Advertisement

Adherence to healthy food choices during the COVID-19 pandemic in a U.S. population attempting to lose weight

      Highlights

      • Growing evidence suggests that food preferences and consumption behaviors are modified in populations during stressful events.
      • Self-reported servings of fresh fruit and vegetable intake decreased from pre-to during-COVID, while intake of red meat and starchy vegetables increased.
      • More men than women increased their intake of red meat and processed meat.
      • Less overall change in fruit and vegetable consumption was seen in those 66 and older, compared to aged 18–35.
      • More subjects aged 18–35 years reduced their intake of caffeine, desserts, lean meat and salads compared to older participants.

      Abstract

      Background and aims

      Food preferences are often modified in populations during stressful, unanticipated events. We examined how a U.S. population's food choices changed during the beginning of the COVID-19 stay-at-home orders, specifically during the spring of 2020.

      Methods and results

      Daily dietary intake data from a digital behavior change weight loss program, which includes an interface for logging meals, beverages, and snacks, were analyzed to assess self-reported food choices from March 5-March 11, 2020 ("Start-COVID") and during the first week of the COVID-19 lockdown (March 12-March 18, 2020; "during-COVID"). The final sample consisted of 381,564 participants: 318,076 (83.4%) females, the majority who were aged 45–65 years (45.2%). Results indicate that self-reported servings of fresh fruit and vegetable intake decreased from start-to during-COVID, while intake of red meat and starchy vegetables increased. More men than women increased their intake of red meat and processed meat. Less overall change in fruit and vegetable consumption was seen in those 66 and older, compared to aged 18–35. Lean meat and starchy vegetable intake increased in older participants, but the change was negligible in younger subjects. More subjects aged 18–35 years reduced their intake of caffeine, desserts, lean meat, and salads compared to older participants. No changes were observed in snack or alcohol intake logged.

      Conclusion

      This study revealed that particular food groups were altered according to age and gender during the first weeks of COVID lockdown. Understanding changes in food choices during a crisis may be useful for preparing supply chains and public health responses.

      Keywords

      Introduction

      In the spring of 2020, the COVID-19 pandemic caused widespread changes in the lifestyle of populations all across the globe. Lockdown and quarantine orders decreased mobility while increasing anxiety [
      • Flanagan E.W.
      • Beyl R.A.
      • Fearnbach S.N.
      • Altazan A.D.
      • Martin C.K.
      • Redman L.M.
      The impact of COVID-19 stay-at-home orders on health behaviors in adults.
      ]. Due to this almost unprecedented event, American news agencies reported stocking up of canned foods and decreased purchase of fresh foods [
      • Creswell J.I.
      ‘I just need the comfort’: processed foods make a pandemic comeback.
      ]. Yet, little is known on how U.S. consumers changed their consumption habits as lockdown policies took effect and how these choices may reflect consumers' emotional status. While countries issued lockdowns and travel bans due to the COVID-19 pandemic, European surveys reported that people increased snacking [
      • Pietrobelli A.
      • Pecoraro L.
      • Ferruzzi A.
      • Heo M.
      • Faith M.
      • Zoller T.
      • et al.
      Effects of COVID-19 lockdown on lifestyle behaviors in children with obesity living in Verona, Italy: a longitudinal study.
      ]. However, few studies have reported what kinds of foods are chosen in an American population directly before and after a crisis.
      Many factors drive food choice; however, studies have shown that emotion plays a pivotal role in times of crisis. For example, a recent study reported that high levels of stress correlated with emotional eating, and stress-related food choice was associated with mood and convenience in U.S. survey respondents during June of 2020 [
      • Shen W.
      • Long L.M.
      • Shih C.-H.
      • Ludy M.-J.
      A humanities-based explanation for the effects of emotional eating and perceived stress on food choice motives during the COVID-19 pandemic.
      ]. Comfort foods have been described as foods that induce a psychological change in state and promote comfort and pleasure for a person when consumed [
      • van den Bos R.
      • de Ridder D.
      Evolved to satisfy our immediate needs: self-control and the rewarding properties of food.
      ]. Comfort foods can be defined as foods specifically chosen for soothing properties, either physiologically, in the case of calorically dense snacks, rich desserts, chocolate, or ice cream, or psychologically, such as warm meals with meats or hearty soup that evoke childhood memories. In a review discussing determinants of food choice Leng and colleagues concluded intake is guided by diverse emotional states, such as stress, as demonstrated by studies performed in controlled laboratories, but that real world observations of stress-related food choice were lacking [
      • Leng G.
      • Adan R.
      • Belot M.
      • Brunstrom J.
      • De Graaf K.
      • Dickson S.
      • et al.
      The determinants of food choice.
      ]. It was of interest to explore how food choices changed in a large population in response to the sudden onset of a real-world stressor like the COVID-19 pandemic. Thus the purpose of this study is to examine how a population attempting to lose weight via an on-line weight loss program altered food choices and adherence to foods low in calorie density during the COVID-19 stay-at-home orders in the United States. It was expected that differences in food choices would be seen across age groups and gender and would tend towards high-calorie density, e.g., 'comfort' connotation vs. healthful (low-calorie density).

      Methods

      Study population

      Noom users were initially self-referred and signed up for the program through the app store (iTunes/Google Play). Age, gender, height, and weight are requested upon signup, but no information on ethnicity or income status is collected. Informed consent to participate in research is included during the signup phase of the program, in which users can choose to opt-out. This study, and informed consent, were approved by an external IRB (Advarra, Pro00017565).
      To be included in the study, users enrolled were (a) 18 years and older, (b) living in the U.S., and logged food at least twice during the weeks of March 5–11 and March 12–18. Users were excluded from the study if: (a) total usage of the program was less than three weeks, (b) users did not enter food log data during both weeks of March 5–11, and March 12–18, and (c) users did not use the English language version of the program.

      Data collection

      This is an observational, retrospective, cohort study. Data was collected via the behavioral change weight loss program, Noom. Noom is a mobile intervention that allows users to log their weight, meals, and physical activity via a smartphone interface and gives access to a virtual 1:1 behavior change coach, support group, and daily curriculum that includes diet and exercise-, and psychology-based content [
      • Jacobs S.
      • Radnitz C.
      • Hildebrandt T.
      Adherence as a predictor of weight loss in a commonly used smartphone application.
      ].
      Noom users’ self-reported food choices during the pandemic were analysed via an existing food logging database that had been previously developed by Noom. Data were collected from the Start-COVID time period (March 5-March 11) and during the first week of the COVID-19 lockdown (March 12-March 18). March 12 was designated as the first day after COVID-19 stay-at-home orders based on news articles advising people in several states to avoid non-essential shopping and to remain at home [
      • Mervosh S.
      • Lu D.
      • Swales V.
      See which states and cities have told residents to stay at home.
      ]. A relatively short time period was chosen to evaluate how abrupt news dissemination on a potential health crisis immediately affected food choices, and to avoid potential interactions from seasonal food availability.

      Analysis of food groups

      Analysis of food groups was extracted from food categorizations in the Noom food database, which contains thousands of food products available in the United States, and allows users to designate their preferred serving measurement unit: metric (g) or Imperial (oz), including units such as tablespoons, cups, bowls, plates, and typical serving units for specific foods (e.g., slices for apples or oranges). Products are categorized into food types reflecting the main component of the food or preparation of the food, which allows further compiling of food types into food groups (e.g., chips servings are compiled into a 'snacks' category, along with popcorn, nuts, pretzels, cheese puffs, and other salty snack foods). The food group's complex meals/sandwiches include foods containing several food groups such as meat, grains, dairy, or vegetables that are prepared as a main meal. 'Sweetened drinks' include sweet, carbonated drinks, as well as juice and juice drinks. Since the Noom program encourages healthy eating, foods containing whole grains, such as whole wheat flour, are given a different designation than foods made from white flour, such as regular pasta or white bread. Thus 'whole grains' designates cereals primarily made with whole-grain ingredients, such as brown rice, corn, or whole wheat. 'High-fat condiments' include butter, oils, high-fat dressings, sauces, and cream. 'Red meat' includes fatty cuts of meat or preparations such as hamburgers, steaks, or lamb or pork chops. 'Lean protein' includes lean cuts of pork and beef, chicken, seafood, soy-based foods, and protein shakes. 'Processed meats and sausage' includes fatty sausage, bacon, and other highly processed meats.

      Adherence to low calorie density food choices

      In order to assess overall changes to the food choices related to weight loss goals at the start of the lockdown and during the lockdown, we analysed consumption of foods with low calorie density (fruits and vegetables) vs high calorie density (desserts, candy, processed meat). The Noom weight loss program encourages participants to choose healthy, low calorie density food via a food categorization system based on Volumetrics, a system developed by Barbara Rolls and colleagues [
      • Rolls B.J.
      • Drewnowski A.
      • Ledikwe J.
      Changing the energy density of the diet as a strategy for weight management.
      ]. Rolls demonstrated that consumption of low calorie density, high volume foods such as fruits and vegetables aided weight loss by improving satiation and the sensation of fullness [
      • Rolls B.J.
      • Drewnowski A.
      • Ledikwe J.
      Changing the energy density of the diet as a strategy for weight management.
      ]. Noom uses a food color categorization system based on calorie density to encourage participants to choose low calorie density foods (termed ‘green’ foods) vs moderate calorie density foods (termed ‘yellow’ foods) vs high calorie density foods (termed ‘red’ foods, typically foods high in sugar and fat) [
      • Michaelides A.
      • Major J.
      • Pienkosz Jr., E.
      • Wood M.
      • Kim Y.
      • Toro-Ramos T.
      Usefulness of a novel mobile diabetes prevention program delivery platform with human coaching: 65-week observational follow-up.
      ]. We analyzed the ratio of foods consumed from the green food vs total food intake and the ratio of red food intake compared to total food intake according to gender and age range during the Start-COVID time period (March 5-March 11) and during the first week of the COVID-19 lockdown (March 12-March 18). Foods are categorized as 'green' when they contain less than one calorie/g or less than 0.4 calories/mL. Foods are categorized as 'red' when they contain more than 2.4 calories/g or more than one calorie/ml. Foods are categorized as 'yellow' when they contain between 1 and 2.4 calories/g or between 0.4 and 1 calorie/ml.

      Statistical analysis

      Descriptive statistics were calculated for users’ baseline characteristics and expressed in means, medians, and standard deviations for continuous variables and frequencies and percentages for categorical variables (Table 1). Frequencies and percentage of changes for food consumption data were compared before and after March 12 (Table 2, Table 3, Table 4). Differences in start- and during-COVID diet ratios within gender and age groups were assessed by two-tailed paired t-tests. Differences in during-COVID diet ratios between gender and age groups adjusted for start-COVID ratios were analyzed by one-way analysis of covariance (ANCOVA) with Tukey post hoc correction. All statistical tests were 2-sided, with an α of 0.05, and carried out in R (version 3.6.0).
      Table 1Demographic and weight information of participants who logged food during March 5–11 and March 12–18.
      Characteristicsn

      (N = 381,564) (%)
      MeanSDMedian
      Gender
      Male63,468 (16.6%)
      Female318,076 (83.4%)
      Age (years)
       All381,564 (100%)47.713.548.0
      18–3581,845 (21.4%)
      36–4586,743 (22.7%)
      45–65172,299 (45.2%)
      66 +40,677 (10.7%)
      Height (inches)333,237 (87.3%)65.73.565.0
      First weight on record (kg)372,158 (97.5%)89.721.286.1
      Most recent weight (kg)372,158 (97.5%)85.520.482.1
      Table 2Number of servings according to the food group for March 5–11 (Start-COVID) and March 12–18 (During COVID); change in the number of servings per week and % change in food group servings was also extracted.
      COVID Era food groupsStart-COVID

      # Servings logged
      During COVID

      # Servings logged
      Change in

      # servings
      % change in # servings
      Red meat356,529374,97218,4435.2%
      Starchy vegetables176,334183,07467403.8%
      White bread, rice pasta, cereal721,605711,389−10,216−1.4%
      Sweeteners218,630209,209−9421−4.3%
      Alcohol166,455158,962−7493−4.5%
      Soups98,85093,111−5739−5.8%
      Whole grain bread, pasta, rice, cereal839,470784,493−54,977−6.5%
      Salty snacks44,540,10041,821,400−271,8700−6.1%
      High fat condiments648,211603,103−45,108−7.0%
      Dairy1,126,0861,046,808−79,278−7.0%
      Dessert427,435395,728−31,707−7.4%
      Complex meals/sandwiches601,213535,513−65,700−10.9%
      Vegetables1,429,6501,264,268−165,382−11.6%
      Fruits1,390,6581,229,266−161,392−11.6%
      Sweetened drinks214,359188,084−26,275−12.3%
      Lean protein623,604539,380−84,224−13.5%
      Salads478,529387,682−90,847−19.0%
      Tea/coffee beverages33,15724,531−8626−26.0%
      All food groups15,443,28413,780,004−1,663,280−9.7%
      Table 3Percent of users reporting consumption of specific food groups according to gender (Start: Start-COVID, March 5–11; During: During COVID, March 12–18).
      FemaleMale
      % Users Start% Users During% Change% Users Start% Users During% ChangeFemale vs. male
      Fruits79.1%74.1%−5.0%75.1%70.9%−4.2%−3.2%
      Salad35.6%29.1%−6.5%34.6%28.7%−5.9%1.3%
      Vegetables73.3%67.3%−6.3%68.8%64.0%−4.7%−3.3%
      Lean protein32.3%28.5%−3.9%33.4%29.7%−3.6%1.3%
      Whole grain bread/pasta/rice/cereal29.1%27.7%−1.4%28.6%27.4%−1.2%−0.3%
      Red_meat26.9%29.9%2.9%33.9%36.5%2.5%6.6%
      Processed meat/sausage37.1%36.2%−0.9%44.9%43.1%−1.8%7.0%
      Starchy vegetables34.6%35.0%0.3%33.5%34.8%1.2%−0.2%
      High fat condiments28.7%26.1%−2.6%24.5%22.6%−2.0%−3.6%
      White bread/pasta/rice/cereal29.0%27.8%−1.1%28.7%27.9%−0.8%0.0%
      Salty snacks15.1%13.7%−1.4%13.7%12.8%−0.9%−0.9%
      Desserts25.6%23.2%−2.4%21.7%20.1%−1.6%−3.1%
      Dairy19.3%18.0%−1.4%16.3%15.5%−0.9%−2.5%
      Sweetened drinks10.0%8.7%−1.3%12.4%10.6%−1.9%1.9%
      Sweeteners11.7%11.4%−0.4%9.6%9.5%−0.1%−1.8%
      Alcohol13.0%11.8%−1.2%16.1%14.6%−1.4%2.8%
      Soups11.4%10.5%−0.9%13.2%12.2%−1.0%1.7%
      Complex meals/sandwiches13.6%11.7%−1.9%17.4%15.2%−2.2%3.5%
      Tea/coffee beverages6.4%4.8%−1.6%3.7%3.0%−0.8%−1.8%
      Eggs30.3%28.7%−1.6%31.2%29.9%−1.2%1.2%
      Table 4Percent of users reporting consumption of specific food groups according to age (Start: Start COVID, March 5–11; During: During COVID, March 12–18).
      18–3536–4546–6566 and above
      % Users

      Start
      % Users

      During
      % change% Users

      Start
      % Users

      During
      % change% Users

      Start
      % Users

      During
      % change% Users

      Start
      % Users

      During
      % change
      Fruits73.1%67.7%−5.4%76.8%70.6%−6.2%80.4%75.8%−4.6%83.2%81.3%−1.9%
      Vegetables69.3%62.1%−7.3%72.8%64.9%−7.9%73.7%68.1%−5.6%75.6%74.2%−1.4%
      White bread/pasta/rice/cereal28.5%27.2%−1.3%29.6%28.2%−1.4%28.6%27.6%−1.0%29.7%29.3%−0.4%
      Dessert25.0%22.1%−2.9%25.9%22.8%−3.0%24.3%22.4%−1.8%25.4%24.4%−1.0%
      Salad29.8%23.9%−6.0%34.2%26.7%−7.4%37.8%31.3%−6.4%39.5%34.9%−4.6%
      High fat condiments26.0%23.4%−2.5%28.6%25.1%−3.5%29.9%27.3%−2.6%33.0%31.7%−1.3%
      Starchy vegetables28.7%28.0%−0.7%31.8%31.1%−0.7%36.4%37.3%0.9%43.0%46.7%3.7%
      Processed meat/sausage33.4%32.2%−1.2%37.9%36.4%−1.5%39.9%39.1%−0.8%43.9%43.3%−0.6%
      Red meat22.9%24.5%1.6%27.0%28.9%1.9%30.2%33.6%3.4%32.6%37.8%5.3%
      Lean meat24.2%21.0%−3.2%26.7%22.8%−3.8%27.9%24.7%−3.2%28.9%27.1%−1.9%
      Complex meals/sandwiches14.7%12.8%−1.8%15.1%13.0%−2.1%14.0%12.5%−1.5%12.7%11.6%−1.1%
      Dairy16.0%14.7%−1.2%17.3%15.8%−1.5%17.5%16.4%−1.1%18.8%18.3%−0.5%
      Alcohol11.8%10.7%−1.1%13.7%12.5%−1.3%14.0%12.7%−1.3%14.5%13.6%−0.9%
      Sweeteners10.7%10.3%−0.4%11.6%11.0%−0.6%11.2%11.0%−0.2%12.6%12.7%0.2%
      Soups8.9%8.0%−1.0%10.7%9.3%−1.4%12.5%11.7%−0.8%16.0%15.6%−0.4%
      Eggs26.5%25.6%−0.9%29.9%28.2%−1.8%31.7%30.0%−1.7%34.0%32.7%−1.2%
      Snacks12.7%11.3%−1.4%14.3%12.5%−1.8%14.8%13.4%−1.4%15.0%14.4%−0.6%
      Sweetened drinks10.9%9.1%−1.8%10.3%8.5%−1.7%10.0%8.8%−1.2%11.7%10.7%−1.0%
      Whole grain bread/rice/pasta/cereal20.8%19.8%−1.0%21.9%20.5%−1.4%23.6%22.6%−1.1%26.5%26.3%−0.2%
      Tea/coffee beverages8.7%6.5%−2.3%6.9%5.1%−1.9%4.7%3.6%−1.1%3.1%2.4%−0.8%

      Results

      Of the 381,564 participants who met inclusion criteria, 318,076 (83.4%) were females, and the mean age of the sample was 47.76 (SD = 13.59) (Table 1). The majority of users included in our population were aged 45–65 years old (45.2%). The average self-reported first weight on record for participants is 89.72 kg (SD = 21.2), and the most recent average weight is 85.57 kg (SD = 20.4).

      Changes in consumption of specific food categories

      To assess changes in dietary patterns, data from Noom users who logged meals during the week of March 5-March 11 (Start-COVID) and during the week of March 12–18 (During-COVID) a descriptive analysis was performed according to food grouping. Users' overall logging of food group servings decreased by 9% in terms of actual items of food logged (Table 2). Results indicate that users increased intake of red meat (5%) and starchy vegetables (potatoes, corn, peas, and winter squash, 4%). Lean meat, salads, and caffeinated drinks show large decreases in logging (i.e., those food groups exhibiting a larger overall drop than the general decrease in food group logging). Moderate decreases in vegetables, fruit, and sweetened drinks were also observed during this time period.
      When food group data were analyzed according to self-identified gender, several differences in overall consumption patterns between males and females were observed. The largest differences in self-reported intake were seen in the greater percentage of men versus women reporting meat food groups: red meat (6.6% difference) and processed meats/sausage (7% difference), and also in the complex meals/sandwiches group (3.5% difference) (Table 3). A higher percentage of female users reported intake of fruits (3.2%) and dessert (3.1%) than male users. Both men and women had an increase in the percentage of users reporting red meat (men: 2.5%; women: 2.9%) during the first week of COVID stay-at-home mandates. Overall, more food groups decreased in terms of the percentage of users reporting consumption: Notably fruits (men: −4.2%; women: −5.0%), salad (men:−5.9%; women: −6.5%) vegetables (men: −4.7%; women: −6.3%), and lean protein (men: −3.6%; women: −3.9%) dropped in terms of users reporting intake.
      When data was analyzed among age groups, other patterns emerged on how older versus younger users changed their diet during the initiation of COVID stay-at-home policies (Table 4). The self-reported intake of food groups changed much less in users aged 66 and older than users aged 18–35 years from the week of March 5 (Start-COVID) to the week of March 12 (During-COVID). There were decreases in the percentage of younger users reporting consumption of fruits and vegetables (−5.4% and −7.3%, respectively) during the week of March 12 (During-COVID), yet negligible change in the percentage of older users' consumption of fruits and vegetables. Conversely, the increase in users aged 66 and over who consumed red meat was substantially higher (5.3%) than younger age groups. A similar pattern was also seen in the percentage of older users consuming starchy vegetables (+3.7% for users 66 and older vs. −0.7% for users aged 18–35 years). Meanwhile, the percentage of younger users consuming lean meat rose and slightly decreased in users 66 and older.
      All groups showed a drop in salad intake (−6%, −7.4%, −6.4%, and −4.6% for age groups 18–35, 36–45, 46–65, and 66 and over, respectively). While the percentage of users consuming alcohol did not change according to age, caffeinated drinks such as tea and coffee decreased in the proportion of users aged 18–35 years (−2.3%) but only marginally decreased in users aged 35 and older.

      Changes in consumption of foods according to calorie density (adherence to weight loss guidelines)

      In order to assess how well participants were adhering to weight loss guidance via selection of foods with low calorie density (fruits, vegetables) vs. high calorie density (desserts, processed meats) we analysed consumption of foods categorized as ‘green’ (calories/g < 1), ‘yellow’ (calories/g ≥ 1 and < 2.4) or ‘red’ (calories/g ≥ 2.4). The ratio of green (low calorie density) food consumption versus consumption of higher calorie density food categories yellow and red decreased after the start of the lockdown for both men and women (Table 5). In contrast, the ratio of red (e.g., high calorie density) food consumption versus yellow and green foods increased after the start of lockdown. From ANCOVA analysis, we observed that gender was significantly associated to during-COVID red food ratio intake where men showed higher red food intake, adjusting for Start-COVID red food ratio intake (Supplementary Tables 1 and 2). Male users' adjusted average during-COVID red food ratio is 0.01 higher than females' ratios. When the ratio of green vs. yellow vs. red food consumption were examined according to age range, we observed that those aged 18–35 years, 36–45 years, or 46–65 years significantly decreased their green food and yellow consumption and increased their red food consumption from the start of COVID lockdown to during COVID lockdown (Table 5). However, those older than 66 years did not change their green food ratio and appeared to have less change in their yellow and red food consumption than younger groups. Via ANCOVA analysis, we confirmed that the 66+ group had less change in their food color ratio consumption than the 18–35 years group (Supplementary Tables 3 and 4) (see Table 6).
      Table 5Change in adherence to weight loss-promoting food choices according to gender at the start of the lockdown period (Start-COVID) and during the COVID lockdown (During-COVID). The ratio of foods with low calorie density (green food ratio), moderate calorie density (yellow food ratio) and high calorie density (red food ratio).
      Male (n = 54,289)Female (n = 243,834)
      Start- COVIDDuring COVIDMean difference (95%CI)P-valueStart- COVIDDuring- COVIDMean difference (95%CI)P-value
      Green food ratio, Mean (SD)0.239 (0.16)0.237 (0.18)0.002 (0.001, 0.004)t = 2.922, df = 54,288, p-value = 0.003480.255 (0.16)0.251 (0.18)0.004 (0.003, 0.005)t = 11.352, df = 243,833, p-value < 2.2e–16
      Yellow ratio, Mean (SD)0.398 (0.17)0.394 (0.19)0.003 (0.001, 0.005)t = 3.3232, df = 54,288, p-value = 0.00089060.401 (0.17)0.400 (0.19)0.001 (0.000, 0.002)t = 3.0146, df = 243,833, p-value = 0.002573
      Red ratio, Mean (SD)0.364 (0.19)0.369 (0.21)−0.005 (−0.007, −0.003)t = −5.4924, df = 54,288, p-value = 3.983e–080.346 (0.18)0.351 (0.2)−0.005 (−0.006, −0.004)t = −12.449, df = 243,833, p-value < 2.2e–16
      Table 6Change in adherence to weight loss-promoting food choices according to age range at the start of the lockdown period (Start-COVID) and during the COVID lockdown (During-COVID). The ratio of foods with low calorie density (green food ratio), moderate calorie density (yellow food ratio), and high calorie density (red food ratio).
      18–35 (n = 61,839)36–45 (n = 66,968)46–65 (n = 136,782)66 and above (n = 32,396)
      Start- COVIDDuring COVIDMean difference (95%CI)P- valueStart- COVIDDuring COVIDMean difference (95%CI)P- valueStart- COVIDDuring-COVIDMean difference (95%CI)P- valueStart-COVIDDuring COVIDMean difference (95%CI)P- value
      Green ratio food, Mean (SD)0.238 (0.17)0.236 (0.19)0.002 (0.001, 0.004)t = 2.6992, df = 618, p-value = 0.0060.245 (0.16)0.240 (0.18)0.004 (0.003, 0.006)t = 6.4028, df = 669, p-value = 1.5e–100.258 (0.16)0.253 (0.18)0.005 (0.004,0.006)t = 10.708, df = 136,781, p-value < 2.2e–160.2654 (0.16)0.2651 (0.17)0.0002 (−0.0014, 0.0019)t = 0.28267, df = 323, p-value = 0.777
      Yellow food ratio, Mean (SD)0.402 (0.18)0.401 (0.2)0.001 (−0.001, 0.003)t = 0.75711, df = 618, p-value = 0.4490.401 (0.17)0.400 (0.19)0.001 (−0.001,0.002)t = 0.76923, df = 669, p-value = 0.440.400 (0.17)0.397 (0.18)0.002 (0.001, 0.003)t = 4.1639, df = 136,781, p-value = 3.11e–050.401 (0.16)0.399 (0.17)0.002 (0.000, 0.004)t = 2.1953, df = 323, p-value = 0.028
      Red food ratio, Mean (SD)0.361 (0.2)0.364 (0.22)−0.003 (−0.005,−0.001)t = −2.8999, df = 618, p-value = 0.0030.356 (0.19)0.361 (0.21)−0.005 (−0.007, −0.003)t = −6.046, df = 669, p-value = 1.4e–090.343 (0.18)0.351 (0.2)−0.007 (−0.008, −0.006)t = −13.062, df = 136,781, p-value < 2.2e–160.335 (0.17)0.338 (0.18)−0.002 (−0.004, −0.000)t = −2.3695, df = 323, p-value = 0.017

      Discussion

      Our study sought to explore Noom users’ food choices during the COVID-19 stay-at-home orders imposed throughout the United States. Results showed that fruit and vegetable consumption, especially salads, appeared to decrease during the initial week of US COVID stay-at-home mandates, while red meat, processed meat, and starchy vegetable consumption increased. Overall dietary pattern changes were confirmed via analysis of low calorie density foods versus high calorie foods according to gender and age. In all age groups, consumption of red foods, e.g., high calorie density foods (fatty meat, desserts), significantly increased during lockdown. A higher percentage of men and older participants (66 years and older) reported red meat and starchy vegetable intake during March 12–18 (During-COVID) vs. March 5–11 (Start-COVID). However, older participants tended to maintain consumption of healthful, low calorie density foods while younger groups decreased their intake.
      Few studies have examined food choices after a stressful, widespread event. Prior to 2020, three studies reported food choice changes after a natural disaster. Women living in Christchurch, New Zealand, engaged in emotional eating (more snacking and less fruit and vegetable consumption) after a 7.1 Richter scale earthquake [
      • Kuijer R.G.
      • Boyce J.A.
      Emotional eating and its effect on eating behaviour after a natural disaster.
      ]. Another study assessed food group intake in evacuees after the 2011 Fukushima Earthquake in eastern Japan and reported that diet quality worsened in those reporting higher stress, i.e., less intake of fruits and vegetables, meat, soy, and dairy [
      • Uemura M.
      • Ohira T.
      • Yasumura S.
      • Otsuru A.
      • Maeda M.
      • Harigane M.
      • et al.
      Association between psychological distress and dietary intake among evacuees after the Great East Japan Earthquake in a cross-sectional study: the Fukushima Health Management Survey.
      ]. A follow-up study in the same population also demonstrated that living in non-home environments also contributed to poor diet quality [
      • Zhang W.
      • Ohira T.
      • Abe M.
      • Kamiya K.
      • Yamashita S.
      • Yasumura S.
      • et al.
      Evacuation after the great east Japan earthquake was associated with poor dietary intake: the Fukushima health management survey.
      ]. Since the spring of 2020, several studies of food choices after COVID-19 lockdown have been reported, using methods such as surveys or purchase data. For example, a survey of Chinese consumers showed decreases in fresh fruit and vegetables and increases in starchy, staple foods [
      • Jia P.
      • Liu L.
      • Xie X.
      • Yuan C.
      • Chen H.
      • Guo B.
      • et al.
      Impacts of COVID-19 lockdown on diet patterns among youths in China: the COVID-19 Impact on Lifestyle Change Survey (COINLICS).
      ], while a Polish survey noted decreased consumption of fruits and vegetables and increased intake of meat and dairy [
      • Sidor A.
      • Rzymski P.
      Dietary choices and habits during COVID-19 lockdown: experience from Poland.
      ]. Other studies have employed purchasing data, such as an Italian report where higher sales of shelf-stable food were observed during early springtime [
      • Bracale R.
      • Vaccaro C.M.
      Changes in food choice following restrictive measures due to Covid-19.
      ]. Food selection after a stressful event may reflect cultural values or what was easily available in a particular region, but predominantly trends were towards shelf-stable and high energy choices. However, many of these reports are derived from data collected weeks after lockdown initiation or are purchase data and do not necessarily reflect immediate intake.
      Using food logging data from 381,564 Americans participating in a weight loss program, the main food categories which exhibited increases after COVID lockdown guidance were red meat and starchy vegetables such as potatoes, both of which have connotations of familiarity and simplicity, e.g., the saying 'meat and potatoes' harkens to essential nourishment. The current set of data agrees with previous reports of decreased fruit and vegetable selection in favor of more shelf-stable foods and red meat from both scientific and lay public literature [
      • Creswell J.I.
      ‘I just need the comfort’: processed foods make a pandemic comeback.
      ,
      • Jia P.
      • Liu L.
      • Xie X.
      • Yuan C.
      • Chen H.
      • Guo B.
      • et al.
      Impacts of COVID-19 lockdown on diet patterns among youths in China: the COVID-19 Impact on Lifestyle Change Survey (COINLICS).
      ,
      • Sidor A.
      • Rzymski P.
      Dietary choices and habits during COVID-19 lockdown: experience from Poland.
      ,
      • Bracale R.
      • Vaccaro C.M.
      Changes in food choice following restrictive measures due to Covid-19.
      ]. Thus, it appears that food choices during this period may be reflecting desires to stock up and consume core, high calorie density food groups. Indeed, similar results have been reported in a study examining children's eating behavior during the initial COVID lockdown [
      • Shen W.
      • Long L.M.
      • Shih C.-H.
      • Ludy M.-J.
      A humanities-based explanation for the effects of emotional eating and perceived stress on food choice motives during the COVID-19 pandemic.
      ]. Meanwhile, the percentage of users consuming alcohol, regardless of age or gender, did not differ during the two consecutive weeks, while caffeine consumption appeared to decrease in the youngest age range (18–35 years), which may be due to less opportunity or desire to drink caffeinated beverages at home than in a workplace.
      There are several limitations to this study. One limitation is that all data are self-reported meal logging, which is susceptible to flawed recall, omissions, or lapses in logging. While some studies have indicated a reasonable correlation of digital food logging data to more traditional methods, results can vary according to the platform used [
      • Rangan A.M.
      • Tieleman L.
      • Louie J.C.
      • Tang L.M.
      • Hebden L.
      • Roy R.
      • et al.
      Electronic Dietary Intake Assessment (e-DIA): relative validity of a mobile phone application to measure intake of food groups.
      ,
      • Boushey C.J.
      • Spoden M.
      • Delp E.J.
      • Zhu F.
      • Bosch M.
      • Ahmad Z.
      • et al.
      Reported energy intake accuracy compared to doubly labeled water and usability of the mobile food record among community dwelling adults.
      ]. Additionally, the participant pool is skewed more towards women (83.4%) than men (16.6%). Lastly, changes in COVID-related eating behavior depend on the extent of lockdown in different regions of the U.S. and may not have occurred at the same time periods. However, the large number of participants included in the study (n = 381,564) adds robustness to the findings.
      In conclusion, this observational study explored how food choices changed during the initiation of US COVID stay-at-home orders according to age and gender in a population engaged in weight loss. Further research is needed to understand the factors that influenced food choices and how food choices are modified in similar and dissimilar populations.

      Authors' contribution

      EM, Q.Y., H.B., L.D., and P.S. are employed by Noom, Inc. and receive a salary and stock options.

      Funding

      This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. All authors were involved in the writing and preparation of this manuscript and have approved the final article.

      Declaration competing interest

      The authors have no additional conflicts of interest.

      Appendix A. Supplementary data

      The following is the Supplementary data to this article:

      References

        • Flanagan E.W.
        • Beyl R.A.
        • Fearnbach S.N.
        • Altazan A.D.
        • Martin C.K.
        • Redman L.M.
        The impact of COVID-19 stay-at-home orders on health behaviors in adults.
        Obesity. 2020; https://doi.org/10.1002/oby.23066
        • Creswell J.I.
        ‘I just need the comfort’: processed foods make a pandemic comeback.
        New York Times, 2020
        • Pietrobelli A.
        • Pecoraro L.
        • Ferruzzi A.
        • Heo M.
        • Faith M.
        • Zoller T.
        • et al.
        Effects of COVID-19 lockdown on lifestyle behaviors in children with obesity living in Verona, Italy: a longitudinal study.
        Obesity. 2020; https://doi.org/10.1002/oby.22861
        • Shen W.
        • Long L.M.
        • Shih C.-H.
        • Ludy M.-J.
        A humanities-based explanation for the effects of emotional eating and perceived stress on food choice motives during the COVID-19 pandemic.
        Nutrients. 2020; 12: 2712
        • van den Bos R.
        • de Ridder D.
        Evolved to satisfy our immediate needs: self-control and the rewarding properties of food.
        Appetite. 2020; 47: 24-29
        • Leng G.
        • Adan R.
        • Belot M.
        • Brunstrom J.
        • De Graaf K.
        • Dickson S.
        • et al.
        The determinants of food choice.
        Proc Nutr Soc. 2017; 76: 316-327
        • Jacobs S.
        • Radnitz C.
        • Hildebrandt T.
        Adherence as a predictor of weight loss in a commonly used smartphone application.
        Obes Res Clin Pract. 2017; 11: 206-214
        • Mervosh S.
        • Lu D.
        • Swales V.
        See which states and cities have told residents to stay at home.
        New York Times, 2020
        • Rolls B.J.
        • Drewnowski A.
        • Ledikwe J.
        Changing the energy density of the diet as a strategy for weight management.
        J Am Diet Assoc. 2005; 105: 98-103
        • Michaelides A.
        • Major J.
        • Pienkosz Jr., E.
        • Wood M.
        • Kim Y.
        • Toro-Ramos T.
        Usefulness of a novel mobile diabetes prevention program delivery platform with human coaching: 65-week observational follow-up.
        JMIR mHealth and uHealth. 2018; 6: e93
        • Kuijer R.G.
        • Boyce J.A.
        Emotional eating and its effect on eating behaviour after a natural disaster.
        Appetite. 2012; 58: 936-939
        • Uemura M.
        • Ohira T.
        • Yasumura S.
        • Otsuru A.
        • Maeda M.
        • Harigane M.
        • et al.
        Association between psychological distress and dietary intake among evacuees after the Great East Japan Earthquake in a cross-sectional study: the Fukushima Health Management Survey.
        BMJ Open. 2016; 6e011534
        • Zhang W.
        • Ohira T.
        • Abe M.
        • Kamiya K.
        • Yamashita S.
        • Yasumura S.
        • et al.
        Evacuation after the great east Japan earthquake was associated with poor dietary intake: the Fukushima health management survey.
        J Epidemiol. 2017; 27: 14-23
        • Jia P.
        • Liu L.
        • Xie X.
        • Yuan C.
        • Chen H.
        • Guo B.
        • et al.
        Impacts of COVID-19 lockdown on diet patterns among youths in China: the COVID-19 Impact on Lifestyle Change Survey (COINLICS).
        Appetite. 2020; https://doi.org/10.1016/j.appet.2020.105015
        • Sidor A.
        • Rzymski P.
        Dietary choices and habits during COVID-19 lockdown: experience from Poland.
        Nutrients. 2020; 12: 1657
        • Bracale R.
        • Vaccaro C.M.
        Changes in food choice following restrictive measures due to Covid-19.
        Nutr Metabol Cardiovasc Dis. 2020; 30: 1423-1426
        • Rangan A.M.
        • Tieleman L.
        • Louie J.C.
        • Tang L.M.
        • Hebden L.
        • Roy R.
        • et al.
        Electronic Dietary Intake Assessment (e-DIA): relative validity of a mobile phone application to measure intake of food groups.
        Br J Nutr. 2016; 115: 2219-2226
        • Boushey C.J.
        • Spoden M.
        • Delp E.J.
        • Zhu F.
        • Bosch M.
        • Ahmad Z.
        • et al.
        Reported energy intake accuracy compared to doubly labeled water and usability of the mobile food record among community dwelling adults.
        Nutrients. 2017; 9: 312