Previous - Next
This is the old United Nations University website. Visit the new site at http://unu.edu
Evaluation of the 24-hour individual recall method in China
Fengying Zhai, Xuguang Guo, Barry M. Popkin, Linmao Ma, Qing Wang, Wentao Yu Shuigao Jin and Keyou Ge
We compared two dietary survey methods using the 1991 nationwide China Health and Nutrition Survey. Data were collected over three consecutive days by 24-hour dietary recall and a household inventory for 3,563 households with 13,606 individuals. We studied the absolute difference between the two methods and the relative differences expressed as dietary in take per capita per day. There was only a 74-kcal difference between the methods for average daily calorie intake; the relative difference was 1%. Ratios were larger for average daily protein (5%) and fat (3%) intakes. Analysis of covariance was used to compare the means of the intakes of nutrients when adjusting for other confounding variables. The largest difference was in households with guests eating at home, where fat intake was significantly higher. Particularly important was an adjustment for household cooking oil consumption used to modify the recall results. Removing this adjustment greatly expanded differences in the two methods.
In China, the standard approach to collecting accurate dietary data was the three-day household inventory to measure household food consumption and nutrient intake. Measuring individual food consumption was regarded as less accurate and was thought to be unacceptable before the 1980s. A multiple-day household inventory method is time-consuming and expensive, as it requires a high-quality, durable scale and trained field workers. Moreover, the method provides only nutrient intake and food consumption data at the household level. It neither differentiates the intake of guests from usual household members nor measures food eaten away from home (AFH). For these reasons, it is very difficult to assess nutrition status and conduct a range of studies at the individual level with the inventory method. Nevertheless, nutritionists in China and elsewhere in Asia believe the increased cost and effort are justified by the increased quality of measurement.
Since the 1980s, several large-scale nutrition surveys have been completed to observe the impact of socioeconomic changes on food consumption and chronic diseases in China. However, there has been no systematic analysis of the quality of 24-hour dietary recall, and little work has been done to consider ways to improve these dietary-assessment methods. Although more and more studies and discussions have focused on such methods in the world, minimal research has assessed the reliability of 24-hour dietary recall, and little attention has been paid to improving the quality of the dietary surveys in China.
Surprisingly little work has been undertaken to validate dietary intake methods in Asia, and many gaps hamper our understanding of the probity of the data they generate around the world [1, 2]. No established clear standards and no biochemical or other biological markers provide the necessary details to validate a dietary method [3-5]. When scales are calibrated properly, weighing (household inventory) is precise. As is standard practice for a wide range of dietary intake-validation studies, the more detailed household inventory is assumed to be the gold standard [1]. For this reason, the accuracy (repeatability and validity) of 24-hour recall is measured by comparison with it. Comparison is undertaken mainly by determining the absolute and relative differences in the results of the two methods. The same Chinese food composition tables are used for both sets of results. In other words, no biochemical or other biological markers are used to validate these methods.
We compared the household diet inventory with individual dietary recall using data from the China Health and Nutrition Survey (CHNS). Detailed data were collected for the same three consecutive days from all respondents. The study also evaluated the methodology whereby the quality of the 24-hour recall was enhanced by use of a household measure of cooking oil consumed.
There are clear reasons for collecting 24-hour dietary intake data. China has conquered the problems of food scarcity at the national level and has undergone a remarkable transition in the structure of food consumption [6]. This has gone hand in hand with marked changes in eating behaviour. For instance, AFH food consumption has increased in response to the dynamic changes in real disposable income and market labour force patterns. Within-household variations in food intake and eating patterns appear to be expanding. Food and nutrition policy is focusing less on food security needs and more on the health-related needs of selected age and gender groups. As this occurs, data on individual dietary intake become more valuable.
Of course, the major disadvantage of 24-hour recall is its reliance on respondents' ability to remember not only each food consumed but also the quantity of each item eaten during the previous day. The accuracy and precision of this method vary, therefore, depending on the population, types of foods consumed, and dietary practices [7]. In China, an additional problem is related to the measurement of cooking oil and other condiments used in food preparation, since the Chinese food composition table is based on items as purchased and not as cooked. For this reason, measuring cooking oil would have been omitted from the traditional 24-hour recall.
The traditional Chinese eating pattern is to prepare and serve a limited number of complex dishes and have each individual place various portions of each dish on his or her plate or bowl. Group consumption from common plates increases the difficulty of obtaining accurate measurement of individual consumption. People are not accustomed to estimating portion sizes. In addition, snacks and food eaten away from home depend on a respondent's memory. Estimating individual quantities is difficult, in particular for children under the age of 11 years, whose intake has to be recalled by mothers or other adults. These limitations are crucial in de creasing the validity and precision of the individual 24-hour recall method.
During the past decade, significant changes have taken place in the composition and level of dietary intake in China [8]. Two sets of large-scale nutrition surveys were conducted, both of which employed household inventories to assess dietary intake at the national and provincial levels. The longitudinal study of the CHNS and the 1992 China Third Nationwide Nutrition Survey (CNNS III) combined household inventory and individual recall to assess individual consumption. In this study, we used the CHNS 1991 survey data to compare dietary intake by 24-hour recall for the individuals with data from changes in household food inventories made during the same survey period. This analysis not only examines basic differences but uses multivariate methods to try to explain some of the reasons for differences in the results of the two methods.
In Western countries, dietary fat intake is calculated directly from food composition tables that incorporate recipes or processed foods. China has not measured and edited the nutrient contents of recipes. The available food composition tables consist of raw food items [9]. Thus, individual fat intake determined by 24-hour recall appears to be underestimated if the amount of household cooking oil used is ignored. For this reason, a method was developed to allocate edible oils as well as other common condiments (sugar, starch, soya sauce, salt) consumed in the household by each member to obtain more accurate estimates of intake.
The study selected 13,606 individuals with dietary data from 3,563 households from the 1991 CHNS. This was the second round of the longitudinal study designed in 1989 and followed up in 1991 and 1993. The cleaned files of the most recent CHNS results were not available at the time of this analysis. The design and selection of the original sample are described in detail elsewhere [8].
All fieldworkers were trained nutritionists who were professionally engaged in nutrition work in their own counties and had participated in other national nutrition surveys. Almost all interviewers were graduates of either medical colleges or vocational schools. About 160 interviewers (20 for each province) were trained for 10 days over a 2-week period in methods of collecting food-consumption data at the household and individual levels. The training covered the collection of a wide range of individual and household socio-economic and health data as well as data from other community services.
Detailed household food-consumption data were collected from changes in household food inventories for three consecutive days, in combination with a weighing and measurement technique. The three days were randomly allocated from Monday to Sunday. Chinese balances with a maximum limit of 15 kg and a minimum of 20 g were used to measure household consumption by inventory change from the beginning to the end of each day. All purchases and the collection of home production were recorded each day, as well as all available foods stored in the storage room and refrigerator at the beginning of the survey. Wastage (e.g., spoiled rice and other dishes, discarded cooked food fed to pets or animals) was estimated when weighing was not possible. At the end of the survey, all remaining foods were again weighed and recorded. The number and personal characteristics of all household members and guests were recorded.
Individual dietary data for the same three consecutive days were recorded for all household members, regardless of age or relationship to the household head. This was achieved by asking each individual, except children under 12 years of age, to report daily all food eaten AFH on a 24-hour recall basis. For younger children, the mother or a mother substitute who handled food preparation and feeding in the household was asked to recall consumption.
The same daily interview was used to collect at home individual consumption. Using food models and picture aids, trained field interviewers recorded the types, amounts, type of meal, and place of consumption of all food items during the previous day. Respondents were prompted about snacks and shared dishes. Foods eaten at restaurants, canteens, and other locations AFH were systematically recorded. Housewives and other household members were encouraged to provide additional information to determine the amounts of particular food items in dishes consumed in the household.
The amount of each dish was estimated from household inventory, and the proportion of each dish consumed was reported by each person interviewed. Thus, individual consumption was determined by the total amount and proportion each person ate.
All samples were separated into three groups based on eating habits at the household level: households with guests, households with AFH food consumption, and households with neither guests nor AFH food consumption. The definitions of guests and AFH were that on any of the three days, one or more guests were present in the household, and one or more individuals consumed one or more food items AFH.
As noted earlier, the household inventory measured the food consumed by guests. Clearly, one would expect the difference between results obtained by this method and the sum of household members' 24-hour recall to be greater in households with guests than in households without guests. In contrast, the household inventory did not include the amount of food eaten and the number of meals AFH. Thus, it would be expected that the sum of the individual dietary intakes from all household members in households with AFH eating would be greater than the results of the weighing method obtained only for the food of these members consumed at home. For these reasons, we defined and grouped the study samples according to AFH and guest status.
The total sample of 3,563 households was separated into 2,096 households without guests and AFH consumption, 152 with guests only, and 1,170 with AFH consumption only. In the 1,170 households with AFH consumption, 4,084 individuals consumed food away from home. Of these 4,084 individuals, 1,589 ate items at home that they purchased outside the home, and the remaining 2,495 ate at least one item at a restaurant, food stand, canteen, or friend's house during any of three days.
Dietary intake is expressed by the consumption per capita per day. The dietary data either from household inventory or from 24-hour individual recall were linked with a nutrient data bank for the new version of Chinese food composition tables. Individual daily consumption of cooking oils and condiments was estimated from household dietary data.
The procedure includes four main steps. First, based on the individual's characteristics (age, gender, physical activity, physical status), each individual who consumed meals in the household was converted into a "reference man" (male, weighing 65 kg, age 18 45 years, undertaking light physical activity) with the aid of Chinese recommended dietary allowances for energy. Second, these values were summed to obtain a total number of reference men for the period of dietary survey at the household level. Third, the daily consumption of oil and condiments per reference man was calculated for each household. Finally, the total amount of oil and condiments consumed was recalculated as the amount of oil per reference man for each household and allocated to each individual on the basis of the proportion of the reference man for the household. The estimated amount of individual oil and condiments consumed was added to the 24-hour dietary recall to measure individual nutrient intake.
The dietary intake figures presented in the results are the mean differences between the two methods and the ratio of the sum of the 24-hour recall for all household members divided by the household inventory method. All results are the average for each figure. This is important to point out, as the differences and the ratios can be different, since the ratio is the average of each household's ratio.
TABLE 1. Mean daily intakes of selected nutrients according to two traditional methods and one modified methoda
| Methods | Energyb (kcal) |
Proteinb (g) |
Fatb (g) |
Calories from fat (%) |
| Household inventory | 2,425 | 70.8 | 62.0 | 23.1 |
| no guest, no AFH | 2,428 | 71.1c | 58.6 | 21.8 |
| guest onlyd | 2,610 | 75.5c | 64.5 | 23.0 |
| AFH onlyc | 2,391 | 69.6c | 66.9 | 25.0 |
| Traditional dietary recall | 2,113 | 69.3 | 29.8 | 12.7 |
| no guest, no AFH | 2,127 | 69.6 | 27.2 | 11.6 |
| guest only | 2,287 | 73.2 | 27.9 | 11.3 |
| AFH only | 2,060 | 68.1 | 34.0 | 14.7 |
| Modified dietary recallc | 2,351 | 69.9 | 54.3 | 20.8 |
| no guest, no AFH | 2,355 | 70.1 | 51.1 | 19.6 |
| guest only | 2,508 | 73.8 | 51.0 | 18.9 |
| AFH only | 2,319 | 68.9 | 60.5 | 23.2 |
Analysis of covariance (ANCOVA) was used to obtain the means of selected nutrient intake adjusted for other covariates. Both dietary intake and eating patterns (household guest and AFH) are expected to vary with household income, size, and place of residence (urban status, region of residence). All analyses were carried out with SAS, version 6.08 (SAS Institute, Cary, North Carolina, USA).
Table 1 shows the mean daily intakes of selected nutrients by inventory, traditional 24-hour recall, and modified 24-hour recall. The mean intake was separated into three groups that represented different eating patterns that are relevant to subsequent analysis—no guest-no AFH, AFH only, and guest only. There were distinct differences in the mean intake from the household inventory and the 24-hour recall methods. The differences for fat intake were larger than those for protein intake and varied for the three groups. The most relevant group was the no guests-no AFH group, in which the mean intake of fat was 58.6 and 27.2 g for household inventory and traditional 24-hour recall, respectively. This is a very large and statistically significant difference, and is reflected in large differences in the energy and percentage of calories from fat measures. The modified dietary recall measure narrowed this difference significant]y. Although the difference was statistically significant, it was only 7.5 g (13%) for fat and 73 kcal (3%) for energy. It is important to note that the adjustments change the protein measure very slightly and have their major effects on fat intake. In subsequent analyses, we used the ratio of each household's total inventory and total recall results. Table 2 shows the overall summary for the comparison of these two techniques (unadjusted basic results). The ratios of selected nutrient intake from the sum of individual recall and household inventory were remarkably close. For the overall sample, there was only a 1% difference for daily calorie intake between the two methods. This increased to above 2% comparing households with and without guests. Similar results were found in households with AFH. Moreover, the ratios for daily protein and fat intake were larger than that for calories.
Comparisons by area of residence are presented in sable 3. The ratios were close to 1. Among rural residents, the ratios for calories and protein intake were almost 1. In contrast, urban residents were more likely to eat AFH than rural residents, and the sum of 24-hour intakes was greater than the household intake. The main sources of AFH consumption were workplace and school canteens, restaurants, and street stalls. Respondents were also divided into three regions—the north, which includes the provinces of Liaoning and Shandong; the central region with the provinces of Henan, Jiangsu, and Hubei; and the south, which contains Guangxi, Hunan, and Guizhou provinces. The ratios of the nutrient intake were closer in the south than in the north. This variation might reflect different eating habits in this country.
The study also examined the differences of dietary intake by household economic status. Households were stratified into the upper, middle, and lowest income tertile based on per capital income using 1990 figures. For the three groups, the ratios of daily calories, protein, and fat intake were consistent but smallest in the low-income group. The absolute differences were larger in the high-income group than in the other two groups. Two possible factors contributed to explain these differences: higher-income individuals were most likely to eat snack foods and eat AFH, and thus were least likely to provide accurate dietary data; and urban residents and families with female heads were apt to be modern factory workers and thus in the high-income group, and less likely to cooperate fully with the interview process.
TABLE 2. Comparison of dietary assessment methods: Modified individual dietary intake versus the average of house hold dietary intake
| Features of households | No. of households | Energy | Protein | Fat | Energy from fat (%) | ||||
| Diffa | Ratiob | Diff | Ratio | Diff | Ratio | Diff | Ratio | ||
| Total sample | 3,563 | -74.2 | 1.01 | -0.9 | 1.03 | -7.6 | 0.96 | -2.3 | 0.96 |
| No guest, no AFH | 2,096 | -72.7 | 1.01 | -0.9 | 1.02 | -7.6 | 0.94 | -2.3 | 0.94 |
| Guest only | 152 | -101.6 | 1.02 | -1.7 | 1.03 | -13.5 | 0.88 | -4.1 | 0.89 |
| AFH only | 1,170 | -72.0 | 1.02 | -0.7 | 1.05 | -6.4 | 1.03 | -1.9 | 1.0 |
a. Absolute difference.
b. Relative difference.
TABLE 3. Comparison of dietary assessment methods stratified by place of residence and income: Modified individual dietary intake versus the average of household dietary intake
| Features of households | No. of households | Energy | Protein | Fat | Energy from fat (%) | ||||
| Diff | Ratio | Diff | Ratio | Diff | Ratio | Diff | Ratio | ||
| Place of residence | |||||||||
| urban | 1,178 | -44.5 | 1.05 | -0.8 | 1.06 | -8.0 | 0.97 | -3.0 | 0.94 |
| rural | 2,385 | -88.9 | 0.99 | -0.9 | 1.02 | -7.5 | 0.96 | -1.9 | 0.97 |
| Region | |||||||||
| north | 885 | -83.3 | 1.03 | -1.6 | 1.04 | -7.4 | 0.96 | -2.2 | 0.95 |
| central | 1,338 | -69.6 | 1.02 | -0.3 | 1.05 | -8.7 | 0.96 | -2.6 | 0.97 |
| south | 1,340 | -72.8 | 0.99 | 1.0 | 1.01 | -6.7 | 0.97 | -2.0 | 0.97 |
| Household income tertile | |||||||||
| low | 1,175 | -101.9 | 0.99 | -1.9 | 1.01 | -6.2 | 0.96 | -1.4 | 0.97 |
| middle | 1,193 | -49.0 | 1.01 | 0.0 | 1.03 | -6.3 | 0.97 | -2.0 | 0.97 |
| high | 1,195 | -72.2 | 1.03 | -0.8 | 1.05 | -10.4 | 0.95 | -3.4 | 0.93 |
Tables 2 and 3 illustrate all the results of the modified 24-hour recall individual dietary intake—a combination of the 24-hour recall method and the use of cooking oil consumption from the household inventory. To evaluate 24-hour recall, the added cooking oil was removed for each individual and the sum of each household's traditional 24-hour recalls was calculated. Comparisons of the traditional method and the household inventory method are presented in sable 4. Daily dietary intake was much less comparable when cooking oil from the household inventory was ignored. Greater differences were found for daily calories (300-328 kcal) and fat (31-37 g) intake. The change of the percentage of calories from fat followed calorie and fat changes. In particular, the measurement of fat intake and also that of the proportion of energy from fat were most notably affected by this change in method.
TABLE 4. Comparison of dietary assessment methods: Traditional individual dietary intakea versus the average of household dietary intake
| Features of households | No. of households | Energy | Protein | Fat | Energy from fat (%) | ||||
| Diff | Ratio | Diff | Ratio | Diff | Ratio | Diff | Ratio | ||
| Total sample | 3,563 | -309.9 | 0.91 | -1.5 | 1.02 | -32.3 | 0.53 | -10.4 | 0.57 |
| No guest, no AFH | 2,096 | -299.7 | 0.91 | -1.4 | 1.02 | -31.4 | 0.49 | -10.3 | 0.53 |
| Guest only | 152 | -323.1 | 0.92 | -2.3 | 1.02 | -36.6 | 0.46 | -11.6 | 0.51 |
| AFH only | 1,170 | -327.7 | 0.91 | -1.4 | 1.04 | -33.0 | 0.60 | -10.4 | 0.64 |
a. Traditional 24-hour dietary recall was not modified by information from the household inventory method.
TABLE 5. Adjusted comparison of dietary assessment methods: Modified individual dietary intakea versus the average of household dietary intake
| Features of households | No. of households | Energy | Protein | Fat | Energy from fat (%) | ||||
| Diff | Ratio | Diff | Ratio | Diff | Ratio | Diff | Ratio | ||
| Total sample | 3,563 | -74.2 | 1.01 | -0.9 | 1.03 | -7.6 | 0.96 | -2.3 | 0.96 |
| No guest, no AFH | 2,096 | -67.9 | 1.01 | -0.7 | 1.02 | -7.8 | 0.93 | -2.4 | 0.93 |
| Guest only | 152 | -105.4 | 1.02 | -1.9 | 1.03 | -13.7b | 0.88 | -4.1b | 0.89 |
| AFH only | 1,170 | -79.0 | 1.02 | -1.0 | 1.05 | -6.0 | 1.04c | -1.6 | 1.02d |
a. Modified dietary intake adjusted for household income,
size, region, and urban or rural residence.
b. Difference of adjusted means in the guest-only group is
significantly different from those in all other groups except the
no guest, no AFH group (p < .001).
c. Ratio of adjusted mean in the AFH group is significantly
different from that in the guest-only group (p < .WI).
d. Ratio of adjusted mean in the AFH group is significantly
different from that in other groups (p< .001).
Clearly, a range of socio-economic and demographic factors can affect consumption behaviour. In particular, whether one eats at home or away from home and whether guests are present are associated with income and other similar factors. Analysis of covariance was used to obtain the least-square means for selected nutrient intake. The differences of least-square means from the sum of all individual dietary recall and household inventory became smaller than in the previous description. During this analysis, controls were introduced for household income, region of residence, urban-rural status, and household size (sable 5). In general, the results for sable 2 (unadjusted results) and sable 5 do not differ and the conclusions are similar. This means that there was a difference in fat intake patterns, in particular for households that had guests.
This study shows considerable agreement of selected nutrient intake between the adjusted 24-hour dietary recall and the household inventory methods using 1991 CHNS data. The study was based on a large sample in which individual and household data were collected for the same three consecutive days. The comparison of two dietary methods indicated considerable similarity for selected nutrient intake, despite known errors in 24-hour recall [10].
Along with most other Asian countries, China uses the method of food as purchased in preparing its food composition tables. Each of the three versions of the tables prepared by the Institute of Nutrition and Food Hygiene, Chinese Academy of Preventive Medicine, was edited based on raw food items. The updated version completed in 1991 did not include nutrient contents for food recipes. Thus, it is important to combine 24-hour individual recall with edible oil consumption to improve the quality of individual dietary data.
However, the modified method has some weaknesses. The allocation of oil among household members was based on the proportion of reference men rather than total food intake, and as such, an additional level of error is introduced for the analysis of individual intake. It is important to note that this error is glossed over in the results presented here, which summarize results for the entire family. On the other hand, these results do omit adjustment for the oil consumption for foods eaten AFH. In this survey population, a small proportion of energy and fat was consumed AFH ( < 5 % ), as a large proportion of those who ate food AFH ate only one item. An effort is under way by this research team to develop recipe files for China and to ascertain oil levels for food items commonly consumed AFH. A method of estimating this AFH oil intake should enhance the quality of the fat intake data.
Determining household use of edible oil and developing a method specific to each country to allocate this oil to each individual in the family appear to be important as means of improving the quality of 24-hour recall data. The large differences found between the 24-hour and household inventory methods when the adjustment for cooking oil was removed demonstrate this clearly. Few lower-income countries have developed food composition tables that incorporate the food-processing method. Moreover, there might be such variability in the use of edible oil that even such a table might produce misleading results.
Since we demonstrated elsewhere significant increases in oil intake in China and other Asian countries during the past decade, this methodological improvement is likely to be of increasing importance in the future in China [8]. The total consumption of fats and oils has steadily increased during the past decades in many countries of the world [11, 12]. For instance, the 1980s saw a per capita increase in the total amount of oil available for consumption of 50% to 115% in China, Indonesia, Thailand, and Malaysia [13]. This appears to be highly income-elastic in Asia, and similar increases with economic development occur in other regions (unpublished data show this is the case in North Africa and Latin America).
The study modified the traditional 24-hour individual dietary recall by adding household edible oil consumption. This improvement was more likely to provide accurate data, as was shown when compared with current dietary methods used in China. This modified dietary method allowed CHNS researchers to obtain a powerful predictor for body mass index changes among Chinese adults [6, 16]. In fact, these analyses are among the few population-based studies to link diet and obesity [3, 14, 15].
These studies, which related dietary variations over time and space to body composition changes, are our only ones that actually or approximately validate 24-hour recall data. The results show that these data only approximate those of the assumed gold standard, the household inventory, so caution must be used in interpreting them.
The findings provide clear evidence that if weighing and measurement techniques were used to collect data on edible oil in China, the quality of dietary data would be improved and more accurate values of intake would be obtained. Moreover, this should eliminate the need for collecting both 24-hour and weighing and measurement data in China. Presumably other researchers in Asia should consider following this model.
Preparation of this article was supported in part by grants from the US National Institutes of Health. We thank Frances Dancy for administrative assistance. We also thank all food and nutrition workers of the Department of Food Hygiene and Nutrition of the Anti-epidemic Station at the county and provincial levels from eight provinces (Liaoning, Shandong, Henan, Hubei, Jiangsu, Hunan, Guangxi, and Guizhou).