This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
Although Pinterest has become a popular platform for distributing influential information that shapes users’ behaviors, the role of recipes pinned on Pinterest in these behaviors is not well understood.
This study aims to explore the patterns of food ingredients and the nutritional content of recipes posted on Pinterest and to examine the factors associated with recipes that engage more users.
Data were collected from Pinterest between June 28 and July 12, 2020 (207 recipes and 2818 comments). All samples were collected via 2 new user accounts with no search history. A codebook was developed with a raw agreement rate of 0.97 across all variables. Content analysis and natural language processing sentiment analysis techniques were employed.
Recipes using seafood or vegetables as the main ingredient had, on average, fewer calories and less sodium, sugar, and cholesterol than meat- or poultry-based recipes. For recipes using meat as the main ingredient, more than half of the energy was obtained from fat (277/490, 56.6%). Although the most followed pinners tended to post recipes containing more poultry or seafood and less meat, recipes with higher fat content or providing more calories per serving were more popular, having more shared photos or videos and comments. The natural language processing–based sentiment analysis suggested that Pinterest users weighted
Although popular pinners tended to post recipes with more seafood or poultry or vegetables and less meat, recipes with higher fat and sugar content were more user-engaging, with more photo or video shares and comments. Data on Pinterest behaviors can inform the development and implementation of nutrition health interventions to promote healthy recipe sharing on social media platforms.
Healthy eating patterns and their effect on disease prevention have been demonstrated robustly across the scientific literature [
Social media has become a new and efficient way to distribute and consume influential information that shapes people’s dietary behaviors [
Pinterest, launched in 2010, is a unique social media platform where users can save images (
To the best of our knowledge, no prior study has evaluated the recipe content on Pinterest. This study provides a first glimpse of this domain to advance the understanding of the relationship between social media use and dietary behavior. We aim to achieve the following 2 goals. First, we aim to examine the patterns of food ingredients and nutrients prescribed by recipes posted on Pinterest. Second, by employing both traditional content analysis and a natural language processing (NLP) technique, we sought to understand the factors that distinguish the most popular recipes among users.
Data were collected between June 28 and July 12, 2020. Although there is no “rule of thumb” on how long the data collection should persist, we adapted a proper time frame based on previous literature that specifically focused on Pinterest [
For the content analysis, following the 2015-2020 United States Department of Agriculture Dietary Guidelines, food ingredients were classified as dark green vegetables, red and orange vegetables, legumes (beans and peas), starchy vegetables, other vegetables, fruits, seafood, meats, poultry, eggs, nuts or seeds or soy products, dairy, oil, and butter [
For the comment analysis, 3 keyword dictionaries were created with keywords related to health (eg,
Descriptive analyses were performed for each type of food ingredient and their corresponding nutrient content. In addition, the popularity of recipe ingredients was assessed by the number of recoded followers (presented in tertiles). The level of engagement for each recipe was also evaluated by categorizing comments and shared photos or videos into tertiles, with regard to the fat, sugar, and fiber content of the recipes. Comments and shared photos or videos were chosen as indicators of engagement based on prior literature that suggested that, in the context of Pinterest, the number of
Process of data collection and analysis.
The ingredient and nutrient distributions of meat, poultry, seafood, and vegetable recipes.
Recipes | Meata | Poultrya | Seafooda | Vegetablea,b with eggsc (n=59), mean (SD) | ||||||||||||||
|
Meat only (n=25), mean (SD) | Meat with vegetablec (n=45), mean (SD) | Poultry only (n=35), mean (SD) | Poultry with vegetablec (n=76), mean (SD) | Seafood only (n=6), mean (SD) | Seafood with vegetablec (n=13), mean (SD) |
|
|||||||||||
|
||||||||||||||||||
|
Dark vegetable | N/Ae | 75.6 (0) | N/A | 49.6 (14.1) | N/A | 29.5 (15.4) | 113.3 (0) | ||||||||||
|
Red and other vegetable | N/A | 90.6 (189.8) | N/A | 29.5 (21.3) | N/A | N/A | 72.5 (88.6) | ||||||||||
|
Legumes and beans | N/A | 42.5 (0) | N/A | 81.5 (35.1) | N/A | N/A | 70.8 (0) | ||||||||||
|
Starchy vegetable | N/A | 132.3 (26.8) | N/A | 81.5 (35.1) | N/A | N/A | 144.7 (129.2) | ||||||||||
|
Meat | 124.8 (128.5) | 100.7 (98.2) | N/A | N/A | N/A | N/A | N/A | ||||||||||
|
Poultry | N/A | N/A | 106.7 (94.5) | 100.4 (88.9) | N/A | N/A | N/A | ||||||||||
|
Seafood | N/A | N/A | N/A | N/A | 122.8 (66.2) | 118.4 (73.0) | N/A | ||||||||||
|
Eggs | 51.4 (111.0) | 51.4 (111.0) | 14.5 (5.7) | 16.6 (6.3) | N/A | N/A | 28.3 (23.4) | ||||||||||
Total energy (calories per serving) | 490.9 (280.1) | 473.8 (246.8) | 433.3 (225.5) | 442.7 (185.7) | 320.6 (126.5) | 329.3 (178.5) | 293.6 (131.6) | |||||||||||
|
||||||||||||||||||
|
Fat | 30.1 (17.9) | 25.6 (17.4) | 22.3 (13.8) | 20.9 (13.0) | 18.5 (9.6) | 16.7 (13.7) | 14.8 (9.7) | ||||||||||
|
Percentage of energy from fat (%)f | 56.6 | 48.5 | 46.7 | 41.8 | 50.5 | 40.4 | 44.8 | ||||||||||
|
Protein | 26.4 (11.9) | 28.5 (19.2) | 32.6 (19.5) | 36.8 (50.2) | 23.5 (14.9) | 27.1 (16.6) | 12.3 (10.0) | ||||||||||
|
Percentage of energy from protein (%)f | 24.2 | 25.2 | 30.9 | 34.9 | 30.7 | 33.1 | 16.4 | ||||||||||
|
Carbohydrates | 24.7 (17.7) | 27.3 (20.0) | 20.9 (24.2) | 27.1 (24.6) | 12.6 (17.1) | 16.3 (17.1) | 29.1 (24.1) | ||||||||||
|
Percentage of energy from carbohydrates (%)f | 22.7 | 25.6 | 17.9 | 23.8 | 14.5 | 24.9 | 38.3 | ||||||||||
|
Fiber | 2.2 (2.) | 3.8 (7.6) | 1.3 (1.2) | 2.9 (3.6) | 0.8 (0.7) | 1.7 (2.1) | 4.0 (5.1) | ||||||||||
|
Sodium (mg per serving) | 861.2 (655.9) | 775.6 (683.3) | 911.0 (510.6) | 767.0 (559.5) | 489.8 (431.6) | 777.3 (659.1) | 483.6 (389.1) | ||||||||||
|
Sugar | 6.5 (12.9) | 6.7 (11.6) | 5.7 (8.0) | 6.6 (7.9) | 1.3 (1.0) | 3.5 (4.2) | 4.73 (7.81) | ||||||||||
|
Cholesterol | 114.6 (135.5) | 97.2 (113.1) | 126.5 (71.4) | 109.3 (69.8) | 98.5 (100.6) | 152.9 (177.2) | 47.6 (66.5) |
aRecipes that included a main dish only and those that included a main dish served with vegetables were mutually exclusive. For example, a meat-only recipe was defined as a recipe that only included meat, whereas recipes that included both meat and vegetables were listed in the meat with vegetable category.
bNo recipes were purely vegan; therefore, we reported ovo-lacto recipes.
cThe calculation of sample average used only complete data, that is, some of the denominators were smaller than 45 and did not have standard errors.
dThe food ingredients were categorized based on the guidelines of the United States Department of Agriculture.
eN/A: not applicable.
fThe sum of column percentages of each recipe class may exceed 100% because each value was calculated separately as the percentage of energy from specific nutrients divided by the total energy provided in a recipe class.
The bar charts in
Comment samples and polarity.
Comment | Sentiment polarity | Keyword | Topic |
“How long do I leave them in the oven?” | Neutral | N/Aa | N/A |
“How many calories is this?” | Neutral | Calories | Health |
“I do not like brown sugar in my meatloaf...ugh” | Negative | N/A | N/A |
“Definitely way too salty and too greasy for me.” | Negative | Salty and greasy | Taste |
“It’s easy! I did this again and LOVED it!” | Positive | Easy | Complexity |
“It turned out amazing!! Very delicious.” | Positive | Delicious | Taste |
aN/A: not applicable.
Pinterest users’ attitudes toward different aspects of recipes.
In this study, recipes posted on Pinterest were collected, analyzed, and compared for ingredients and nutrients. We found that, in most cases, recipes using seafood or vegetables as the main ingredient, for example, tuna salad (main ingredient: tuna; other ingredients: celery, onion, flat-leaf parsley, mayonnaise, mustard, and black pepper), had, on average, fewer calories and less sodium, sugar, and cholesterol than meat- or poultry-based recipes, for example, crispy chicken wraps (main ingredient: popcorn chicken; other ingredients: tomatoes, cheddar cheese, buffalo wing sauce, and flour tortillas) and Mongolian beef (main ingredient: flank steak; other ingredients: cornstarch, canola oil, ginger, garlic, soy sauce, dark brown sugar, and scallions). Recipes using meat as the main ingredient, for example, creamy herbed pork chops (main ingredient: pork chops; other ingredients: milk, Montreal steak sauce, butter, flour, basil, black pepper, and instant beef bouillon granules), provided more energy by fat. Although the most followed pinners tended to post recipes containing more poultry or seafood and less meat, recipes serving higher fat or providing more calories per serving were more popular, having more shared photos or videos and comments. Sentiment analysis based on text mining showed that Pinterest users, in general, valued taste more than health qualities when making comments or sharing photos or videos.
With the sharp increase in the number of social media users, platforms such as Pinterest have become influential mechanisms to transform knowledge sharing and acquisition, including dietary choice [
From the perspective of content providers, we found that the most popular pinners, by sharing recipes containing more seafood and poultry (
From the users’ perspective, they are often learners in pursuit of inspirational recipes [
There appears to be a discrepancy between what pinners posted and how users consumed information, leading to an opportunity for future health interventions via Pinterest. Previous studies have shown that social media interventions can have a positive effect on nutritional outcomes [
This study had some limitations. First, because of the restrictions imposed by Pinterest, the content scrolling process is not automated. The manual data collection resulted in a relatively small sample size and a large margin of error. To address the issues related to the small sample size, we applied a machine learning technique to mine text from the comments. A total of 100 comments were randomly selected to assess sentiment error rates. We found that the error rate was 18%, which is better than the acceptable level used in previous studies by convention [
In this study, we used both content analysis and NLP techniques to analyze recipes posted on Pinterest. Seafood-based recipes and vegetarian recipes had fewer calories and less sodium, sugar, and cholesterol than meat-based recipes. Although the most popular pinners tended to exhibit more health consciousness by posting recipes with more seafood, poultry, and vegetables and less meat, recipes with higher fat and sugar content had higher user engagement, as demonstrated by the higher numbers of photo or video shares and comments. Population health could be improved with targeted interventions to address this disparity through efforts to enhance interest in and adoption of healthy recipes by Pinterest users.
Relationship between pinners’ popularity and their recipe ingredients.
Association between popularity and recipes.
natural language processing
valence aware dictionary and sentiment reasoner
HX, XC, and SL made significant contributions to the conception and study design. HX, XC, and SL conducted data analyses. HX, KW, AH, DG, XZ, JW, and LC provided significant support for the interpretation of results. HX, XC, SL, KW, AH, DG, XZ, JW, and LC drafted the paper. All authors approved the final manuscript submitted.
None declared.