2025 Volume 7 Issue 11 Pages 1109-1115
Background: We developed the SaFIS, a structured self-administered food intake survey for patients with cardiovascular diseases needing dietary improvements.
Methods and Results: The SaFIS comprised questions on 31 items, including staple foods (7 items) and food ingredients (24 items), based on the foods recommended by the Japanese Atherosclerosis Society for the Prevention of Arteriosclerotic Disease. To evaluate reproducibility, SaFIS was administered twice with a 1-month interval using the intraclass correlation coefficient (ICC [1,1]). Validity was assessed by comparing energy and nutrient content from a weighed dietary record (WDR) with the first SaFIS survey using Bland-Altman plots. Reproducibility values were energy (ICC 0.90, 95% confidence interval: 0.80–0.95), protein (0.90, 0.80–0.95), fat (0.76, 0.54–0.88), saturated fatty acids (0.78, 0.58–0.89), monounsaturated fatty acids (0.76, 0.56–0.88), polyunsaturated fatty acids (0.82, 0.66–0.91), carbohydrates (0.89, 0.79–0.95), and dietary fiber (0.86, 0.71–0.93). Bland-Altman values indicated SaFIS-based energy, protein, carbohydrate, monounsaturated fatty acids, and dietary fiber were 67.2 kJ, 5.7 g, 3.9 g, 0.5 g, and 1.8 g greater than WDR-based values, respectively. SaFIS-based fat, saturated, and polyunsaturated fatty acid values were 0.7 g, 1.3 g, and 0.39 g less than WDR-based values, respectively.
Conclusions: SaFIS demonstrated high reproducibility and validity with clinical potential. Subsequent introduction to digital health from the usability perspective is planned.
Lifestyle modifications using digital technology has been adopted as a common approach in modern health care. Among these, dietary modification is one of the core components, as well as enhancing exercise habits, and has been indicated for a wide range of subjects, including patients with cardiovascular disease (CVD) or at risk of CVD. Guidelines for treating patients with CVD recommend assessing their initial energy and nutrient intake when individualized intervention is indicated.1,2 Increased risk of incident atherosclerosis is associated with higher levels of cumulative non-high-density lipoprotein cholesterol, which reflects long-term exposure to atherogenic lipoproteins, and with elevated levels of low-density lipoprotein cholesterol, a widely used clinical indicator.3,4 Therefore, assessing dietary intake of saturated fatty acids, which are known to elevate these lipid markers, is critically important.
Various dietary assessment methods for assessing dietary habits and the nutritional composition of the diet are available, such as food records, 24-h dietary recall, and the food frequency questionnaire (FFQ).5 Food records and 24-h dietary recalls are standard approaches to assessing the actual diet consumed, but have been reported to be burdensome for patients and healthcare providers and dependent on patients’ culinary knowledge and memory.5,6 The FFQ is primarily used in epidemiological studies and does not include detailed dietary records, thus providing limited information for personalized counseling.5 Analysis of energy and nutrients from meal photographs is a common approach in digital health. However, the method has been validated using instant diets at hospital with a small number of participants.7 Therefore, from the perspective of generalizability, it is difficult to apply this method to individuals with illnesses. To establish personalized nutritional interventions for patients with CVD or other non-communicable diseases and their precursor groups, a survey should provide the energy and nutrients ingested in detail as validated data for a given candidate.
Therefore, we devised a structured self-administered dietary intake survey (SaFIS) that can be converted into an application for patients with CVD and other individuals with diseases who require dietary improvements. The SaFIS was developed according to the following requirements: (1) it can evaluate nutrient and energy intake targets for personalized interventions aimed at reducing CVD risk factors, (2) it can be evaluated repeatedly with an appropriate number of answer items, and (3) photographs of the measuring unit can be grasped. We also examined the reproducibility and validity of the SaFIS.
Concept of the SaFIS The SaFIS is a self-administered nutrition survey designed to replace the FFQ and 24-h dietary recall surveys, aiming to assess dietary patterns that may influence the onset or recurrence of CVD and its risk factors, including obesity and diabetes mellitus. The questionnaire comprised 31 items, including questions assessing staple foods (7 items) and food ingredients (24 items), to evaluate the consumption of nutrients that influence CVD progression according to the Japan Atherosclerosis Society (JAS) Guidelines for Prevention of Atherosclerotic Cardiovascular Diseases 2017.8 To enhance the compatibility of the SaFIS with digital health platforms, the system was adjusted such that: (1) nutrients and energy intake targets for CVD correction could be evaluated, (2) the number of answer items was reduced so that evaluations could be repeated in a short period, and (3) the measurement unit could be grasped from a photograph.
Questionnaire Items of the SaFIS The 7 staple food items included rice, bread, noodles, and others. The other 24 food items included fatty or lean beef; pork; chicken; blue fish; soy products; green, yellow, and white vegetables; and fruits, categorized as foods to be actively consumed and foods to be avoided. The detailed food list is presented in the Supplementary Table.
The intake of each ingredient was estimated according to a predetermined unit; for example, the size/number of bowls for rice and the number of pieces for bread. Non-staple foods, butter, margarine, fats, and oils, were evaluated using spoon measures (teaspoons and tablespoons) for portion estimation. The ingredients in milk, yogurt, and soymilk were estimated using cups of a constant size. Other ingredients were evaluated using 3 plate sizes (small, medium, and large, approximately 10, 15, and 20 cm in diameter, respectively). The weight of the ingested food was calculated according to the predetermined weight of each bowl; for example, 40, 80, and 160 g for small, medium, and large plates, respectively. A similar estimation was made for the weights measured in cups or pieces.
Furthermore, questions regarding frying, stir-frying, baking, simmering, and steaming were included to estimate the differences in energy content and each nutrient based on the cooking methods. Questions regarding the cooking methods used for chicken skin and fatty meat were further included to estimate the intake of saturated fatty acids more accurately. The respondents were advised to take photos of the meals they consumed and to refer to them when completing the SaFIS to reduce recall bias. Using a photo of a meal, even first-time users could complete the SaFIS assessment within 5 min.
Energy and Nutrient Calculations Built Into the SaFIS The questionnaire has a built-in database providing numerical values for the nutrients in various foods, based on the Standard Tables of Food Composition in Japan (published by the Ministry of Education, Culture, Sports, Science and Technology).9 SaFIS can calculate the energy, macronutrients, saturated fatty acids, monounsaturated and polyunsaturated fatty acids, and dietary fiber, which are recommended for evaluation in the guidelines. Furthermore, other nutrient databases can be added as needed.9
The energy and nutrient intakes for each of the food groups shown in the SaFIS were calculated from the average of the representative food items among the multiple items presented in each question. Further, energy and nutrient values were established for each cooking method to account for variations.
Study DesignThe participants included 28 healthy adults (median age: 54 years [interquartile range (IQR) 20–65]; median body mass index: 20.9, IQR, 19.4–22.4 kg/m2; 9 men, 19 women). Registered dietitians in the researchers’ private research network recruited participants from among their families/friends.
The study design is illustrated in Figure 1. The registered dietitian created 10 different menus and prepared the dishes using the weighed dietary records (WDR)10 to avoid generating differences in cooking content and ingredients between the first and second surveys. The median values of energy and each of the constituent nutrients in the 10 menus were 2,234 kJ of energy, 22.8 g of protein, 18.0 g of fat, 4.1 g of saturated fatty acids, 7.1 g of monounsaturated fatty acids, 4.6 g of polyunsaturated fatty acids, 74.7 g of carbohydrates, and 5.6 g of dietary fiber.
Design of the SaFIS reproducibility and validity study. (SaFIS1 and SaFIS2, first and second SaFIS assessments.) Survey 1, After photographing the cooked meal, participants ingested the meal and filled it in SaFIS while reviewing the photos in the evening. Survey 2, 1 month after the first survey, a second SaFIS survey was conducted using the same one-meal menu. SaFIS, self-administered food intake survey.
The participants were randomly assigned to receive one of 10 different one-meal menus, and meals were prepared for the participants at home by one of the 13 registered dietitians in charge of cooking who collaborated with the researchers (21 of 28 participants were recruited from among their family members). After photographing the cooked meal, participants ingested the meal, and filled in the relevant information in SaFIS while reviewing the photos in the evening. The validity was verified by comparing the true values with the values calculated from the results of the first survey. At 1 month after the first survey, a second SaFIS survey was conducted using the same one-meal menu. Reproducibility was verified using values calculated from both surveys. When the cooking staff changed the amount of spices or ingredients on the menu, this was reported to the researchers.
After verifying the validity and reproducibility of the SaFIS results, we conducted a dietary assessment in 20 patients with atherosclerotic CVD (ASCVD) or risk factors for CVD, to evaluate the responsiveness of SaFIS to specific trends in nutrient intake.
Statistical AnalysisContinuous nonparametric data are expressed as medians with IQR. To assess reproducibility, the intraclass correlation coefficient (ICC [1,1]) was applied to evaluate the consistency of energy and nutrient measurements between the first and second SaFIS surveys.
Bland-Altman analysis plotting was performed to examine the agreement between energy and nutrient values calculated using SaFIS and WDR data, to ensure the validity of the SaFIS method.
The intake of protein, fat, saturated fatty acids, carbohydrates, and dietary fiber by patients with ASCVD or CVD risk factors was assessed based on the JAS Guidelines for prevention of atherosclerotic cardiovascular diseases 2022.11 Intake within the target range was classified as adequate, whereas intake outside this range was categorized as inadequate. Descriptive statistics were performed based on the classification results. Statistical analysis was performed using EZR software (Jichi Medical University, Saitama, Japan).12
Ethics ApprovalThis research was approved by the Ethics Committee of Chubu Rosai Hospital (approval number: 202108–01). The survey participants were informed of the purpose of the study and gave consent by signing an informed consent form. The study was performed in accordance with the Declaration of Helsinki.
The Table summarizes the energy content and nutrient composition calculated by the WDR (reference values) and SaFIS (first and second surveys). The ICCs for each indicator in the first and second SaFIS are as follows: energy (ICC 0.90, 95% confidence interval, 0.80–0.95), protein (0.90, 0.80–0.95); fat (0.76, 0.54–0.88); saturated fatty acids (0.78, 0.58–0.89); monounsaturated fatty acids (0.76, 0.56–0.88); polyunsaturated fatty acids (0.82, 0.66–0.91); carbohydrates (0.89, 0.79–0.95); and dietary fiber (0.86, 0.71–0.93).
Reproducibility of Energy and Nutrient Intake of the Food Intake Survey
| WDR (reference values) |
Survey 1 | Survey 2 | ICC (95% CI) | |
|---|---|---|---|---|
| Energy (kJ) | 2,234 (2,159–2,550) | 2,275 (2,057–2,486) | 2,319 (2,157–2,465) | 0.90 (0.80–0.95) |
| Protein (g) | 22.8 (19.4–27.2) | 27.9 (25.0–33.5) | 28.7 (25.3–31.0) | 0.90 (0.80–0.95) |
| Fat (g) | 18.0 (14.7–18.3) | 15.6 (13.4–18.9) | 17.8 (14.2–20.3) | 0.76 (0.54–0.88) |
| Saturated fatty acid (g) | 4.1 (3.5–4.6) | 2.6 (1.9–3.4) | 2.6 (1.8–3.6) | 0.78 (0.58–0.89) |
| Monounsaturated fatty acid (g) | 7.1 (6.5–7.4) | 7.7 (5.4–8.9) | 8.5 (5.8–9.7) | 0.76 (0.56–0.88) |
| Polyunsaturated fatty acid (g) | 4.6 (3.8–5.4) | 4.0 (3.4–5.0) | 4.0 (3.6–4.7) | 0.82 (0.66–0.91) |
| Carbohydrate (g) | 74.7 (73.6–79.4) | 75.1 (71–83.8) | 73.6 (70.9–82.1) | 0.89 (0.79–0.95) |
| Dietary fiber (g) | 5.6 (4.3–6.6) | 6.8 (5.6–8.7) | 6.9 (5.8–8.5) | 0.86 (0.71–0.93) |
CI, confidence interval; ICC, intraclass correlation coefficient; WDR, weighed dietary records.
Figure 2 shows the Bland-Altman plot for energy and the 3 major nutrients. The energy content was 67.2 kJ higher on average for the SaFIS than for WDR. Similarly, protein and carbohydrate values calculated by the SaFIS were higher on average than those calculated by the WDR (by 5.7 g and 3.9 g, respectively). Fat values calculated by the SaFIS were lower on average than those calculated by the WDR by 0.7 g. The overall agreement for energy and the 3 major nutrients was high, with a few extreme outliers outside the 95th percentile.
Bland-Altman plots of the difference between the energy and 3 macronutrients calculated by the food intake survey vs. WDR. The solid line denotes the average difference between the 2 methods: the smaller the difference, the smaller the systematic bias. The dotted lines denote the 95% limits of agreement representing the range in which most differences are expected to fall: the smaller the range, the smaller the difference between the two methods for most individuals. SaFIS, self-administered food intake survey; WDR, weighed dietary records.
Figure 3 shows the Bland-Altman plot for saturated fatty acids, monounsaturated fatty acids, polyunsaturated fatty acids, and dietary fiber. Saturated and polyunsaturated fatty acids showed lower average values by 1.3 g and 0.39 g, respectively. Conversely, monounsaturated fatty acids and dietary fiber exhibited higher average values by 0.5 g and 1.6 g, respectively. The agreement for saturated fatty acids, monounsaturated fatty acids, polyunsaturated fatty acids, and dietary fiber was also high, with a few extreme outliers outside the 95th percentile.
Bland-Altman plots of the difference between each nutrient calculated by the SaFIS and WDR. The solid line denotes the average difference between the 2 methods: the smaller the difference, the smaller the systematic bias. The dotted lines denote the 95% limits of agreement representing the range in which most differences are expected to fall: the smaller the range, the smaller the difference between the two methods for most individuals. SaFIS, self-administered food intake survey; WDR, weighed dietary records.
As shown in Figure 4, dietary assessment of 20 patients with ASCVD or risk factors for CVD revealed a tendency toward excessive intake of lipids and saturated fatty acids, as well as insufficient intake of dietary fiber.
Ratio of adequate to inadequate intake for each nutrient (n=20). The adequacy of nutrient intake was assessed based on the following target values: protein (15–20% of total energy), fat (20–25% of total energy), saturated fatty acids (<7% of total energy), carbohydrate (50–60% of total energy), and dietary fiber (≥20 g/day).
In this study, we developed the SaFIS for CVD prevention and disease management with an assuming application view and examined its reproducibility and validity. The results of our analysis using ICC and Bland-Altman plots showed good reproducibility and validity for energy and each nutrient. Our findings indicated that the SaFIS is applicable as a clinical dietary assessment tool for use with patients with CVD or CVD risk factors.
The SaFIS was highly reproducible for energy and nutrients. The reproducibility was superior to that reported in previous studies using 24-h dietary recall and FFQ.13 To complete the task outlined in SaFIS, the participants referred to actual photographs when responding to questions on each food item; for example, food arranged on small, medium, and large plates. The participants were also recommended to respond with reference to photographs of the foods they consumed. The most significant feature of the SaFIS is that it is a self-administered survey of food intake, rather than of a meal menu, as it is difficult to collect information on the target ingredients from meal menus. In addition, to be able to answer the SaFIS in a short time, we narrowed down the number of food items by asking summary questions about similar foods from a nutritional point of view. Specifically, it consisted of 31 items: staple foods (7 items) and other food (24 items). With regard to recall bias, the number of response items is an important factor for reproducibility, and in fact, reproducibility tends to be low for surveys with more than 100 food selections.14,15 However, the optimal number of responses for dietary evaluation has not been established and needs to be examined further.
This study showed high validity of the energy and nutrient values calculated from the food consumed. Among the 3 surveys described above, variation was observed for the FFQ depending on the nutrient, leading to relatively low validity.13 The number of food items in the FFQ is another factor that contributes to low validity, with several studies having more than 100 items.16–18 As the number of food items increases, the time and cost increase and the accuracy of the collected data is reduced.19 In this study, we confirmed the intake status of food items in the same way as in the FFQ, but the number of food items was carefully selected to a total of 31 items, including staple foods (7 items) and ingredients (24 items). Compared with the 24-h dietary recall and food record methods commonly used in clinical practice, SaFIS offers several advantages: (1) it is simple and time-efficient; (2) it does not require an interviewer; (3) it provides reference standards to guide dietary intake responses; (4) it allows for more accurate nutritional assessment by accounting for cooking methods; and (5) it enables repeated assessments over time. In addition, we created detailed answer branches for questions, such as the amount of chicken skin and fat consumed, cut and type of meat, and cooking method. These answer branches were developed to allow detailed assessment of protein, saturated fatty acid, and unsaturated fatty acid intake. For high-fiber foods, such as staple foods, vegetables, and fruits, response branches were created so that the amount could be measured using familiar tableware, such as a bowl of tea or a small bowl. We speculate that these well-crafted schemes may have contributed to increased accuracy and, in turn, improved the validity.
For validity, recall bias is another factor to be considered. In a validation study using the 24-h recall method, large interindividual variation was indicated.13 To counteract this recall bias, the use of food photographs when responding has been reported to improve the validity in the 24 h-camera method.20 In previous studies, the correlation coefficients between the WDR and 24-h recall method for energy, protein, fat, and carbohydrate were 0.774, 0.855, 0.769, and 0.763, respectively. The correlation coefficients were high, exceeding 0.75.20,21 Similar to their study, the participants in the present study were asked to answer the questions on the same day they consumed the food while looking at a photo of the food, which may have reduced the recall bias.
Being able to reliably assess diet is essential in the era of digital health; however, a simple, reliable, and validated assessment method has not yet been developed. For example, the assessment method using food photos, which is widely used, has the advantage that the measurement is low-burden, as it only requires taking a photo. However, its validity has been tested only with hospital and instant meals,7 and its generalizability is low. The calculation of energy and each nutrient from photographs requires the accumulation of a database of dishes and evaluation of 3D images.22 This method has also been reported to underestimate the amount of energy by 846 kJ compared with the actual amount,23 making it difficult to apply it to participants who require accurate assessments, such as patients with medical conditions.
In this study, we verified the reliability and validity of the SaFIS against the WDR, and a high degree of generalizability was demonstrated, indicating that the SaFIS is fully applicable for nutritional interventions with a clinical application. By automating the relationship between foods and nutrients within the same system, the SaFIS can be also developed into an application that allows patients themselves to recognize the nutrients they have ingested, which could make a significant contribution to dietary interventions not only in terms of quality but also convenience. As the SaFIS is designed to be mainly applied within digital health systems, we plan to incorporate it in the future. In addition, we plan to accumulate clinical data on the use of the SaFIS for nutritional assessment and interventions for patients with CVD. No age- or sex-related effects were observed among participants with extreme outliers, and the discrepancy was likely influenced by the minimum intake unit available in the application interface.
Study LimitationsFirst, the dietary assessment assumes people in their 40s to early 80s with preserved cognitive function and heart disease or risk factors, but the study included 30% of people in their 20s. The inclusion of a younger participant group may not correctly reflect the assumed target population, such as the older population in whom heart disease is more prevalent. Recall bias with the SaFIS may vary by age; however, we believe that the effect of recall bias was low because responses on the day of intake were made in conjunction with photographs of the foods. Second, the high proportion of young women in this study is likely to be a confounding factor affecting the generalization of the results. The assumed target population is predominantly male,24 and the assumed female target population is older. In dietary surveys to evaluate food ingredients, women may be more likely to precisely answer the questions than men because of greater familiarity with food ingredients.25 To the best of our knowledge, this effect of sex on dietary assessment has not been reported to date and is a subject for future study. Finally, this study evaluated food ingredients in a single meal. Compared with the 24-h recall method, the higher validity of responses related to only 1 meal in this study may have reduced the recall bias. However, mechanisms to reduce various response biases were incorporated in the SaFIS, such as reference meal pictures and response branches; hence, the number of meals provided is unlikely to have an effect. We also recognize that other unknown confounding factors may exist. Nevertheless, the SaFIS can be used as a clinical tool for dietary assessment in patients with CVD and risk factors.
In conclusion, the SaFIS demonstrated high reproducibility and validity, indicating wide scope for generalization and the possibility of it being used as a precise dietary assessment tool for patients with CVD and associated risk factors.
The authors thank the participants and registered dietitians for their collaboration with this study.
This study was supported by Japan Agency for Medical Research and Development, the grant number JP21le0110020.
The authors have no conflicts of interest related this research to declare.
This research was approved by the Ethics Committee of Chubu Rosai Hospital (approval number: 202108–01).
This study is not available for secondary use.
Please find supplementary file(s);
https://doi.org/10.1253/circrep.CR-25-0032
