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.
Presently, dietary management approaches are mostly oriented toward using calorie-counting and diet-tracking tools that draw our attention away from the nutritional value of our food. To improve individuals’ dietary behavior, primarily that of people with type 2 diabetes, a simple technique is needed to increase their understanding of the nutritional content of their food.
This study aimed to design, develop, and evaluate a customized nutrient-profiling tool called
We evaluated the utility of
The intervention group (n=9) exhibited a statistically significant change between the pre- and postexposure results of their HbA1c (
This study adds to the evidence base that a nutrient-profiling strategy may be a modern adjunct to diabetes dietary management. In conjunction with reliable dietary education provided by a registered dietician,
Keeping up with a healthy, well-balanced diet is by no means easy. This is primarily challenging for patients with type 2 diabetes because it is related to maintaining a certain metabolic goal. The traditional dietary methods to manage diabetes include carbohydrate intake monitoring and
Another problem with the current traditional dietary management methods is the focus on single nutrients instead of the overall nutritional value. Carbohydrate-restricted, fat-restricted diets are examples of these kinds of diets. The American Diabetes Association (ADA) released a new statement in 2013 summarizing the nutrition therapy recommendations for people with type 2 diabetes. As people eat food and no single nutrient such as carbohydrates, protein, and fat, this new statement designates a new section on eating patterns or plans. Patients with diabetes can still enjoy the food they like while, at the same time, keeping their diabetes under control. In addition, educating people about nutrition has to be aimed at modifying the factors that influence their dietary behavior rather than being aimed at increasing their knowledge about nutrition. Bader et al [
In addition, one more issue concerning the current dietary management tools is identified by the ADA in their new position statement on nutrition therapy recommendations. The ADA states that there is no single eating pattern that is best for everyone [
Motivated by the aforementioned issues, we have developed a customized easy-to-follow dietary tool called
Although genetics are an important consideration in health, during the past half-century, our genes have not measurably altered, and yet, we are significantly more overweight, obese, and prone to lifestyle-related diseases. As of 2014, more than one-third (36.5%) of US adults have obesity, according to the National Health and Nutrition Examination Survey data (2011-2014) [
Managing diets is essential when it comes to diet-related chronic diseases. For healthy individuals, it is a preventative measure to maintain a healthy weight and facilitate overall well-being. Essentially, there are 4 different approaches to manage diets, summarized in a study by Arens-Volland et al [
Many studies have been conducted to develop and evaluate computerized dietary management approaches that are based on diet recall and food records [
First, the underlying method, where users have to log their daily food intake, can suffer from the issue of recall. Examples of these food and calorie tracker apps include
To better manage diets and sustain a healthy lifestyle, one has to be aware and knowledgeable about the nutritional content of the food consumed. Being educated and aware of macronutrition and micronutrition would contribute to one’s overall health. This knowledge leads to a healthy behavior, which, in turn, holds the promise of preventing the onset of chronic diseases or mitigating their severity. Individualized diet education is effective to both understand diet requirements and control body weight and blood sugar levels [
Nutritional profiling aims to rank food based on their nutritional quality. It is driven by the focus on food quality instead of quantity. Individuals who follow high-score food choices would most likely improve their dietary behavior.
Simplicity is key when it comes to presenting nutritional information. It has been suggested that nutritional information on mobile phones should be easy to read and understand. Arsand et al [
This approach emphasizes the idea that carbohydrate and calorie intake counting has become less preferred in favor of more generalized nutritional information about the quality of food compositions. As noted, nutrient profiling is defined as the science of ranking foods according to their nutritional composition for reasons related to preventing disease and promoting health, as stated by the World Health Organization. This idea has led to the creation of many nutritional rating systems.
Driven by the idea of nutrition profiling, the traffic-light diet was developed by Epstein et al in the 1970s [
Many studies have been conducted utilizing the
In this study, we adopted the presentation approach of the traffic-light diet to present the nutrients contained in a food recipe. However, it was not a strict tricolor output. Rather, it was a color-coded food rating scale of eight values as it takes into consideration five different nutrients and not only the caloric count. It scales food based on its nutritional quality from red, for extremely unhealthy choices, to green, for optimal healthy choices, through intermediate colors.
This study aimed to provide a nutrition educational tool to help people with type 2 diabetes learn about the nutritional content of the food they eat and, hence, improve their dietary behavior by choosing healthier, more nutritional recipes.
This study followed the design science research (DSR) approach suggested by Hevner and Chatterjee [
Design science research methodology.
Using
This scale will give the user an initial indication of how nutritious the chosen recipe is. If the user is interested to know more about the nutrients that lower the overall nutritional quality, he/she can click on the
As the user saves and adds food recipes to his/her meal plan, he/she will be given either a gold, silver, or bronze trophy based on how active they are with the app. The app will categorize the user into one of the three different groups based on the number of recipes he/she adds to his/her meal plan.
EasyNutrition Sign up screen.
EasyNutrition Home page.
The overall nutritional value of the “Falafel salad” recipe.
Nutrition break down of the recipe.
We developed the
All these nutrients’ amounts/percentages are based on one’s daily calorie intake. As the recommended calorie intake differs based on age, gender, height, and weight, we applied the Harris-Benedict equation to determine the recommended calorie intake or the BMR based on these factors [
The cursor would start right in the middle of the traffic-light scale as an initial score for any given food recipe, as can be seen in
The nutrition limits per daily recommended percentages/amounts of total calorie intake.
Nutrients | Intake | |
|
||
|
Carbohydrates |
45%-65% |
|
Fats |
25%-35% For those with hypertension, no more than 7% of this percentage should come from saturated fat |
|
Protein |
15%-20% |
|
||
|
Dietary fibers |
20-30 grams |
|
Sodium |
No more than 2300 mg No more than 1500 mg daily for diabetics who have hypertension |
Calculate the basal metabolic rate (BMR) based on gender, age, height, and weight based on the Harris-Benedict equation
Divide the number by 3 assuming 3 meals a day, to get the ideal recipe calories for a particular user
Recommended recipe calories (RRC)=BMR/3
Obtain the chosen recipe calories (CRC) and compare the value with RRC
If CRC≤RRC, and is within the recommended range, CRCw=1, otherwise CRCw=−1
Obtain the amount of carbs in grams
Convert this amount into calories (1 gram of carbs provides 4 calories)
0.45%
If it is within the recommended range, Cw=1, otherwise Cw=−1; if
Obtain the amount of fats in grams
Convert this amount into calories (1 gram of fat provides 9 calories)
0.25%
If it is within the recommended range, Fw=1, otherwise Fw=−1; if
If the patient with diabetes has hypertension, then another test for saturated fat will be conducted: fat sugars should not exceed 0.07% of the total fat
Obtain the amount of protein in grams
Convert this amount into calories (1 gram of protein provides 4 calories)
0.15%
If it is within the recommended range, Prw=2, otherwise Prw=−2; if
Obtain the amount of sodium in milligrams
If
If the patient with diabetes has hypertension, then the amount has a lower cut-off value (<1500/3 mg)
Obtain the amount of dietary fibers in grams
6
If
The nutritional score presented behind a traffic-light scale.
During this study, we evaluated the utility, efficacy, and quality of the app,
The evaluation plan (subjects, measures, and analysis).
Subjects were recruited from the diabetes treatment center (DTC), Loma Linda University Health. The center offers classes on diabetes education in 4 different sessions: the first and second sessions are on nutrition, the third one is on physical activity, and the last one is after 2 to 3 months of follow-up with the patients to provide a plan for them to move forward and live with diabetes. Diabetes educators and registered dieticians facilitate those sessions. We attended the sessions to introduce
In the first class, we started the recruitment by introducing ourselves and our study and explained how relevant it is to their diabetes education session. This included playing a 2-min short demo about
First, as part of the DTC standard of care (SOC), subjects’ blood was drawn to measure their blood sugar (glycated hemoglobin, HbA1c), and this was self-reported to the research personnel. The HbA1c is measured in units of mmol/mol. The HbA1c test measures how much hemoglobin in the blood has become glycated (chemically bonded with glucose).
Second, researchers assigned participants either to the intervention group (group number 1) or to the control group (group number 2), based on the kind of smartphone he/she had (patients with an Android-based smartphone were assigned to group number 1).
Third, participants answered a quick questionnaire about their EH using the Health-Promoting Lifestyle Profile (HPLP) nutrition subscale. The administered questionnaire was adapted from the HPLP, a 48-item questionnaire of self-reported health-promoting lifestyle habits. HPLP consists of 6 scales: self-actualization, health-responsibility, exercise, nutrition, interpersonal support, and stress management. The final structure of the 48-item HPLP was found to have high internal consistency with an alpha coefficient of .922. Sets of items assigned to each of the 6 factors were examined for their reliability as subscales. The nutrition subscale is found reliable with an alpha coefficient equal to .757.
We utilized the Nutrition subscale, a 10-item, 4-point Likert scale that ranges from Never (1), Sometimes (2), Often (3), to Routinely (4) to indicate how often an individual engages in each behavior. The minimum score was 10, and the maximum score was 40 [
On the basis of participants’ answers to these questions, we generated a score reflecting their EH. This score was used as a baseline for postcomparison for both groups. An increase in this score suggests improvement in the patient’s dietary behavior. This can establish a potential association between the intervention (interacting with
If the patient was assigned to the intervention group, he/she was asked to download the app,
Fourth, at the end of the study, all subjects filled in the same questionnaire about their EH. In addition, their HbA1c was measured again to compare it with the benchmark data from the presurvey data. The follow-up classes at the DTC we attended were at the following dates/times: Monday, February 26, 2018 (5:30-7:30 PM); Wednesday, February 28, 2018 (9 AM-11 AM); Monday, March 26, 2018 (1:30 PM-3:30 PM); and Wednesday, March 28, 2018 (5:30 PM-7:30 PM). One class was designated for only those with prediabetes, and it was offered under the diabetes prevention program. This was on Wednesday, March 14, 2018 (5:30 PM-7:30 PM).
In addition, all subjects were asked to fill in the nutrition self-efficacy (NSE) questionnaire [
We compared the mean of both eating habit composite scores and HbA1c before and after the study for both groups (intervention and control). Thus, a series of independent and dependent sample one-tailed
Subjects are patients with either type 2 diabetes or prediabetes. Out of the 28 patients who signed up in the third stage, 21 (75%) completed their participation. Of these, 9 patients were men and 12 were women, with the majority being white (11 out of 21). To measure
Participants in both groups (9 subjects in the intervention group and 12 subjects in the control group) were administered the presurvey to determine the pre-exposure composite scores of their EH and, hence, establish a benchmark healthy score. The mean composite score for EH was 23.33 for the intervention group and 25.83 for the control group. As for the HbA1c, the mean was 8.13 for the intervention group and 7.13 for the control group (
Descriptive statistics on the eating habits composite score and glycated hemoglobin before and after the experiment.
Group | Values | |||
|
Mean (SD) | SE mean | ||
|
||||
|
|
|||
|
|
Pre-HbA1c | 8.13 (2.83) | 0.94 |
|
|
Post-HbA1c | 6.72 (1.27) | 0.42 |
|
|
|||
|
|
Pre-EH | 23.33 (3.08) | 1.02 |
|
|
Post-EH | 24.33 (3.54) | 1.18 |
|
||||
|
|
|||
|
|
Pre-HbA1c | 7.13 (1.49) | 0.43 |
|
|
Post-HbA1c | 6.90 (1.64) | 0.47 |
|
|
|||
|
|
Pre-EH | 25.83 (5.27) | 1.52 |
|
|
Post-EH | 26.08 (2.94) | 0.85 |
aHbA1c: glycated hemoglobin.
bEH: eating habits.
Pre-exposure independent
Group | Levene test | |||||
|
Significance ( |
Significance (two-tailed; |
Mean difference | SE difference | ||
Eating habits | 3.39 | .08 | 1.27 | .22 | 2.50 | 1.98 |
Glycated hemoglobin | 6.39 | .02 | −1.05 | .31 | −1.00 | 0.95 |
Both groups exhibited an increase in their composite dietary habits score in the postexposure results and a decrease in HbA1c (
First, a paired samples
In addition, the second analysis consists of two independent samples
Changes in both eating habits composite score and glycated hemoglobin.
Group | Before | After | |
|
|||
|
HbA1ca | 8.10 | 6.60 |
|
EHb | 23.3 | 24.4 |
|
|||
|
HbA1c | 7.10 | 6.90 |
|
EH | 25.9 | 26.10 |
aHbA1c: glycated hemoglobin.
bEH: eating habits.
Dependent samples
Group | Values | ||||
|
Mean (SD) | SE mean | Significance ( |
||
|
|||||
|
Pre-HbA1ca to post-HbA1c | 1.41 (1.75) | 0.58 | 2.42 (8) | .04 |
|
Pre-EHb to post-EH | −1.00 (2.74) | 0.91 | −1.09 (8) | .31 |
|
|||||
|
Pre-HbA1c to post-HbA1c | 0.23167 (0.62) | 0.18 | 1.30 (11) | .22 |
|
Pre-EH to post-EH | −0.25000 (4.31) | 1.24 | −0.20 (11) | .84 |
aHbA1c: glycated hemoglobin.
bEH: eating habits.
Between-group differences for both glycated hemoglobin and eating habits.
Group | Levene test | |||||
|
Significance ( |
Significance (two-tailed; |
Mean difference | SE difference | ||
Glycated hemoglobin difference | 12.67 | .002 | 1.94 (19) | .04 | 1.18 | 0.54 |
Eating habits difference | 3.30 | .09 | −0.456 (19) | .65 | −0.75 | 1.64 |
The quasi-experiment conducted for this study showed that there is a statistically significant difference within and between subjects in terms of their HbA1c. This provides strong preliminary evidence about the efficacy of using a nutrient profiling–based dietary app to present the nutritional value of different food recipes in a customized meal plan. Combined with the nutrition classes offered by DTC, interacting with
One might argue that the nutrition classes offered by DTC do have an effect on patients in lowering their HbA1c. However, we tried to counterbalance this potential effect by having a control group. Obtaining pretest measurements on both the intervention and control groups allows us to assess the initial comparability of the groups. The assumption is that if the intervention and the control groups are similar at pretest, there is a smaller likelihood of important confounding variables differing between the 2 groups [
Nutrition self-efficacy descriptive statistics.
Type | Values | ||
|
Mean (SD) | SE mean | |
|
|||
|
Control (n=12) | 15.25 (3.22) | 0.93 |
|
Intervention (n=9) | 14.67 (1.66) | 0.55 |
Nutritional self-efficacy between-group difference.
Type | Levene test | 95% CI of the difference | |||||
|
Significance ( |
Significance (two-tailed; |
Mean difference | SE difference |
|
||
Nutritional self-efficacy | 2.06 | .17 | 0.49 | .63 | 0.58 | 1.18 | −1.89 to 3.06 |
For patients with diabetes attending the DTC or any other diet management center, keeping up with a healthy nutrient-dense diet has always been a challenge. According to personnel from the DTC, most patients fall back into their old dietary habits as soon as they leave the center. By having a customized dietary app, such as
This study contributes to the body of knowledge on 2 broad levels: to society and to science. The first level of contribution is a new method for nutrition education. The
In addition, the steps of the algorithm can be viewed as a set of design principles. The algorithm can be tailored to tackle different health conditions. Both the nutrients and the criteria for each nutrient can be tailored according to the health condition being treated. For example, cardiovascular diseases (CVDs) have certain nutrition therapy recommendations that are slightly different than those for patients with diabetes. Diabetes gives priority to carbohydrate consumption, whereas CVDs give special attention to fat consumption.
The artifact,
Although the results of this study are promising, there are some limitations that allow room for further improvement and map out new directions for future research. The first set of limitations concerns the design artifact,
The second set of limitations concerns the Intelligent Nutrition Engine artifact. This algorithm is inspired by the novel concept of
The third set of limitation concerns the study design. In the third stage of this study, we conducted a quasi-experiment to evaluate the effect of using
In the third stage, participants in the intervention group exhibited significant changes in their HbA1c. Participants in the control group, however, exhibited minimal and nonstatistically significant changes in their HbA1c. In addition, postresults of the independent samples
American Diabetes Association
Application Programming Interface
basal metabolic rate
cardiovascular disease
design science research
diabetes treatment center
eating habits
glycated hemoglobin
Health-Promoting Lifestyle Profile
Institutional Review Board
nutrition self-efficacy
standard of care
type 2 diabetes mellitus
This Project was funded by the Deanship of Scientific Research (DSR) at King Abdelaziz University, Jeddah, under grant no G-276-612-1440. The authors, therefore, acknowledge with thanks DSR for technical and financial support. Samir Chatterjee was partly funded for a portion of this study by a Fellowship from Schoeller Research Center for Business and Society in Nuremberg, Germany.
None declared.