFFQ設計之參考。CFFQ架構因考慮學童之回憶能力，分為三餐次、點心及飲料五大類，研究將衛生署「家庭飲食與健康研究」收集之24小時飲食回憶(24-h recall)資料的625項食物歸類為96類食物項目後再計算各食物項目的營養素貢獻率，依據營養素貢獻率決定放入CFFQ的飲食項目共80項。CFFQ之飲食頻率為從未～大於7次／週共6種，飲食份量採單一份量。研究為了解CFFQ內容設計的適當性，將原24h-recall(n=110)的飲食資料填入CFFQ後，選擇主要的10種營養素攝取量比較CFFQ與原24h-recall之計算結果。CFFQ預試，由古亭國小五、六年級學童(n=55)自塡FFQ及1天飲食記錄，將兩種評估方式結果進行相對效度分析。統計方法以paired-t-test及Spearmon Correlation比較營養攝取量之差異及相關性。研究結果發現：1.評估CFFQ內容設計之適當性所選譯的10個營養素，攝取量評估值的平均正確率為92％，營養素攝取量相關性為0.37～0.78(P＜0.01)。2.國小五、六年級學童大部分能自塡CFFQ，審閱評定為無效問卷者佔29％。3.學童塡寫之1天飲食記錄結果普遍低估，只有維生素C及鈣之攝取量在兩種評估法達顯著相關(r=0.34 & 0.37, p＜0.05)。未來信效度研究應增加樣本數及飲食記錄的天數，以利此問卷進一步之發展。

The purpose of this study was to develop a self-administered Chinese food frequency questionnaire (CFFQ) for school children. The development for the CFFQ included three stages. The first stage was to understand the pros and cons of various dietary assessments in children. The second stage was to collect the children's dietary data by 24-h recalls and to design the contents of the CFFQ. Both food list and standardized portion sizes were developed using 24h-recall data from 197 school children respondents from "Family diet, nutrition and health study in Taipei metropolitan area". The CFFQ divides into five parts including breakfast, lunch, dinner, snacks and drink in consideration of the memory ability of children. Foods reported by these children (625 different foods) were grouped into 96 similar food items. Total 80 food items and standardized portion sizes were decided on the basis of their contribution to intakes of energy, protein, carbohydrate, fat, cholesterol, vitamin B1, vitamin B2, vitamin C, Fe and Ca, total 10 key nutrients of the study population, and represented over 90% of each of these nutrients. The third step was to analyze the adequacy of food lists and portion size and conduct the pilot study of CFFQ. The results showed the mean nutrient estimates accuracy rate for 10 selected key nutrients was 92%. The correlation coefficients between nutrients estimated by 24-h recalls and CFFQ we4re 0.34~0.78 (p<0.01). The results of this study are following: First, senior grades of school children (mean age: 11.3±0.3) could complete FFQ in 15-30 minutes by self. Second, the invalid questionnaires are 29%. Third, because one-day food records by children had lower estimate generally, the correlation coefficients for energy and 16 nutrients intake ranged from 0.37 for calcium to 0.01 for niacin. Only vitamin C and calcium were significantly correlated with 2 methods (p<0.05). In conclusion, CFFQ has become popular for the assessment of usual diet intake. This questionnaire could be self-administered by school children over 10 years old. The validity of the questionnaire remains to be further tested.