Individual-level dietary data were acquired from the National Health and Nutrition Examination Survey (NHANES) waves 2007–2008, 2009–2010, 2011–2012, and 2013–2014 from 35,507 individuals 2+ y.[27] NHANES is a cross-sectional, continuous survey that collects data on demography, diet, and health behaviors from approximately 5,000 individuals per year, and data are released on a two-year cycle. NHANES is maintained by the National Center for Health Statistics. Dietary data were acquired specifically from What We Eat In America (WWEIA), the dietary component of NHANES. Individuals complete a 24-hour recall administered by a trained interviewer using United States Department of Agriculture’s (USDA) Automated Multiple Pass Method,[28] and a subset of the study population completes a subsequent 24-hour recall by telephone on a non-consecutive day. Only data from day 1 were used because this represents per capita intake (i.e., mean intake of all individuals); whereas data from multiple days of intake are needed to estimate usual intake distributions, which was not the focus here[29]. WWEIA provides dietary data as reported consumed at home and away from home by individuals, which, in most cases, is in the form of mixed dishes composed of multiple foods, such as a cheeseburger. Data on nutrient content of each mixed dish are available, but WWEIA does not disaggregate most mixed dishes into their component foods or ingredients.

The Food and Nutrient Database for Dietary Studies (FNDDS)[30] is often used to disaggregate WWEIA dishes into their component foods, but these data are not provided at the resolution needed for this study. For example, FNDDS provides information on the amount of cheese, hamburger patty, and bun in a cheeseburger, but does not provide information on the amount of individual ingredients in a bun, such as wheat flour and oil. Disaggregation of WWEIA dishes into component ingredients was achieved with the Food Commodity Intake Database (FCID),[31] which was developed by the US Environmental Protection Agency (US EPA). FCID (2005–2010) provides data on the weight of nearly 500 ingredients included in each dish listed in WWEIA. FCID is the only source of data on the amount of individual ingredients in each WWEIA dish and, therefore, represents the most comprehensive and reliable data available on dish recipes.

The Healthy Eating Index-2015 (HEI-2015)[32] was used to assess diet quality for each individual in WWEIA (2007–2014). HEI-2015 provides a measure of compliance with the 2015–2020 Dietary Guidelines for Americans,[33] and includes 13 components, nine of which assess adequacy (total fruit, whole fruit, total vegetables, greens and beans, whole grains, dairy, total protein foods, seafood and plant proteins, and unsaturated:saturated fats) and four of which assess moderation (refined grains, sodium, added sugars, and saturated fats). Consumption amounts of each component were acquired from the Food Patterns Equivalents Database (FPED; 2007–2014), which provides reported consumption data from WWEIA converted to HEI components.[34] All consumption amounts are standardized to a 1,000 calorie basis (except for unsaturated:saturated fats), and each component has distinct scoring standards that range from 0–5 or 0–10, with higher scores being more favorable. Moderation components are reverse scored so that ultimately greater scores are favorable for each component. The component scores were summed to compute the overall HEI-2015 score with a maximum of 100. HEI-2015 scores were computed for each individual providing dietary data in WWEIA, and individuals were grouped by quintile of HEI-2015 score, where quintile 1 represents the lowest diet quality and quintile 5 represents the highest diet quality. Mean HEI-2015 scores for each quintile were appropriately computed using the National Cancer Institute’s population-ratio method, which has been demonstrated to generate scores that more closely reflect the true population mean than the arithmetic mean of individual-level scores.[35] The population-ratio method involves calculating the ratio of the reported intake of each food group for each individual to the reported energy intake, then computing the mean of these ratios across all individuals, then computing the HEI score using this summary ratio. [35]

Data on food waste were derived from the USDA Loss-adjusted Food Availability data series (LAFA).[36] LAFA is curated by USDA Economic Research Service (ERS) and provides estimates of food loss (including waste) at multiple stages in the food system for over 200 individual foods. Food loss represents the portion of food that is not consumed for any reason, including spoilage, cooking loss, and plate waste. At the consumer level, LAFA provides data on the proportion of purchased food that is inedible (such as banana peels), but does not provide information on the proportion of edible food wasted at the plate level, so these data were derived using the computations described in S1 Fig. We assumed that uneaten portions/cooking loss represents waste for the purpose of this study.

Cropland waste was estimated using the US Foodprint Model, a simulation model that represents the US as a closed food system.[37] The US Foodprint Model estimates cropland use associated with user-inputted data on food intake. Embedded computations use data on food losses and waste that occur at multiple points in the food system, the conversion of raw agricultural crops into edible food products, crop and grazing yields, livestock feed requirements, suitability of available land for agricultural uses, agricultural land area, and population size. Additional calculations account for multi-use crops (i.e., crops that are used to produce multiple products from equivalent mass) and multi-use cropland (cropland used to produce multiple crops during different parts of the year). Additional details are available elsewhere.[37] We modified the US Foodprint Model to produce estimates of the amount of cropland associated with food waste. Specifically, data on per capita food waste (described above) were added to the model instead of data on per capita food intake, and the embedded calculation adjusting for cooking loss and uneaten food was removed. Data on crop yields were updated to reflect the most recent five-year mean for each crop (2011–2015 for most crops),[38] and population data was updated to the 2015 US population size.[39]

National agricultural water irrigation rates, fertilizer application rates, and pesticide application rates (amount applied per land area per year) were acquired from USDA Agricultural Surveys (2002–2016)[38] and USDA Farm and Ranch Irrigation Surveys (2003–2013),[40] maintained by the National Agricultural Statistics Service (2002–2016). USDA Agricultural Surveys collect data primarily by telephone interviews with producers in all states for all major crops. Producers are selected based on the size of their operation, with larger producers having a greater likelihood of being selected to participate in the survey. Approximately 65,000–81,000 producers are surveyed for each crop annually.[41] USDA Farm and Ranch Irrigation Surveys are conducted every five years, and data are collected primarily by mailed surveys. Approximately 35,000 producers are surveyed for each data release.[40] Data for chemical use on hay and pasture are not available in USDA Agricultural Surveys, so these data were estimated based on personal communication with agricultural Extension Service agents in top producing states.

The proportion of each food item wasted at the consumer level (from LAFA) was linked with specific foods in FCID that best matched its description (Fig 1). Two investigators matched data independently, and consensus was used to resolve infrequent differences. Successful matches were achieved for 92% of the FCID foods; the remainder typically represented foods consumed infrequently and in minute amounts by the general population (such as passionfruit, arrowroot, and seaweed), and were not included in the analyses (S1 Table). Subsequently, linked FCID-LAFA data (which provided the proportion of each food wasted in each dish) were coupled with data on intake of mixed dishes (and their nutrient content) from WWEIA (which provided the amount of each dish and nutrient consumed) to disaggregate mixed dishes into component ingredients to estimate waste of individual ingredients, mixed dishes, and nutrients. For example, the portion size of a cheeseburger reported by a participant in WWEIA was disaggregated into its component ingredients (beef, cheese, wheat, etc.) using FCID. Each of those ingredients were then linked to a waste proportion from LAFA, which provided enough information to estimate the amount of each ingredient wasted for each individual in our WWEIA sample (S1 Fig).

The wasted amount (by weight) of individual ingredients were grouped according to the categorization scheme used in the US Foodprint model, which includes 22 distinct food categories. Mixed dishes were grouped primarily according to the FNDDS categorization scheme, which resulted in 36 distinct mixed dish categories. In cases where a mixed dish included disparate foods (e.g., meat and grains), the mixed dish was categorized in FNDDS according to the predominant food in the mixed dish (for example, steak with a side of rice was categorized as a mixed meat dish). The amount of nutrients associated with food waste was estimated for 57 individual nutrients.

For each individual in our WWEIA sample, HEI-2015 scores were computed, and the entire sample was divided into quintiles based on HEI-2015 scores. The mean waste of each ingredient was estimated for each HEI-2015 quintile, and these data were entered into the US Foodprint Model to estimate the annual amount of cropland associated with food waste, by HEI-2015 quintile, which was reported for ten distinct cropland uses. Estimates of cropland amounts associated with food waste were applied to application rates of irrigation water, fertilizers (nitrogen, phosphorus-P₂O₅, and potash-K₂O), and pesticides for major cropland uses to estimate waste of agricultural resources.

The relationship between daily per capita food waste amount (grams) and HEI-2015 quintiles was estimated using simple linear regression models to test for trend, with food waste as the dependent variable and quintile of diet quality as the independent variable. Statistical significance was set at P<0.05. All analyses were adjusted for the complex, multistage, probability sampling design of WWEIA using standardized procedures and variables provided by the National Center for Health Statistics.[42] Total annual amount of cropland associated with food waste was estimated using the US Foodprint model. A Monte Carlo simulation with 1,000 random, non-replacement draws was embedded into the US Foodprint model to provide measures of variation, based on inter-individual variation in reported food intake from WWEIA. The relationship between each of the sustainability measures (cropland use, and application of irrigation water, pesticides, and fertilizers) and HEI-2015 quintiles was estimated using simple linear regression models to test for trend, with a given sustainability measure as the dependent variable and quintile of diet quality as the independent variable. Statistical significance was set at P<0.05. SAS 9.4 (SAS Institute; Cary, NC) was used to estimate population-ratio HEI-2015 scores using the modified code and macros provided by the National Cancer Institute.[43] Stata14 (StataCorp; College Station, TX) was used for data management and all other analyses.

US consumers wasted 422 g (95% CI: 409–434 g)–nearly one pound–of food per person per day from 2007–2014 (Table 1 and S2 Table). Fruits and vegetables and mixed fruit and vegetable dishes accounted for 39% of food waste, followed by dairy (17%), meat and mixed meat dishes (14%), and grains and grain mixed dishes (12%). Remaining foods and dishes each accounted for less than 10% of total food waste: other foods and dishes (mostly candy, soft drinks, and other beverages), salty snacks, soup, potatoes and mixed potato dishes, nuts and seeds, Mexican dishes, eggs and mixed egg dishes, and table oils and salad dressing.

Nearly 26% (95% CI: 25–26%) of food was wasted by US consumers every day from 2007–2014 (S3 Table). Soup, fruits and vegetables and mixed dishes, and other foods and dishes had the highest waste rate (approximately 30% each). Nuts and seeds, potatoes and mixed potato dishes, and table oils and salad dressing had the lowest rates of food waste (12–18% each). Over 800 kcal (795–840 kcal) were wasted per person per day, representing about 29% of total daily energy intake (S4 Table). Of all nutrients, carotenoids had the greatest percent waste (31%) and vitamin D had the lowest percent waste (25%).

Annually, wasted food was grown on the equivalent of over 30 million acres (95% CI: 29.3–30.8 million acres) of cropland, representing 7.7% (7.5–7.9%) of all harvested cropland in the US (Fig 2A). Hay (8.9 million acres, 6.9–9.2 million acres) and feed grains and oilseeds (7.7 million acres, 7.6–7.9 million acres) accounted for over half (56%) of all cropland used to produce wasted food. Over 60% (58–64%) of land used to grow fruit was wasted, followed by vegetables (56%, 52–59%), and sweeteners (30%, 29–31%; Fig 2B). Cropland categories with the lowest proportion of waste were nuts (2.3%, 2.1–2.4%) and legumes (2.5%, 2.4–2.6%).

Using novel, interdisciplinary methods, we demonstrate that the average person living in America wastes nearly one pound of food daily. Furthermore, we find that higher diet quality is associated with greater food waste but less cropland waste, and greater waste of agricultural irrigation water and pesticides. This is largely due to the greater amount of fruits and vegetables included in higher quality diets, which have higher rates of waste, lower cropland needs, and higher application rates of agricultural inputs compared to other crops. Food waste is a critical component of environmental sustainability that, until now, has not been rigorously analyzed alongside diet quality. The current results suggest that simultaneous efforts to improve diet quality and reduce food waste may be critical. Practically, increasing consumers’ knowledge about how to prepare and store fruits and vegetables will be an essential component to reducing food waste. A number of important efforts have been proposed or are underway to reduce and repurpose food waste at the individual and institutional levels, yet further research is needed to better understand the comparative effectiveness of these efforts. Additional research is also needed to better understand how reducing food waste can contribute to monetary savings at the household level, especially for those with limited food budgets.