Gemesi Kathrin, Holzmann Sophie Laura, Böhm Markus, Leipold Nadja, Hauptmann Hanna, Lurz Martin, Groh Georg, Hauner Hans, Krcmar Helmut, Holzapfel Christina
Institute for Nutritional Medicine, School of Medicine and Health, TUM University Hospital, Technical University of Munich, Munich, Germany.
Department of Informatics, University of Applied Sciences Landshut, Landshut, Germany.
BMJ Nutr Prev Health. 2025 Jun 20;8(1):e001162. doi: 10.1136/bmjnph-2024-001162. eCollection 2025.
Photo-based nutrition diaries might be useful to assess dietary intake without much effort and maybe even without nutrition expertise. This proof-of-concept study aimed to investigate the principle of 'the wisdom of crowds' by examining how accurately both nutrition experts and non-experts can rate nutritional values of meals presented in digital pictures.
An online survey was conducted among adults in Germany from 2016 to 2017. Participants rated a random selection of six meal pictures according to their nutritional values (energy, carbohydrates, sugar and fat content) and their healthiness. Rating results were compared with the nutritional value calculated by the German Nutrient Database or according to the manufacturer's information ('truth'). Descriptive statistical analysis, Mann-Whitney-U test and multiple linear regression analysis were performed using RStudio.
In total, 110 (92.7 % women, mean age: 38.7±14.0 years) nutrition experts and 233 (31.3 % women, mean age: 21.2±2.6 years) non-experts participated. Overall meal pictures, experts overestimated the average content of all nutritional values (sugar: 3.8 %, energy: 4.9 %, carbohydrates: 4.9 % and fat: 10.4 %). Non-experts overestimated the average energy content by 10.4 %, fat content by 17.1 % and sugar content by 27.5%. The average carbohydrate content was underestimated by 9.0%. A statistically significant difference between the two crowds' ratings was found for energy (p=0.03), carbohydrates (p<0.001) and sugar (p<0.001), but not for fat (p=0.44). An increasing deviation of nutritional value ratings from the truth towards overrating was associated with decreasing rating of healthiness (all p<0.001).
This study suggests that both experts and non-experts rate nutritional values of meals in digital pictures in an appropriate manner, although both crowds occasionally deviated significantly from the truth, especially over-rating occurred with decreasing rating of healthiness. Due to the proof-of-concept approach and the limited generalisability of the results, the principle of 'the wisdom of crowds' is not fully supported. Studies with a large representative population are necessary. However, findings suggest that crowd-based meal picture ratings could be a method of digital dietary self-monitoring in combination with gamification elements.
基于照片的营养日记可能有助于轻松评估饮食摄入量，甚至可能无需营养专业知识。这项概念验证研究旨在通过考察营养专家和非专家对数字图片中呈现的餐食营养价值的评估准确性，来探究“群体智慧”原则。
2016年至2017年对德国成年人进行了一项在线调查。参与者根据餐食的营养价值（能量、碳水化合物、糖和脂肪含量）及其健康程度，对随机抽取的六张餐食图片进行评分。将评分结果与德国营养数据库计算得出的或根据制造商信息得出的营养价值（“真相”）进行比较。使用RStudio进行描述性统计分析、曼-惠特尼-U检验和多元线性回归分析。
共有110名（92.7%为女性，平均年龄：38.7±14.0岁）营养专家和233名（31.3%为女性，平均年龄：21.2±2.6岁）非专家参与。对于整体餐食图片，专家高估了所有营养价值的平均含量（糖：3.8%，能量：4.9%，碳水化合物：4.9%，脂肪：10.4%）。非专家高估了平均能量含量10.4%，脂肪含量17.1%，糖含量27.5%。碳水化合物平均含量被低估了9.0%。发现两组人群在能量（p = 0.03）、碳水化合物（p < 0.001）和糖（p < 0.001）评分上存在统计学显著差异，但在脂肪评分上无差异（p = 0.44）。营养价值评分与真相的偏差朝着高估方向增加与健康程度评分降低相关（所有p < 0.001）。
本研究表明，尽管两组人群偶尔会与真相有显著偏差，尤其是随着健康程度评分降低出现高估，但专家和非专家都能以适当方式对数字图片中餐食的营养价值进行评分。由于采用了概念验证方法且结果的普遍性有限，“群体智慧”原则未得到充分支持。有必要进行具有大量代表性人群的研究。然而，研究结果表明，基于群体的餐食图片评分可能是一种结合游戏化元素的数字饮食自我监测方法。
