Dataset Information

Nutrient Estimation from 24-Hour Food Recalls Using Machine Learning and Database Mapping: A Case Study with Lactose.

ABSTRACT: The Automated Self-Administered 24-Hour Dietary Assessment Tool (ASA24) is a free dietary recall system that outputs fewer nutrients than the Nutrition Data System for Research (NDSR). NDSR uses the Nutrition Coordinating Center (NCC) Food and Nutrient Database, both of which require a license. Manual lookup of ASA24 foods into NDSR is time-consuming but currently the only way to acquire NCC-exclusive nutrients. Using lactose as an example, we evaluated machine learning and database matching methods to estimate this NCC-exclusive nutrient from ASA24 reports. ASA24-reported foods were manually looked up into NDSR to obtain lactose estimates and split into training (n = 378) and test (n = 189) datasets. Nine machine learning models were developed to predict lactose from the nutrients common between ASA24 and the NCC database. Database matching algorithms were developed to match NCC foods to an ASA24 food using only nutrients ("Nutrient-Only") or the nutrient and food descriptions ("Nutrient + Text"). For both methods, the lactose values were compared to the manual curation. Among machine learning models, the XGB-Regressor model performed best on held-out test data (R2 = 0.33). For the database matching method, Nutrient + Text matching yielded the best lactose estimates (R2 = 0.76), a vast improvement over the status quo of no estimate. These results suggest that computational methods can successfully estimate an NCC-exclusive nutrient for foods reported in ASA24.

SUBMITTER: Chin EL

PROVIDER: S-EPMC6950225 | biostudies-literature | 2019 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Nutrient Estimation from 24-Hour Food Recalls Using Machine Learning and Database Mapping: A Case Study with Lactose.

Chin Elizabeth L EL Simmons Gabriel G Bouzid Yasmine Y YY Kan Annie A Burnett Dustin J DJ Tagkopoulos Ilias I Lemay Danielle G DG

Nutrients 20191213 12

The Automated Self-Administered 24-Hour Dietary Assessment Tool (ASA24) is a free dietary recall system that outputs fewer nutrients than the Nutrition Data System for Research (NDSR). NDSR uses the Nutrition Coordinating Center (NCC) Food and Nutrient Database, both of which require a license. Manual lookup of ASA24 foods into NDSR is time-consuming but currently the only way to acquire NCC-exclusive nutrients. Using lactose as an example, we evaluated machine learning and database matching met ...[more]

PMID: 31847188

Similar Datasets

Project description:BackgroundStatistical methods to model the usual dietary intake of foods in a population generally ignore the additional information on the never-consumers. The objective of this study is to determine the added value of Food Frequency Questionnaire (FFQ) data allowing distinguishing the never-consumers from the non-consumers while modeling the usual intake distribution.MethodsThree food items with a different proportion of never-consumers were selected from the database of the Belgian food consumption survey of 2004 (N = 3200). The usual intake distribution for these food items was modeled with the Statistical Program for Analysis of Dietary Exposure (SPADE) and modeling parameters were extracted. These parameters were used to simulate (a) a new database with two 24-h recalls per respondent and (b) a "true" usual intake distribution. The usual intake distribution from the new database was obtained by modeling the 24-h recalls with SPADE, once without and once with the inclusion of the FFQ data on the never-consumers. Ratios were calculated for the different percentiles of the usual intake distribution: the modeled usual intake (g/day) (for both SPADE with and without the inclusion of FFQ data on never-consumers) was divided by the corresponding percentile of the simulated "true" usual intake (g/day). The closer the ratio is to one, the better the model fits the data.ResultsInclusion of the FFQ information to identify the never-consumers did not improve the estimation of the higher percentiles of the usual intake distribution. However, taking into account this FFQ information improved the estimation of the lower percentiles of the usual intake distribution even when the proportion of never-consumers was low.ConclusionsThe inclusion of FFQ information to identify the never-consumers is beneficial when interested in the whole usual intake distribution or in the lower percentiles only, no matter how low the proportion of never-consumers for that food item may be. However, when interest is only in the higher percentiles of the usual intake distribution, inclusion of FFQ information to identify the never-consumers will have no benefit.

Project description:BackgroundRegulating meal timing may have efficacy for improving metabolic health for preventing or managing chronic disease. However, the reliability of measuring meal timing with commonly used dietary assessment tools needs characterization prior to investigating meal timing and health outcomes in epidemiologic studies.ObjectivesTo evaluate the reliability of estimating meal timing parameters, including overnight fasting duration, the midpoint of overnight fasting time, the number of daily eating episodes, the period with the largest percentage of daily caloric intake, and late last eating episode (> 09:00 pm) from repeated 24-h dietary recalls (24HRs).MethodsIntraclass correlation coefficients (ICC), Light's Kappa estimates, and 95% CIs were calculated from repeated 24HR administered in 3 epidemiologic studies: The United States-based Interactive Diet and Activity Tracking in AARP (IDATA) study (n = 996, 6 24HR collected over 12-mo), German EPIC-Potsdam Validation Study (European Prospective Investigation into Cancer and Nutrition Potsdam Germany cohort) (n = 134, 12 24HR collected over 12-mo) and EPIC-Potsdam BMBF-II Study (Federal Ministry of Education and Research, "Bundesministerium für Bildung und Forschung") (n = 725, 4 24HR collected over 36 mo).ResultsMeasurement reliability of overnight fasting duration based on a single 24HR was "poor" in all studies [ICC range: 0.27; 95% CI: 0.23, 0.32 - 0.46; 95% CI: 0.43, 0.50]. Reliability was "moderate" with 3 24HR (ICC range: 0.53; 95% CI: 0.47, 0.58 in IDATA, 0.62; 95% CI: 0.52, 0.69 in the EPIC-Potsdam Validation Study, and 0.72; 95% CI: 0.70-0.75 in the EPIC-Potsdam BMBF-II Study). Results were similar for the midpoint of overnight fasting time and the number of eating episodes. Reliability of measuring late eating was "fair" in IDATA (Light's Kappa: 0.30; 95% CI: 0.21, 0.39) and "slight" in the EPIC-Potsdam Validation study and the EPIC-Potsdam BMBF-II study (Light's Kappa: 0.19; 95% CI: 0.15, 0.25 and 0.09; 95% CI: 0.06, 0.12, respectively). Reliability estimates differed by sex, BMI, weekday, and season of 24HR administration in some studies.ConclusionsOur results show that ≥ 3 24HR over a 1-3-y period are required for reliable estimates of meal timing variables.

Project description:BackgroundThe Automated Self-Administered 24-hour (ASA24) dietary recall system enhances the feasibility of collecting high-quality intake data in population-based studies.ObjectiveThe aim of this study was to assess the accuracy of portion size reporting in the ASA24 compared with interviewer-administered recalls.MethodsTrue intake for 3 meals was ascertained in 81 adults aged 20-70 y from the Washington, DC area. Participants were randomly assigned to complete an unannounced ASA24 or an interviewer-administered Automated Multiple-Pass Method (AMPM) recall the following day. An adapted Bland-Altman approach was used to assess agreement between true and reported portion sizes. Linear regression was used to assess log-scale differences between true and reported portion sizes by recall mode. The proportions of reported portion sizes within 10% and 25% of truth were estimated. Analyses were conducted for all foods and drinks and predetermined categories.ResultsMean differences between true and reported portion sizes were 3.7 g for the ASA24 and 11.8 g for the AMPM. According to the Bland-Altman-type plots, between 92% and 100% (depending on food or drink category and recall mode) of observations fell within the limits of agreement. After adjustment for multiple testing, the mean ratio of reported to true portion sizes was significantly >1 for the categories of all foods and drinks, all foods excluding liquids, amorphous or soft foods, and small pieces among AMPM respondents. Misestimation in the AMPM was significantly different from that in the ASA24 for all foods and drinks and for all foods excluding liquids. Small proportions of reported portions fell within 10% (16.2% for the ASA24 and 14.9% for the AMPM) and 25% (37.5% for the ASA24 and 33.2% for the AMPM) of truth.ConclusionsThe results raise the possibility that digital images tailored to different types and formats of foods may facilitate improved estimation of amounts eaten but highlight the need for continued work in this aspect of dietary assessment. This trial was registered at clinicaltrials.gov as NCT00978406.

Project description:Background: Our aim was to utilize a feasible quantitative methodology to estimate the dietary adequacy of >900 first-trimester pregnant women in poor rural areas of the Democratic Republic of the Congo, Guatemala, India and Pakistan. This paper outlines the dietary methods used. Methods: Local nutritionists were trained at the sites by the lead study nutritionist and received ongoing mentoring throughout the study. Training topics focused on the standardized conduct of repeat multiple-pass 24-hr dietary recalls, including interview techniques, estimation of portion sizes, and construction of a unique site-specific food composition database (FCDB). Each FCDB was based on 13 food groups and included values for moisture, energy, 20 nutrients (i.e. macro- and micronutrients), and phytate (an anti-nutrient). Nutrient values for individual foods or beverages were taken from recently developed FAO-supported regional food composition tables or the USDA national nutrient database. Appropriate adjustments for differences in moisture and application of nutrient retention and yield factors after cooking were applied, as needed. Generic recipes for mixed dishes consumed by the study population were compiled at each site, followed by calculation of a median recipe per 100 g. Each recipe's nutrient values were included in the FCDB. Final site FCDB checks were planned according to FAO/INFOODS guidelines. Discussion: This dietary strategy provides the opportunity to assess estimated mean group usual energy and nutrient intakes and estimated prevalence of the population 'at risk' of inadequate intakes in first-trimester pregnant women living in four low- and middle-income countries. While challenges and limitations exist, this methodology demonstrates the practical application of a quantitative dietary strategy for a large international multi-site nutrition trial, providing within- and between-site comparisons. Moreover, it provides an excellent opportunity for local capacity building and each site FCDB can be easily modified for additional research activities conducted in other populations living in the same area.

Dataset Information

Nutrient Estimation from 24-Hour Food Recalls Using Machine Learning and Database Mapping: A Case Study with Lactose.

Publications

Nutrient Estimation from 24-Hour Food Recalls Using Machine Learning and Database Mapping: A Case Study with Lactose.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets