2.1 Software design
The software was named Nutriecology®, and its design started in May 2018. Development began in June 2019 and was concluded by December 2019. In January 2020, the final tests were carried out. The methodology waterfall life cycle for software development (a multi-stage process methodology) was used to design and develop Nutriecology®. Engineers in informatics and computer systems programmers collaborated in the programming of the software. The multi-stage process described in Figure 1 was used. As can be seen, we included the following five stages regarding requirements and analysis, design, coding, testing, and maintenance: 1) Formative research, 2) Definition of software features, 3) Development of Nutriecology®, 4) Tests for internal validation, and 5) Launch of Live Site. Stages 1 to 4 are presented as part of the methodology (features and development of the software). The results section shows stage 5. It is essential to mention that since the software was designed and developed for Mexico’s context, it was all done in Spanish. This development is already registered in the National Institute of Copyright (INDAUTOR) under the name of Nutriecology® as computational software development with registration number: 03-2022-012812203100-01.
Insert Figure 1. Process followed for the development of the nutritional ecologic software Nutriecology®. Note: FFQ = Food Consumption Frequency Questionnaire.
2.1.1 Ethical considerations
Although no participants were involved in this study, since personal data questions are included in the interface of the software, the study was approved by the Ethics Committee of the University of Guadalajara CEICUC (registration number CEICUC-PGE-004). Also, data collected with this software will be protected by the Federal Law on Protection of Personal Data Held by Private Parties. For further validation studies, the Declaration of Helsinki will be followed. Also, informed consent will be obtained from all subjects involved.
2.2. Stage 1 (Requirements and analysis): Formative Research
Literature research was launched to define the requirements of the software. That included the search for other online nutritional or ecological software to identify the main features provided. Also, a search to identify complete 24-hour recalls, FFQ’s and dietary quality indexes available internationally and those validated for Mexico’s context was performed. Once identified, we interacted with the software available on the web. Besides, informal interviews were done with nutrition and software experts to decide the features to include in Nutriecology®.
At the formative research and international level, the more remarkable nutritional software identified was myfood24®, which is available for the UK, USA, Germany, Denmark, France, Middle East, Norway, Australia, Caribbean, and Peru (3,9). In Mexico’s context, some nutritional software has been designed, such as Nutrimind® (11), Aznutrition® (12), and Nutricloud Nutrición Digital® (10). However, most of them are orientated to provide nutritional consultation, except for Nutricloud Nutrición Digital® (10), a platform designed for nutritional studies that evaluate diet and diet quality using a validated FFQ, 24-hour recall, and a diet quality index (15). Although it does not consider dietary environmental impact assessment, that software was used as a reference for the development of Nutriecology®.
To identify available tools to measure dietary intake in Mexico, we consulted the Nutritools® platform to identify international dietary assessment tools. This platform reports over 20 types of 24-hour recalls and over 79 FFQ’s, but only a Mexican validated FFQ was identified (39). However, elsewhere other FFQ’s used in Mexico at state and national levels were identified. Those include the ones from Macedo-Ojeda et al. (15) and Denova-Gutierrez et al. (5,16), which have been used in the country, both for dietary assessment and WF estimations (6). Concerning diet quality, The most used indexes at the international level are the Healthy Eating Index (HEI) (40,41), the Alternate Index of Healthy Eating Index (AHEI) (42), and the 2015 Dietary Guidelines Adherence Index (DGAI) (43). However, these were developed for specific populations out of Mexico. In the country, some indexes have been created, including the Mexican Diet Quality Index (44), the Mexican Alternate Healthy Eating Index (45), and the Mexican Diet Quality Index (ICDMx) (19).
Although all those were designed based on the Mexican dietary guidelines, each has some limitations, such as no differentiation of meat products, and only the ICDMx includes food sub-groups, was recently validated (19,46), is integrated into Nutricloud Nutrición Digital® (the base for this project) (10), and considers the “correct” diet Mexican term, which refers to a diet to meet the following characteristics: (1) Sufficient as it completely satisfies the nutritional needs, regarding energy, iron, calcium, fiber and water requirements intake; (2) Balanced regarding the proportions of nutrients that make it up (proteins, lipids, and carbohydrates); (3) Complete, as long as it includes foods from three groups: (a) vegetables and fruits, (b) cereals, (c) legumes and animal products; (4) Varied, as it includes different foods from each group, for example, fruits and vegetables of different colors (e.g., red: tomato, strawberries; blue-purple: blackberries, beetroot; yellow-orange: carrots, papaya; green: spinach, lime; and white: onion, banana); different cereals types (e.g., corn, wheat, rice), and different protein sources (e.g., beans, poultry, unsweetened milk); (5) Innocuous, regarding its regular consumption does not represents health risks, for example regarding saturated and polyunsaturated fatty acids, sodium and alcoholic beverages. The sixth aspect, “adequate,” which refers to factors such as the tastes and culture of its consumers and the food's affordability, is not considered in the ICDMx (19,46). The ICDMx gives a score to the participant. A maximum of 100 points represents a diet of the highest quality. To calculate the ‘ICDMx,’ it is necessary to fill out an FFQ or a 24-hour recall (19). However, only the FFQ was considered for this study since it is a more robust instrument to assess the habitual diet (1).
Despite Nutricloud Nutrición Digital® having strengths such as dietary assessment through 24-hour recall and FFQ’s, and diet quality evaluation by the ICDMx, this software has some limitations. The principal one is that the included FFQ lacks several highly consumed foods in the Mexican current and traditional diet, such as pozole, tacos, sopes, chilaquiles, enchiladas (5,6). Also, the ICDMx has significant limitations, especially regarding the aspects “complete”, “innocuous,” and “adequate”. Currently, the “complete” aspect does not differentiate between the foods in the animal-based food group and provides points for consuming and exceeding 120 g of animal products and legumes. No differentiation between animal and plant foods is an important issue both from a nutritional and sustainability perspective. The nutritional composition of milk, yogurt, cheese, eggs, white meats (chicken and fish), red meats (pork and beef), and legumes vary greatly (19,47). Evaluating meats and legumes at the same level generates fundamental problems when addressing diet quality. For example, chronic disease prevention has been related to fish and legumes intake (48,49). Meanwhile, the development of certain cancers has been linked to red meat consumption (50). Regarding environmental impact, the WF of red meats is up to 6 times graters than fish and legumes, for example, a kilogram of fish has a WF of 3,110 liters, a kilogram of beans a WF of 5,789.87 liters, and a kilogram of beef has a WF of 21,566 liters (5,51).
On the other hand, the “innocuous” aspect of the ICDMx does not consider the recommendation of maximum sugar intake in the diet, although it is widely known that excessive sugar intake is related to chronic disease development (52,53). Finally, the ICDMx does not evaluate whether the diet is “adequate” or not according to the concept of the correct diet (19,46). This element is essential in sustainable diets, which must consider economic and social aspects, besides health, nutrition, and the environment (54).
Regarding environmental impact assessment of dietary aspects, the most remarkable softwares were Optimeal® and Agri-footprint® (55), which allow the design of sustainable diets by optimization, and the dietary environmental impact assessment, respectively. Those tools include several environmental impacts indexes, such as terrestrial acidification, freshwater eutrophication, land use, and water consumption (56). However, those tools are not considered nutritional instruments for assessing diet and diet quality in parallel with environmental impact. Also, although those are complete tools for environmental impact assessment, they are only based on the Life Cycle Assessment method and do not provide other environmental perspectives such as the Water Footprint Assessment method (WFA), which is the most used method for evaluating dietary WF (5). Also, the softwares were designed for the European context, for which they do not include highly consumed foods in Mexico and are not contextualized to the country's characteristics. For Mexico’s context, a new methodological approach based on the WFA was recently proposed (5). However, software for dietary environmental impact assessment has not been identified.
2.3. Stage 2 (Design): Definition of software features
A user’s and administrator sections were considered in the design of Nutriecology®. Table 1 presents the characteristics incorporated in Nutriecology® for the user’s segment. In that segment, the following five sections compose the software: 1) Registration, 2) Sociodemographic and nutritional evaluation (body composition data, purchasing information, physical activity), 3) 24-hour recall, 4) FFQ, 5) Diet quality, and 6) Dietary environmental impact assessment.
Insert Table 1. Features of Nutriecology® in users’ section
In section 1, we included the option for the user to register and log in later, as well as an informed consent. For section 2, general information is requested, including the name, last name, phone, email, age, date of birth, sex, place of birth, residence, time living there, and other comments. Educational level was included, with the option to respond according to the Autonomous University of Mexico (UNAM) (57) based on the last group of studies. Following, Nutriecology® included the National Institute of Statistic and Geography (INEGI) classification for occupational level (Supplementary material 1) (60).
Additionally, this section included the type of work, workdays, schedule, and monthly income of the user. All these aspects have been reported as reliable and practical data for assessing sociodemographic aspects in the Mexican population (61). Other aspects relevant to nutrition were included, such as food purchasing data, monthly money spent on food, principal place of purchasing, days a week eating out of home, and types of food consumed out of the house. Also, physical activity evaluation was included, following the International Physical Activity Questionnaire (IPAQ), including questions related to the type of physical activity, weekly frequency, minutes a day, and intensity (low, moderate, and high) (59). Finally, in section 2, an open-ended question about disease presence was added, which included asking for disease type, time since diagnosis, and medication used. Also, Nutriecology® incorporated a section for registering the user's anthropometric and body composition data. These data included estimated self-report (by the user) weight and height in case the user was not nutritional measured (i.e., weighed and measured by the study investigator). In case a nutritional evaluation was performed by the research team using Nutriecology®, blank spaces to fill in were established for information related to body composition. The data included were height, weight, body mass index, percentage of body fat, muscle mass, corporal water, metabolic rate, metabolic age, visceral fat, waist and hips circumferences, and bone mass. Nevertheless, the software allows users to provide self-reported data for these parameters for cases where the researcher asks subjects to enter their information.
Section 3 considered the most common characteristics used in 24-hour recalls. Those included food time divided into breakfast, morning snack, meal, evening snack, and dinner. Also, it included a space for the exact hour of the eating episode, the place of consumption, the menu (set of ingredients included in the dish), or the preparation method (grill, boiled, fried). Besides, it is presented a displayable list of foods with images and quantities of habitual consumption. A displayable list of food portions was included too, for a more precise dietary analysis. Additionally, it was added a question about water consumption.
Section 4 included an adaptation of two of the most used FFQs in Mexico. The base was the questionnaire of Macedo-Ojeda et al. (15), which was used in Nutricloud Nutrición Digital® and included 162 food items. Also, it provides more specific food options such as skim, semi-skim, whole, or condensed milk, or chicken with or without skin, instead of only milk or chicken, as in other FFQ (16,19,62). That FFQ was adapted based on the questionnaire of Denova-Gutiérrez et al. (16), which is used in The National Survey of Health and Nutrition (ENSANUT) from Mexico and includes 140 food items. We also used the food list reported in an exploratory study searching for consumed foods in Mexico (5). Some foods that were not included in the analyzed FFQs, but were reported in that list include chilaquiles, burritos, chickpeas hummus, vegetable drinks such as almond and coconut “milk”, berries, peanut butter, rice cakes, quinoa, sweet potato cooked with sugar, oats prepared with milk and sugar, and whole versions of rice and wheat tortillas, among others (5). Although some of those foods are not considered basics in the Mexican current diet, there is a growing trend toward its consumption (1).
In total, we included a food list of 248 items (Supplementary Materials 2 and 3). The consumption frequencies range options were also adapted to provide a more accurate and comprehensive analysis regarding the consumption temporality (63). The frequencies were extended to yearly frequency, from 1 to 5 times and 6 to 11 times (3 and 8.5 times per year, respectively, on average). The monthly frequencies were divided into 1, 2, and 3 times a month. The same was done for the weekly frequency, from 1 to 7 times a week. In addition, in this adaptation of the FFQ, the frequency per day was included according to the frequency of consumption previously reported (annual, monthly, or weekly). Considered daily frequencies were 1, 2, 3, or 4 or more (4+) times by day. The number of eaten portions was also included, considering established units: ¼, ½, 1, 2, 3, 4, 5, or 6 or more (6+). Therefore, for calculating the average daily amount of consumption of a specific food, according to frequency and number of portions, the following formula must be followed:
Where G is the average amount of food consumed a day; g is the reported consumed portion in grams or milliliters (establish portion in the FFQ); m is the frequency of consumption a month (i.e., five times a week will be multiplied for four weeks a month, so consumption of 5 times a week will be equal to 20 times a month); d is times consumed a day (i.e., in breakfast and dinner); p is the number of portions consumed (i.e., two cups each time the food is consumed); divide between 30 days a month.
The portions were obtained from both questionnaires (15,16) and the food list from Lares-Michel et al. (5) and correspond to portions of habitual consumption. Supplementary Material 3 presents the written version of the designed FFQ. For the digital version of the FFQ at Nutriecology®, reference images of each of the 248 foods were attached, which were taken by us or were obtained from free internet downland (64).
Both for 24-hour recalls and FFQ nutritional composition, we used the Mexican System of Equivalent Foods (SMAE) (47) and the Mexican Tables of Food Composition of Ledesma et al. (65). Also, labels of foods from different brands not available in nutritional composition tables were consulted. Analyzed labels included cookies, sauces, sweet bread, packaged ice cream, milkshakes, granola bars, instant soups, candy, and chips, among others. Complete data regarding label analysis is presented in Supplementary material 4. For food groups and food sub-groups classifications, we also relied on the Mexican System of Equivalent Foods (SMAE) (47) and the Mexican Tables of Food Composition of Ledesma et al. (65). However, the environmental impact of foods was also considered, especially in the food sub-groups classification (5,52). Supplementary Material 2 provides the complete food classification.
Regarding section 5, which corresponds to the diet quality index, the Mexican Diet Quality Index (ICDMx) was used as a reference but was adapted. All the components of ICDMx were included, incorporating the aspects 1) sufficient, 2) balanced, 3) complete, 4) varied, and 5) innocuous. However, the index was modified in its "complete" and "innocuous" aspects. Nutriecology® also included the “adequate” aspect, which is not included in the ICDMx (46). Table 2 presents the complete values considered for the calculations. Supplementary material 5 shows the specific instructions for the calculations. Based on the modifications performed, the adapted version of the ICDMx was called as the Alternate Mexican Diet Quality Index (IACDMx for its initials in Spanish). The dietary data for the automatic calculation of the IACDMx was taken only from the FFQ as it provides dietary data from a longer period regarding the 24-hour recall (1).
The first adapted aspect (complete) was modified based on the subdivision of the group of foods of animal origin and legumes. Instead of evaluating the consumption of 120 g or more of those foods, as suggested in the ICDMx (19), we separately assessed the consumption of 1) at least 60 g of legumes, 2) less than 71 g of red, and industrialized meats, 3) less than 56 g of poultry (chicken and eggs), 4) at least 26 g of fish and shellfish, 5) less than 240 ml of milk, and yogurt, and 6) less than 40 g of cheeses. Additionally, the intake of 7) at least 400 g of fruits and vegetables and 8) at least 200 g of cereals without fats were evaluated. The recommended grams of consumption per day are shown in Table 2 and were obtained from Bonvecchio Arenas et al. (66) and the Ministry of Health (67). This adaptation was carried out by what was suggested by Hawkes et al. (68), who insist that diet quality indexes should not only focus on complying with recommended portions of certain food groups or nutrients, as was traditionally carried out to combat undernutrition but evaluating not exceeding portions must be included, because of the obesity pandemic that the world is facing. In this sense, the evaluation of sugar intake concerning recommendations (not exceeding 10% of energy intake according to the World Health Organization, OMS) was added in the "innocuous" aspect (69).
Regarding the “adequate” aspect, we attached the previously described definition of normativity in Mexico (46). Based on that, a literature search was performed to identify the average weight of Mexican traditional foods and dishes. From the revision, an average amount of 30 g was considered for Mexican foods since is the approximate weight of a corn tortilla, which is one of the most traditional Mexican foods consumed daily by the majority of the population (47,70). An average weight of 180 g was identified for typical Mexican dishes such as pozole, tamales, sopes, tacos, and enchiladas, among others. Nevertheless, based on the revision, it was considered “adequate” to consume at least one traditional Mexican dish a week (47). Therefore, the “adequate” aspect evaluates the consumption of 30 or more grams per day of foods considered Mexican and the intake of 180 g or more of Mexican dishes per week, equivalent to 26 g per day. Likewise, it was evaluated the consumption of less than 30 g of foods considered westernized and the intake of less than 180 g of westernized dishes (<26 g per day). For the determination of western foods' weights, a literature search was also performed, and similar trends were identified for foods such as packaged bread and chips (30 g), and burgers (180 g). Therefore, 30 g/day and 180 g/week (<26 g/day) were also considered for unifying Mexican and Western foods. Supplementary Material 2 (Table S2) shows the food classification of Mexican foods and dishes and western foods and dishes (47,70–75).
Unlike the quantitative diet quality proposed by Macedo-Ojeda et al. (19), this element was evaluated qualitatively. Instead of numbers, the diet was qualified as 'inadequate' or 'adequate' with letters from a to d, according to adherence to the parameters established and shown in Table 2. Additionally, for providing grades to the level of ‘adequate’, a tertile analysis was performed based on the nutritional data reported in Lares-Michel et al. (76), where a dietary and environmental impact assessment was performed in a representative sample of Mexican population. Full description of this analysis is shown in Supplementary Material 5.
It is worth mentioning that the scores of the other components of the ICDMx were rearranged to obtain a maximum score of 20 points per aspect in the IACDMx. In other countries, such as the United States, these modifications to diet quality indices have been reported. An example is the Healthy Eating Index and its modified version, the Alternative Healthy Eating Index (AHEI) (77).
Insert Table 2. Calculation and components of the Alternate Quality Index of the Mexican Diet (IACDMx)
Regarding dietary environmental impact assessment, Nutriecology® included dietary WF calculation. WF is the volume of freshwater needed to produce a good or service, including food. This index comprises three WF types: green, blue, and grey. In a food production context, green WF quantifies the amount of rainwater stored in the soil that is used by crops within the evapotranspiration process. The blue WF estimates the amount of water used for agricultural irrigation. Grey WF is considered the water necessary to assimilate a certain pollutant load, such as water used in food industries, food washing, and cooking (i.e., boiling, scalding, poaching, and stewing). The sum of those three WFs is considered the total WF of a product or food (5,30).
The method of Lares-Michel et al. (5), which is based on WFA method (30), was carefully followed and charged into Nutriecology® for the automatic calculation of the total, green, blue, and grey WF of each of the 248 foods from the FFQ. The first step for WF calculation is identifying the daily consumption of each food (g or ml), by the 24-hour recall or the FFQ. The software identifies if the food requires a process before consumption (washing, cooking, or peeling). For example, a cooked potato needs to be washed, peeled, and cooked; meanwhile, milk is directly consumed as was bought. For converting food already processed to its raw version, the formulas and correction factors provided in Lares-Michel et al. (5) were used (for example, 1.45 for meats). In Supplementary Material 2 we present the foods to which correction factors were applied into Nutriecology®. Once food raw weight is obtained, the software determines if washing and/or cooking is required before consumption. If washing and/or cooking were needed, the software considered the grey WF reported by Lares-Michel et al. (5). That corresponds to 10 liters/kg for cereals and legumes, 1 liter/kg for meats, and 14.44 liters/kg for food in general. The foods for which washing and/or cooking water was considered are presented in Supplementary Material 2. After those steps, the WF of each food is quantified using the formulas and databases provided in Lares-Michel et al. (5), which considers green, blue, and grey WFs. For foods composed of more than one ingredient, the individual's WF of each food was summed according to the corresponding formulas or using already calculated data for Mexican dishes (5).
Considering all those aspects in dietary WF calculation is essential for accurate dietary environmental impact assessments. In previous studies, the WF quantifications have been carried out in cooked (or ready-to-eat) versions, but the environmental data was calculated in raw food (5,78). That generates essential variations that can lead to WFs up to 135% higher or lower than the values calculated, considering correction factors and quantifying the water involved in washing and cooking food. An example of these variations can be provided for beef. Before cooking, beef has an average weight of 145 g, but after cooking, it loses water and weighs 100 g. That would generate to report a WF of 2,156.60 liters instead of 3,127.07 liters (5).
Regarding Nutriecology® administrator section, Table 3 shows the sections integrated there. Section 1 allows the administrator's log in with a pre-established account and password. Once into the software, a menu is presented to the administrator. All available options are presented in Table 3, which includes viewing, editing, adding, or eliminating foods, food groups, and sub-groups. Also, a section for data exportation and importation from excel sheets is available.
Insert Table 3. Features of Nutriecology® in the administrator section
Nutriecology® included 33 nutrients and environmental aspects that are automatically calculated. Those are presented in Table 4, and include energy, macro, and micronutrients, including vitamins and minerals. Regarding the environmental aspect, Nutriecology® currently only includes green, blue, and grey WF, as well as water involved in cooking and food washing.
Insert Table 4. Nutritional composition and environmental impact aspects incorporated in Nutriecology®
2.4. Stage 3 (Coding): Development of Nutriecology®
The development of Nutriecology® was done in collaboration with software developers, engineers in informatics, nutritionists, environmental science researchers, behavioral science experts, and psychologists. The methodology for software development “waterfall life cycle”, was applied in all stages. This process was performed following the workflow shown in Figure 1. During the development of Nutriecology®, in-house testing and feedback were constantly done by the research team and the software programmers. In-house testing and feedback included aspects regarding interface colors, letter type, presentation of the tools, and excel sheets, among other aspects (79).
2.5. Stage 4 (Testing): Tests for internal validation
The validation of Nutriecology® was done by calculating nutritional composition and WF manually and using the software. Then both calculations were compared to ensure the software is providing accurate data. The manual calculations were performed using Excel® sheets and the formulas provided there. Both for the first and second sections of the software, the correct registration and user information were verified manually by checking entered data and exportable excel sheets. For the third section, the nutritional composition and WF of 24-hour recalls were calculated by multiplying or dividing the amounts consumed by each nutrient. For WF data, the method of Lares-Michel et al. (5) was manually performed, applying the correction factors suggested there, accounting for the water used in cooking and washing, and using the WF datasets which were loaded to Nutriecology®. For the fourth section, the nutritional composition and WF of the FFQ (Supplementary Material 3) were calculated according to formula 1, multiplying or dividing by datasets of nutrients and WF. For the five-section, diet quality was both calculated manually and by using Nutriecology®. Supplementary Material 5 presents the manual calculation steps followed for diet quality calculation using the IACDMx. Finally, for WF calculations, both the WF from the 24-hour recall and FFQ sections were re-checked by following the method of Lares-Michel et al. (5).