Are Machine Learning Algorithms More Accurate in Predicting Vegetable and Fruit Consumption Than Traditional Statistical Models? An Exploratory Analysis.

Machine learning (ML) algorithms may help better understand the complex interactions among factors that influence dietary choices and behaviors. The aim of this study was to explore whether ML algorithms are more accurate than traditional statistical models in predicting vegetable and fruit (VF) consumption. A large array of features (2,452 features from 525 variables) encompassing individual and environmental information related to dietary habits and food choices in a sample of 1,147 French-speaking adult men and women was used for the purpose of this study. Adequate VF consumption, which was defined as 5 servings/d or more, was measured by averaging data from three web-based 24 h recalls and used as the outcome to predict. Nine classification ML algorithms were compared to two traditional statistical predictive models, logistic regression and penalized regression (Lasso). The performance of the predictive ML algorithms was tested after the implementation of adjustments, including normalizing the data, as well as in a series of sensitivity analyses such as using VF consumption obtained from a web-based food frequency questionnaire (wFFQ) and applying a feature selection algorithm in an attempt to reduce overfitting. Logistic regression and Lasso predicted adequate VF consumption with an accuracy of 0.64 (95% confidence interval [CI]: 0.58-0.70) and 0.64 (95%CI: 0.60-0.68) respectively. Among the ML algorithms tested, the most accurate algorithms to predict adequate VF consumption were the support vector machine (SVM) with either a radial basis kernel or a sigmoid kernel, both with an accuracy of 0.65 (95%CI: 0.59-0.71). The least accurate ML algorithm was the SVM with a linear kernel with an accuracy of 0.55 (95%CI: 0.49-0.61). Using dietary intake data from the wFFQ and applying a feature selection algorithm had little to no impact on the performance of the algorithms. In summary, ML algorithms and traditional statistical models predicted adequate VF consumption with similar accuracies among adults. These results suggest that additional research is needed to explore further the true potential of ML in predicting dietary behaviours that are determined by complex interactions among several individual, social and environmental factors.

Artificial intelligence in nutrition research: perspectives on current and future applications.

Artificial intelligence (AI) is a rapidly evolving area that offers unparalleled opportunities of progress and applications in many healthcare fields. In this review, we provide an overview of the main and latest applications of AI in nutrition research and identify gaps to address to potentialize this emerging field. AI algorithms may help better understand and predict the complex and non-linear interactions between nutrition-related data and health outcomes, particularly when large amounts of data need to be structured and integrated, such as in metabolomics. AI-based approaches, including image recognition, may also improve dietary assessment by maximizing efficiency and addressing systematic and random errors associated with self-reported measurements of dietary intakes. Finally, AI applications can extract, structure and analyze large amounts of data from social media platforms to better understand dietary behaviours and perceptions among the population. In summary, AI-based approaches will likely improve and advance nutrition research as well as help explore new applications. However, further research is needed to identify areas where AI does deliver added value compared with traditional approaches, and other areas where AI is simply not likely to advance the field. Novelty: Artificial intelligence offers unparalleled opportunities of progress and applications in nutrition. There remain gaps to address to potentialize this emerging field.

Changes in diet quality and food security among adults during the COVID-19-related early lockdown: results from NutriQuébec.

BACKGROUND: The impact that the coronavirus disease 2019 (COVID-19)-related early lockdown has had on dietary habits of the population and on food insecurity is unknown. OBJECTIVE: The aim of this study was to document the change in diet quality and in food insecurity observed during the COVID-19-related early lockdown. We hypothesized that the lockdown was associated with a deterioration in overall diet quality and an increase in food insecurity. METHODS: Data are from a COVID-19 subsample of NutriQuébec, a web-based cohort destined to study temporal changes in dietary habits among adults in Quebec, Canada. Participants completed questionnaires before (between June 2019 and February 2020) and during (April to May 2020) early lockdown, including a validated web-based 24-h recall (n = 853) and a questionnaire on food security (n = 922). Primary study outcomes were temporal changes in diet quality measured by the Healthy Eating Index (HEI)-2015 and in the prevalence of food insecurity. RESULTS: There was a small increase in the HEI-2015 during the COVID-19 early lockdown compared with baseline (+1.1 points; 95% CI: 0.6, 1.5), mostly due to small improvements in the intakes of whole grains, greens and beans, refined grains, total vegetables, total dairy, seafood and plant proteins, added sugar, and total protein subscores of the HEI-2015. Exploratory analyses suggested that individuals aged 18-29 y (+3.6 points; 95% CI: 2.4, 4.7), participants with lower education (+1.9 points; 95% CI: 1.3, 2.6), or with obesity (+3.8 points; 95% CI: 2.7, 4.8) showed particularly important increases in the HEI-2015. The prevalence of food insecurity was reduced from 3.8% at baseline to 1.0% during the early lockdown (prevalence ratio = 0.27; 95% CI: 0.08, 0.94). CONCLUSIONS: Contrary to our hypotheses, diet quality has slightly improved and prevalence of food insecurity was reduced in this sample of adults from Quebec during the COVID-19-related early lockdown. These results may be generalizable only to relatively healthy populations.

A cross-sectional survey examining motivation and beliefs to participating in a web-based prospective cohort study on nutrition and health among individuals with a low socioeconomic status.

BACKGROUND: Prospective cohort studies may support public health efforts in reducing health inequalities. However, individuals with a low socioeconomic status (SES) are generally underrepresented in health research. This study aimed to examine the intention and determinants of intention of individuals with a low SES towards participation in a Web-based prospective project on nutrition and health (NutriQuébec) in order to develop recruitment and retention strategies. METHODS: A cross-sectional survey based on the Theory of planned behaviour was conducted in the Province of Québec, Canada. Low SES individuals (high school or less and annual household income < $55,000 CAN) were recruited through a Web panel of a polling firm to assess intention, attitude, subjective norm and perceived behavioural control (PBC) towards participation in the NutriQuébec project. Linear regression and logistic regression analyses were conducted. RESULTS: Mean age of respondents (184 women, 141 men) was 57.6 y (SD = 13.6). Attitude (ß = 0.54, 95%CI: 0.41-0.68) and PBC (ß = 0.50, 95%CI: 0.37-0.63) were significantly associated with intention. Participants who agreed that participating in the study would contribute to an improvement in 1) collective health (odds ratio [OR] = 2.15, 95%CI: 1.27-3.64) and in 2) one's lifestyle habits (OR = 1.70, 95%CI: 1.04-2.78) were more likely to express positive intention compared to participants who did not agree with these statements. Participants who agreed to participate in the study even 1) in the absence of a financial incentive (OR = 1.43, 95%CI: 1.04-1.99) and even 2) if the completion of questionnaires took up to two hours (OR = 1.78, 95%CI: 1.27-2.48) were also more likely to express high intention. Receiving a personalized brief health assessment (OR = 1.61, 95%CI: 1.13-2.30) and the use of simple questions in the questionnaires (OR = 1.54, 95%CI: 1.05-2.25) were facilitating factors associated with high intention. Participants believing that participation would be too time-consuming were less likely to have positive intention (OR = 0.57, 95%CI: 0.43-0.75). CONCLUSIONS: The development of a positive attitude and a high PBC towards participation in the NutriQuébec project will be necessary to obtain representative data of low SES adults.

Beliefs Related to Participation in a Large Web-Based Prospective Survey on Diet and Health Among Individuals With a Low Socioeconomic Status: Qualitative Study.

BACKGROUND: NutriQuébec is a Web-based prospective study on the relationship between diet and health as well as the impact of food-related health policies in the adult population of Québec, Canada. Recruitment and retention of individuals with a low socioeconomic status (SES) in such a study are known to be challenging, yet critical for achieving representativeness of the entire population. OBJECTIVE: This study aimed to identify the behavioral, normative, and control beliefs of individuals with a low SES regarding participation in the NutriQuébec project and to identify their preferences regarding recruitment methods. METHODS: A total of four focus groups were conducted in community centers located in low-income areas of Québec City, Canada. On the basis of the theory of planned behavior, participants' beliefs associated with attitude, subjective norm, and perceived behavioral control regarding hypothetical participation in the NutriQuébec project were identified. Focus groups were recorded, transcribed, and coded by two analysts. RESULTS: Participants (16 men and 12 women) were aged between 28 and 72 years, and a majority of the participants had an annual household income of Can $19,999 or less. The main perceived advantages of participating in the NutriQuébec project were contributing to improved collective health and supporting research. The only disadvantage identified was the risk of having to fill out too many questionnaires. Participants could not, in general, identify persons from their entourage who would approve or disapprove their participation in the study. The main facilitators identified were obtaining a brief health assessment and the ability to complete questionnaires in a way that is not Web-based. The main barrier was the lack of internet access. The preferred means of recruitment were through social media, television, and community centers. CONCLUSIONS: These results provide insightful information regarding the best methods and messages to use in order to recruit and retain individuals with a low SES in a population-based prospective study on lifestyle and health on the internet.