Application of Logistic Regression and Random Forests to Assess the Relevance of Chrononutrition Information for Prediction of Overweight in Adults: Evidence from the INRAN-SCAI 2005-2006 Italian Nutrition Survey.

May 27, 2026Nutrients

Using Logistic Regression and Random Forests to Predict Adult Overweight Based on Meal Timing and Nutrition: Results from the 2005-2006 Italian Nutrition Survey

AI simplified

Circadian Biology on OpenScience ↗PubMed ↗DOI ↗

Abstract

An analysis of data from 2,312 adults indicates that models using chrononutritional information slightly improved the prediction of overweight status compared to those using daily energy intake alone.

Logistic regression models showed better performance with chrononutrition data, but the differences in diagnostic accuracy were small and possibly due to overfitting.
The area under the curve (AUC) for models trained on chrononutrition was 0.7909 for mean and irregularity of calorie intake across 6 time intervals compared to 0.7850 for whole-day intake.
Random forest models did not show significant improvements with the inclusion of chrononutritional information.
In test sets, models using chrononutrition data performed significantly better than those using whole-day intake only.
Socio-demographic and behavioral variables may provide insight into the timing of energy intake, suggesting that the health impact of calorie timing may not be solely related to overweight.

AI simplified

: Obesity represents a growing public health concern worldwide. Chrononutrition, a research field examining the timing and regularity of food intake, has been shown in animal models to influence body weight regulation and obesity-related outcomes. Previous research has also explored the association between chrononutrition information and BMI. Using INRAN-SCAI 2005/2006 adult nutrition data based on 3-day diet diaries (= 2312), this study aims to assess whether chrononutritional information on the distribution of energy intake during the day is able to improve prediction of overweight status (BMI > 25 kg/m), compared to information on energy from the whole day alone.: This research investigates it using logistic regression and random forest models. For both types of models, three different specifications were compared: a model trained on the mean and irregularity of calorie intake over 3 days for 6 day-time intervals (MI6); a model trained on repeated measures over 3 days of calorie intake from the same 6 time intervals (RM); and a model trained on mean and irregularity of calorie intake over 3 days for the whole day (MID). The performance of the models was compared using risk prediction metrics and ROC curves.When including additional demographic and behavioural predictors beside the energy variables, the results only showed a statistically significant difference in the performance of the logistic regression models if they were trained and tested on the same data. The models trained using chrononutrition information performed better, but the difference in diagnostic accuracy was very small (AUC = 0.7909 for MI6,= 0.0086; 0.7923 for RM compared to 0.7850 for MID,= 0.0072) and possibly attributable to overfitting, as it was no longer significant in the comparison within a testing set (70% training and 30% testing samples). For the random forest models, no significant difference was found. In the same models including only the energy variables, the improved performance of MI6 and RM was significantly better than for MID also in the test set (respectively,= 0.0001 and= 0.0002), and the gap in AUCs became substantial (AUC = 0.622 for MI6, 0.618 for RM and 0.507 for MID), indicating that socio-demographic and behavioural variables encapsulate information on energy intake by time of the day. Typical under-reporting bias present in nutritional epidemiology and the cross-sectional nature of the sample based on 3-day diaries may have affected these results, although use of diet diaries should minimize recall bias.In conclusion, the impact on health of timing and regularity of calorie intake in the day may act through other mechanisms than via overweight and may be captured by other demographic and behavioural variables; larger and prospective longitudinal studies are warranted to thoroughly investigate the added value of time-of-day information. Background/Objectives Methods Results: Conclusions: n p p p p2

Full Text

We can’t show the full text here under this license. Use the link below to read it at the source.

View on DOI ↗View on PubMed ↗

Application of Logistic Regression and Random Forests to Assess the Relevance of Chrononutrition Information for Prediction of Overweight in Adults: Evidence from the INRAN-SCAI 2005-2006 Italian Nutrition Survey.

Abstract

Full Text

You found one interesting study. We’ll send the next 7.

what lands in your inbox each week:

Recent issues from the circadian biology brief

Abstract

Full Text

Related papers

You found one interesting study. We’ll send the next 7.

what lands in your inbox each week:

Recent issues from the circadian biology brief