We considered the sequences and processes from several studies in the literature involving software applications [11-13] to select, develop, and apply an assessment tool complying with the DIET@NET partnership guidelines [14]. We performed a cross-sectional study between August 2018 and December 2018 in public primary health centers in 4 small towns in the northern region of Brazil's Rio Grande do Sul, namely Estação, Erebango, Getúlio Vargas, and Ipiranga do Sul, which had 27,079 inhabitants in the last official census. This study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving research study participants were approved by the ethics committee of the University of Passo Fundo under opinion number 2660304. Written informed consent was obtained from all participants. According to national regulations, participants were not compensated.

For the sample calculation, we considered the total population of elderly patients who were diagnosed with T2DM by the public health control from the cities investigated and attended a primary care center (N=257), with an expectation of a dietary intake 50% in line with the Food and Nutrition Surveillance System (SISVAN) of Brazil guidelines, acceptable error of 5%, and a confidence level of 95%, resulting in a target sample size of 154 individuals. The inclusion criteria were 60 years of age or older; T2DM diagnosis, either by self-report or confirmed by examinations in the patient history (fasting glycemia or glycated hemoglobin [HbA_1c]); and cognitive ability confirmed by the Mini-Mental State Examination [15]. Through phone calls, we randomly called participants to attend the primary health center for an interview. We invited participants until we reached the required sample. All participants signed an informed consent form.

Initially, we performed a systematic mapping to investigate the different methods for assessing the dietary intake of adults and the elderly [5]. We observed that the main assessment method was a food frequency questionnaire. However, the questionnaires are usually applied manually. To optimize time and resources, we wanted to use an electronic method for collecting and processing the questionnaires.

In this study, we used the Dietary Intake Form proposed by SISVAN [16], which is an initiative from the Brazilian Ministry of Health. The form is a food frequency questionnaire with a recall period of the previous 7 days and is comprised of 10 categories, which relate to a healthy diet (eg, daily intake of beans, fresh fruits, vegetables, and milk) or to nonrecommended practices (eg, frequent intake of fried foods, processed meat products, cookies, sweets, and soft drinks). SISVAN guidelines recommend an intake of 7 serves per week of raw salad, pulses and cooked vegetables, fresh fruit, and milk or yogurt; 5 or more serves per week of beans; and 0-1 serves per week of fried foods, processed meat products, savory biscuits, cookies or sweets, and sugary drinks.

Previously, a mobile app addressing SISVAN had been developed for use with inpatients [7]. Hence, we developed a new version, with improved functionalities adapted to primary care outpatients, called Diabetes Food Control 2. The app now includes questions on demographic and economic characteristics (eg, marital status, self-reported skin color, education, income, and occupation), clinical and nutritional history, time from diagnosis, family history of diabetes, medication intake, presence of comorbidities, nutritional monitoring, lifestyle, physical activity, smoking, and alcohol consumption. The Brazilian Institute of Industrial Property granted us a software registration certificate under number BR5120190010540.

The app (Figure 1) was designed so that health care professionals could use it in primary health care centers to assess the dietary intake and health status of outpatients with T2DM, addressing the recommendations of Brazil's Ministry of Health [16]. Using the app, we intended to optimize data collection and make the usage of apps a more consistent practice in primary care settings. The app's report provides straightforward information on whether the patient meets the SISVAN criteria related to food intake and anthropometric measurements. It also provides personalized feedback for patients in the form of a report they can receive by email.

We collected all data with the SISVAN questionnaire using the Diabetes Food Control 2 app. In the public primary health centers selected for the study, the same dietitian researcher interviewed each participant to simultaneously fill out the questionnaire on the app. The dietitian also collected demographic and economic data such as marital status, self-reported skin color, level of education, income, and occupation. The participants followed the use of the app during the entire interview and interacted by filling out their demographic information.

To evaluate nutritional status, we analyzed body weight (kg), height (cm), and waist and hip circumferences (cm). For stature measurement, we used a 200-cm stadiometer with a scale of 0.2 centimeters. To assess weight, we used a Welmy digital scale with a capacity of 150 kg. We took all measurements according to the recommendations of the Brazilian Ministry of Health [17]. Using the measured height and weight, the app calculated the BMI and the classification of nutritional status according to the guidelines of the Pan American Health Organization [18].

The dietitian assessed waist circumference with a measuring tape at the smallest curvature between the ribs and the iliac crest, without compressing the skin. For waist circumference, the categories were low risk (men <94 cm; women <80 cm), high risk (men ≥94 cm; women ≥80 cm), and very high risk (men ≥102 cm; women ≥88 cm). We measured hip circumference in centimeters in a horizontal plane at the region of the greatest gluteal protuberance. We calculated the waist-to-hip ratio according to the cutoff points recommended by the World Health Organization [19]. For waist-to-hip ratio, the categories were low risk (men <0.9; women <0.8), moderate risk (men 0.9-0.95; women 0.8-0.85), and high risk (men >0.95; women >0.85).

To assess glycemia, we asked the primary health center in each town to conduct a HbA_1c test. All blood tests were collected and evaluated by trusted outsourced laboratories that were contracted by the primary health center. All handling procedures were carried out in accordance with national quality and safety standards and under the supervision of the researchers. A value of 8.0% was the standard reference value, as recommended by the American Diabetes Association for individuals aged over 60 years [20]. We also assessed fasting capillary glycemia, which a nursing professional verified at the time of the questionnaire application using an Accu-Check Active glucometer. In this case, 70-110 mg/dL was the standard reference value [20].

During the interview, the dietitian also investigated the clinical and nutritional history of participants, time from T2DM diagnosis, family history of diabetes, and medication intake. The dietitian registered information on the presence of comorbidities, such as hypertension and dyslipidemia, and whether the participant had received any nutritional guidance, engaged in physical activities routinely, smoked, and consumed alcohol. At the end of the interview, the participant received nutritional instructions for better glycemic control and a report generated by the app showing his or her dietary and health status according to the SISVAN guidelines.

We also assessed the mobile app with 20 health professionals, including 4 medical doctors, 8 registered nurses, and 8 community health agents (registered nurses and licensed practical nurse) who worked in the primary health centers. Individually, we presented the app and described its functionalities and form of use. We let each professional test and interact with the app and fill out the questionnaire once with dummy data, which we discarded later. After the professionals tested the app, we applied a questionnaire to assess usability. To assess usability, we used an adaptation of the System Usability Scale [21]. To analyze acceptance, we used a questionnaire from the Technology Acceptance Model (TAM) [22]. Last, we asked the health professionals the following 3 descriptive questions: Have you ever used a mobile app in clinical practice? How was your experience with using the Diabetes Food Control 2? What changes do you suggest for this app to make it more useful or applicable to clinical practice? The opinions of professionals from the primary health centers would allow us to understand the potential for implementing the app in clinical practice.

Participants’ dietary intakes were summarized and compared by all the demographic, nutritional, biochemical, and lifestyle variables. Intake was classified according to whether it was in line with SISVAN guidelines [16]. We used SPSS 22.0 statistical software (IBM Corp, Armonk, NY) to analyze the quantitative data. Basic quantitative data are described as mean, SD, and median. Categorical data are described by simple frequencies. The associations between dietary intake and demographic, nutritional, biochemical, and lifestyle variables were tested using chi-square tests, considering a 5% significance level for all analyses. For the qualitative data, we assessed the data in groups, following the model proposed by Minayo [23].

Table 1 presents the demographic characteristics. Most of the participants were women, married, with self-reported white skin color, residents of an urban area, with <4 years of education, retired, with family income between 4 and 10 minimum wages, and presenting with a family history of diabetes.

Table 2 presents the data on clinical characterization and lifestyle. The majority reported using only oral hypoglycemic agents for T2DM treatment, having dyslipidemia and systemic arterial hypertension, and using medications for hypertension. Most respondents also reported having received some nutritional guidance but not performing nutritional monitoring. Most participants reported having never smoked, not consuming alcohol, and not performing physical activity.

Table 3 presents the characterization of nutritional status. We identified that most participants were obese and had a waist circumference indicating a very high risk of metabolic disorders, a waist-to-hip ratio categorized as high risk, altered capillary glycemia, and altered glycated hemoglobin. As for comorbidities, we identified that 51.9% (80/154) of respondents reported none of the diseases investigated, 31.2% (48/154) with coronary disease, 16.9% (26/154) with diabetic retinopathy, 16.9% (16/154) with depression, 5.8% (9/154) with kidney disease, 1.3% (2/154) with diabetic neuropathy, and 1.3% (2/154) had received a diagnosis of diabetic foot.

Among the participants, 82.5% (127/154) ate breakfast, 50% (77/154) ate a morning snack, 100% (154/154) ate lunch, 70.1% (108/154) ate dinner, and only 9.1% (14/154) ate supper. Most of them (139/154, 90.3%) described a usual diet in the week prior to the study, 85.1% (131/154) reported not adding salt in prepared food, and 52.6% (81/154) stated using vegetable oil or butter. Most participants had an intake of raw salads, fruits, milk products, fried foods, savory biscuits, cookies or sweets, and sugary drinks in line with recommended guidelines. Meanwhile, the intake of beans, pulses and cooked vegetables, and processed meat products were not in line with recommended guidelines. Figure 2 summarizes the food frequency data for the number of serves per week and the classification according to the SISVAN guidelines. Absolute values are available in Multimedia Appendix 1.

We compared the association between dietary intake with the demographic variables (Multimedia Appendix 2) and with lifestyle, nutritional, and biochemical variables (Multimedia Appendix 3). Statistically, we identified that women had healthier behaviors regarding the intake of fried foods (χ²₁=7.963, P=.006) and sugary drinks (χ²₁=6.127, P=.035). As for the place of residence, respondents living in a rural area had a more appropriate intake of cookies and sweets than residents living in an urban area (χ²₁=5.035, P=.034). Participants without a family history of T2DM had a more appropriate intake of beans (χ²₁=15.170, P<.001). We detected a more appropriate fruit intake in participants who engaged in physical activity (χ²₁=8.955, P=.003). We also found that the lower the rate of smoking, the more adequate the intake of raw salad (χ²₁=13.034, P=.001). Participants with coronary disease reported a more appropriate intake of pulses and cooked vegetables (χ²₁=4.223, P=.050).

The usability assessment with the System Usability Scale generates a score that should be ≥68 points to determine satisfactory results. The assessment of the Diabetes Food Control 2 by health professionals resulted in a mean of 71.75 points, confirming satisfactory usability. Figure 3 presents the responses from health professionals to the TAM questionnaire [22]. The health professionals also answered 3 descriptive questions, depicted in Textbox 1.

This study assessed various aspects of the nutrition of elderly patients diagnosed with T2DM in the context of primary health in line with SISVAN guidelines, which is a national parameter in Brazil. For data collection, we developed and evaluated a new app based on recommendations from medical associations and literature studies. Interesting insights are presented about the use of the app and the health professionals’ perceptions about its features and the possibility of using it in the context of primary health.

On nutritional assessments, the profile found was similar to surveys performed in the same circumstances involving adult [24] and elderly [25] patients with T2DM. The anthropometric assessment identified that 73.3% of the study population was overweight, which is a rate higher than that presented in similar studies. This stresses the need for greater attention to elderly patients with T2DM assisted in basic health care, especially due to the evident metabolic syndrome and the worsened transition from adulthood to the elderly.

In this study, we identified that men had a greater intake of fried foods than women with a similar dietary behavior, as in the study by Linde et al [26], in which men with higher BMIs consumed more fried foods than women. Similar findings were also reported by Cahill et al [27] and Sayon-Orea et al [28]. Regarding the intake of sugary drinks, we identified that women had a more appropriate intake than men, corroborating other studies [29,30]. As for the place of residence, respondents living in a rural area had a more appropriate intake of cookies and sweets than residents in an urban area, which was similar to the findings of Farrell et al [31] regarding foods and beverages with high sugar content.

The literature has already shown that people who engage in physical activity have better dietary habits [32]. In our study, respondents who reported engaging in physical activity had a more appropriate fruit intake. These data corroborate the studies by Tan et al [33] and Alakaam and Lemacks [34], which identified a positive relationship between physical activity and the intake of fruits and vegetables. We also identified an inversely proportional association between salad intake and smoking. Sebastian et al [35] had similar findings, in which salad intake seemed more prevalent among former smokers and people who had never smoked. Besides, the intake of pulses was significantly higher in participants with cardiovascular disease. Other studies have also associated the intake of fruits and vegetables with a lower prevalence of coronary disease and the intake of fiber with a lower prevalence of acute myocardial infarction [36,37]. Interestingly, our findings also suggest that respondents without a family history of diabetes had a more appropriate intake of beans. Some prior studies discuss the relationship between T2DM and the intake of beans. For instance, Ley et al [38] affirmed that a diet rich in vegetables, legumes, and unrefined grains was associated with a lower risk of T2DM, and for diagnosed patients, the intake of such foods improves glycemic levels and lipid control.

The app we developed has several benefits emphasized by health professionals. Other researchers who used mobile apps in dietary intake assessments cited similar advantages. Rangan et al [39] mentioned a reduction in data entry and coding workload for researchers. Meanwhile, Bucher Della Torre et al [40] reaffirmed that electronic assessment helps to prevent some pitfalls of paper-based food records, such as errors in transcription and difficulties in reading what participants have written. There is a growing interest in using smartphones to collect data on dietary assessment [6], considering mobile technology offers a wide range of suitable options to achieve the benefits that technology can deliver [41-43].

Among the 20 respondents, none had used apps in their professional practice. This suggests the need for implementing the use of technology in public health, considering the potential help and benefits provided to professionals. Health professionals also reported that, by using the app, they could thoroughly monitor patients in the long term, considering that the questionnaire is easy to apply and contains all the questions needed for proper monitoring in the context of primary health. Community health agents, who are professionals who visit patients’ homes, mentioned that the app could be used by them to counsel the general population regarding nutrition. These professionals receive little prior training on nutrition, and the app would provide a scientific basis so they could provide such care to patients.

The physiological aspect of aging was a limitation of the present study. The potential memory loss of the respondents may have affected their answers to retrospective dietary intake. The length of the questionnaire may have also negatively affected the disposition of participants to respond. Another limitation was the impossibility of measuring the HbA_1c of all respondents in the laboratory, due to the lack of financial resources. The app was only used one time; therefore, the long-term effectiveness and usefulness should be studied further.

The dietary intake of the assessed participants is partly appropriate. However, this population was substantially overweight and presented with metabolic syndrome and poor results in the anthropometric and biochemical assessments. Moreover, there was a low intake of pulses and cooked vegetables and beans and a high intake of processed meat products. Mostly, the findings of this study are similar to others in the literature, which reinforces a concerning reality of the health conditions of elderly patients with T2DM. These findings can guide dietary interventions and health education in similar settings. In addition, this study adds to the health informatics literature that a new app can be used in the context of public health in Brazil. The mobile app was useful in the study and well-rated by health professionals in primary health care settings. Its features are encouraging tools for use in future studies. In this study, the app was developed to completely focus on T2DM in primary care settings in Brazil. For application in other settings or for other health conditions, the development method could be adapted or used in a similar way, but considering the specificity of other chronic diseases and according to the specific guidelines available. Future studies can use this study as an example of a location and features where technology is satisfactorily accepted by health care professionals in the first introduction initiative.