Applicability of a Web-based 24-hour Dietary Recall Tool for Japanese Populations in Large-scale Epidemiological Studies

Background Recent innovations in information and communication technology have made it possible to assess diet using web-based methods; however, their applicability in the general population remains unclear. Hence, we aimed to examine the applicability of a web-based 24-hour dietary recall (24HR) tool to large-scale epidemiological studies by determining the sampling rate and characteristics of randomly selected participants from a Japanese cohort study. Methods In total, 5,013 individuals were recruited from a cohort of 21,537 individuals, and 975 agreed to participate in this study. The participants selected either self-administered web-based dietary 24HR (self-administered 24HR) or interviewer-administered telephone-based 24HR (interviewer-administered 24HR) as the method for the dietary assessment and answered questions regarding the acceptability of the system. Results The response rate of the 975 participants was 19.4%, corresponding to approximately 4.5% of the total study sample. About half of them chose the self-administered 24HR (46.9%). The median time required for the self-administered and interviewer-administered 24HR was 25 and 27 minutes, respectively. In the self-administered 24HR, older people, regardless of sex, tended to require a longer time, and approximately 60% of the participants rated the ease of use of the system as “somewhat difficult” or “difficult.” Conclusion Characteristics of the participants in this study were not systemically different from those of the entire study sample. Improvements in the approach to entering cooking details and the dish name selection may be necessary for better acceptability in order to be accepted as a self-administered dietary recall tool.


Introduction 22
Information obtained through diet and nutrition assessment is the basis for establishing evidence-based policy 23 for preventing disease and improving public health. In large-scale epidemiological studies, food frequency 24 questionnaires (FFQs) are commonly used. 1 However, compared with other tools used to assess dietary intake, 25 FFQs tend to have large measurement errors. 2

26
Recent innovations in information and communication technology have made it possible in various countries 27 to assess diet using web-based 24-hour dietary recall (24HR) tools. [3][4][5][6][7][8][9][10][11][12][13] EPIC-Soft collects data using the 24HR 28 method and can be used to calibrate measurement errors by FFQs. 14 Web applications and software programs 29 enable efficient data collection and manipulation. 30 Japanese diets are complex and involve mixed dishes, and the variety of combinations makes it difficult to 31 investigate meals on a food-by-food basis. This may also cause relatively low validity of the FFQ among the 32 Japanese population. 15 To overcome the complexity and measurement errors of investigation of the Japanese 33 diet, a web-based automated food assessment tool that uses the 24HR method, the Automated Web-based 34 assessment System Using Recipe-Data for Japanese (AWARDJP), was developed. 35 This study aimed to investigate the estimated dietary intake by the AWARDJP for the calibration of the 36 measurement error of the estimated intake by the FFQ in this population. Towards this goal, we examined the 37 applicability of the AWARDJP in large-scale epidemiological studies by determining the sampling rate and 38 characteristics of randomly selected participants (partially entire population) in the Japan Public Health Center 39 Next Generation Prospective Study 16 (JPHC-NEXT) and other cohorts that use the same FFQ as JPHC-NEXT. 40 Furthermore, we clarified the differences in the acceptability of the AWARDJP by the participants that were due 41 to the differences in the response method used in the dietary assessment.   U test was used to evaluate the difference of time required between self-administered 24HR and  98 interviewer-administered 24HR by sex. 99 Additionally, the proportions of participants who answered each question of acceptability were calculated by 100 age group and response method. The χ² test was performed, and the p-value was calculated. Fisher's exact test 101 (direct method) was used when the number of cells with expected values less than 5 exceeded 20% of the total. 102 The significance level was 5% (two-sided test). SAS 9.4 (SAS Institute Inc., Cary, NC, USA) was used for all 103 data analyses. 104 105

106
In total, 975 participants from the 5,013 recruited individuals in this study (response rate: 19.4%), accounted 107 for approximately 4.5% of the total study population. The characteristics of participants who responded to the 108 AWARDJP and those of the study population were similar. These results indicated that the characteristics of the 109 participants represented those of the cohort population (Table 1). Of the 975 participants, 457 (46.9%) and 518 110 (53.1%) participants chose self-administered and interviewer-administered 24HR, respectively. There were 111 differences in the response method among the regional cohorts (Table 2). 112 Table 3 presents the duration of the dietary survey by method, sex, and age. Among men, the median (inter 113 quartile range) time required to complete the dietary survey was 20 (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) and 25 (21-31) minutes in the self-114 administered and interviewer-administered 24HR, respectively. For women, it was 30 (20-40) and 28 (23-35) 115 minutes, respectively. Furthermore, when aggregated by age, the response time of the self-administered 24HR 116 tended to be longer than that of the interviewer-administered 24HR as the participant age increased. In view of 117 the age group, we found a tendency of a broader range variation in self-administered 24HR. In the case of self-118 administered 24HR, there were participants whose assessment time >50 minutes for all age groups and sexes. 119 For interviewer-administered 24HR, an assessment time >50 minutes was rare across all age groups ( Figure 2). 120 A c c e p t e d V e r s i o n 7 percentage of those who answered "somewhat difficult" or "difficult" showed an increasing trend with age for 124 both men and women. Regarding the question about the difficult part to enter, the most common answer among 125 those in the self-administered 24HR was "dish name selection" (about 70%), followed by "entering cooking 126 details" (about 60%). However, there was almost no significant difference by sex (eTable 3-1 and 3-2). 127 128 Discussion 129 To evaluate the applicability of the AWARDJP in large-scale epidemiological studies, this study was 130 conducted with two aims: to examine the sampling rate and respondent characteristics of the participants 131 randomly sampled from the four selected study cohorts, and to clarify the difference in the acceptability among 132 the participants by the assessment method. Referring to the EPIC study, where the random sample size was 5-133 12% of the cohort population for FFQ calibration, this study included only 4.5% of the initial population. 134 However, the characteristics of the participants and those of the study population were quite consistent, 135 indicating the possibility of AWARDJP applicability to large-scale epidemiological studies for 24HR dietary 136 assessment and for FFQ calibration. To use it for correcting the measurement error in FFQs, further studies 137 considering the acceptance rate obtained in this study are required. 138 The median of time required to enter responses into the system was 20 and 25 minutes for men, and 30 and 28 139 minutes for women for the self-administered and interviewer-administered 24HR, respectively. These results 140 were consistent with those of the review study conducted by Eldrige 24 on the evaluation of dietary assessment 141 tools, where it was observed that the average time to complete an assessment ranged from at least 14 minutes to 142 45-60 minutes. This indicated a similarity with the AWARDJP system, from a data entry time perspective. 143 However, there were several differences between the interviewer-administered 24HR and the self-administered 144 24HR. In this study, especially in the case of self-administered 24HR survey, it is not possible to accurately 145 confirm whether the participant was interrupted in the middle or ended the survey prematurely which is the 146 limitation of this study. 147 Additionally, the percentage of those who answered "somewhat difficult" or "difficult" to the question A c c e p t e d V e r s i o n 8 in the self-administered 24HR. In the interviewer-administered 24HR, the interviewer entered the data for the 150 participants, shortening the data entry time, possibly avoiding some of the difficult processes affecting the 151 acceptability of the system. Based on the above, the evaluation of AWARDJP system acceptability may have 152 been biased by age group or sex. 153 The dish name selection and the entering cooking details (food and food volume) were the most difficult 154 points to answer according to the result of acceptability. A better approach for entering cooking details such as 155 the dish name selection option and search keywords may be necessary to improve the system so that the number 156 of web-responders would increase. 157 The result of the questionnaire on AWARDJP acceptability by the web-based self-administered 24HR showed 158 that many respondents answered that it was difficult to select a dish name from the food search interface and 159 enter the cooking details into the system. To use the web-based 24HR for large-scale epidemiological studies, it 160 is necessary to improve the food search interface for easy-to-use and support methods, considering that the web 161 environment is being used by users who are not accustomed to PCs. Furthermore, it will be necessary to examine 162 the accuracy and validity of both the response methods based on accurate indicators such as a biomarker 163   A c c e p t e d V e r s i o n  The percentage of "total" shown in bold is the percentage of the total number of participants in each cohort. Percentages by response method (self-administered / interviewer-administered) are expressed as percentages of the total number of participants who used each method.
a Saku area in the JPHC-NEXT, b Aichi Workers' Cohort Study, c Yuzawa Cohort Study, d Yamagata Cohort Study.
e Reminders are set to 0 times for those who responded to the initial survey request.
f Participants are given a choice to select one out of two days within a week to take the assessment.
Percentage of respondents on the requested day is calculated as follows: "number of participants who took the survey according to our request"/ "total number of participants" ×100.
For those who responded to our request for two days, the dietary survey on the first day was considered for the calculation.
g The seasons are set as follows. Spring from March to May, summer from June to August, autumn from September to November, and winter from December to February. Those who responded for 2 days or more were shown the season when they responded to the dietary survey on the first day.     The "total" percentage is expressed as a percentage of the total number of participants. The percentage of Self-administered / Interviewer-administered responses is expressed as the percentage of the total number of those who used each method. Missing values are excluded for each item. N/A: Not applicable -:No description because the question was answered by the interviewer. a Self-administered web-based dietary 24-h recall. b Interviewer-administered telephone-based 24-h recall. c χ2 test: Fisher's exact test (direct method) was used when the number of cells with expected values less than 5 exceeded 20% of the total. d The denominator is the number of subjects who selected "Somewhat difficult" or "Difficult" in the above question. e The denominator is the number of subjects who selected "Somewhat disagree" or "Disagree" in the above question.