Validation of Dietary Assessment Methods in Preschool Age
Populations
General
Twenty-three validation studies in preschool populations are included
in Table 4.1. Studies with children older than 5 years were included if
they had a substantial number of preschool children. A recent review includes
a few additional studies through 2000 from Senegal, Malawi, and Ghana
as well as reproducibility measurements (137).
Studies
Food Records (FRs)
Two studies (129, 92) found close agreement between food records and the reference
method. Davies (129)
found close agreement (3%) at a group level between energy intake
from 4-day weighed FRs and DLW (Doubly Labeled
Water) total energy expenditure measurements in a group of 81 preschool
children recruited from towns around Cambridge. In this study, mothers
weighed all food consumed at home, but used a notebook to record details
of food consumed away from home. The results lead to the recommendation
to use the 4-day weighed FRs in the UK National Diet and Nutrition Surveys
program, which investigates the diets and nutritional status of representative
samples of British children age 1.5 to 4.5 years.
In a UK study of a preschool population of recent Indo-Asian immigrants,
close agreement (7%) was found between energy intake estimates from a
5-day weighed FR and a diet
history conducted in the homes (92).
However, in this low-literacy population, the use of the Portable Electronic
Tape Recording Automated (PETRA) scale by caregivers was problematic,
requiring intensive instruction and monitoring by investigators and frequent
repair.
Diet History (DH)
Livingston and colleagues (138)
validated a 1- to 2-hour DH interview with the DLW method in a small group
of children in the UK who were age 3 to 5 years. The DH overestimated
total energy expenditure measurements by 9%.
24-Hour Recall (24HR)
Two studies validating a 3-pass 24HR interview with the DLW method were
found. In 2001, a representative sample of 41 preschoolers (age 3 to 4
years) from Glasgow, Scotland, completed a DLW measurement and three telephone-administered
24HR in 7 days with a parent. Mean 24HR energy intakes overestimated DLW
energy expenditure measurements by about 11 percent, with no evidence
that the bias was related to diet composition or weight status. In an
older sample of 24 children (age 4 to 7 years), Johnson found closer agreement
(3% underestimation) between energy intakes reported in three 24HR interviews
and DLW energy expenditure estimates at the group level. The limits of agreement were wide
in both studies, indicating no agreement between methods at an individual
level.
Three studies comparing intake reported in a 24HR interview with direct
observation of intake at one or more meals found agreement between reported
and observed energy intake within 10 percent (139-141). In a study of
a diverse group of 56 preschoolers, teams of observers shadowed children
for 12 hours recording food intake (139). A 24HR interview with the child's
mother the following day underestimated observed energy intake by 7% and
found only 65% agreement in food items. Although socioeconomic status
was not consistently related to reporting errors, racial/ethnic differences
were observed (white and Hispanic mothers tented to underreport and blacks
tended to overreport). In addition, mothers who were not at home with
children were less able to report on their child's diet for a large part
of the day. In a similar study in older Hispanic children (age 4 to 7
years), a 24HR on the day following direct observation of the evening
meal in the child's home overestimated energy intake
by 9% (140). A third study observing children eating with their parents
in a cafeteria found mothers and fathers interviewed separately on the
following day were equally able to recall the child's intake of energy
and nine other nutrients (141).
Two small studies found no significant differences between energy and
nutrient intakes estimated by 24HR interviews and 1- to 3-day weighed
FRs in children age 2 to 4 years (142;143). A third study in older children
(age 4 to 8 years) found close agreement (2%) for energy intake between
a 24HR and 7-day estimated FR, but less agreement for the nine nutrients
examined (128;143). In two of these studies, child care workers participated
in food recording or measurement (128;142;143). One study included the
child care provider in the recall interview (142).
As summarized in a recent similar review (137),
agreement between the food recalls in these validation studies and the
reference measurement varied by food group. Intake of main meal items
was more likely to be reported than intake of desserts and snack food
(141)
and it was more common to omit than add food items (139;140;143).
Portion size estimation was inconsistent with both over- and underestimation
of various foods. Four studies reported correlation coefficients for energy
and macronutrients >0.45 (140-143).
Investigators included very little examination of effect of sex, ethnicity,
or weight status on recall validity; no effect of gender on validity was
found in the two DLW method studies (144;145)
but some reporting differences were found by gender in one direct observation
study (139).
Food Frequency Questionnaires (FFQs)
Although the eight FFQ validation studies examined differed in the actual
FFQ instrument and reference method, all found the FFQ overestimated energy
and other nutrients (Table
4.1). The only validation study using the DLW method as the reference
measurement found the HFFQ (Harvard
Food
Frequency Questionnaire) overestimated energy intake by 59% (146).
No differences were found in reporting by gender, ethnicity, or the body
composition of the child, but the paternal percent body fat was significantly
correlated with misreporting of energy intake. It is speculated that the
use of adult portion sizes on the FFQ may have contributed to the sizeable
overestimation (137).
The three studies comparing various FFQs with multiple (2 to 4) 24HR
recalls spaced at varying intervals, found the FFQ overestimated intake
by 42% (147),
66% to 73% (136),
and 70% (102).
These three studies varied greatly by how information was obtained on
the child's intake while with a child care provider or preschool. One
study contacted alternative care givers for information (102),
another excluded times when food was consumed outside of the parent's
supervision from the recall data (136),
and another included only children whose intake was directly observed
by the 24HR respondent (147).
Two studies examined the validity of the HFFQ in a WIC population with
widely varying results. In one study, the HFFQ (modified for a 1-month
time period) agreed closely with results from three 24HR interviews (administered
10 days apart). The FFQ overestimated energy intake slightly (0.2% overestimation)
and was within 10% of 24HR intakes for 20 nutrients (103).
However, the other validation study compared the HFFQ or the NCI HHHQ
(Health
Habits and History Questionnaire) with three 24HR interviews and found
low correlations for energy (HFFQ = 0.13/HHHQ = 0.14) and also for five
nutrients with each FFQ (25).
Other reference methods for validating the FFQ have found varying results.
Energy-adjusted correlations between a HFFQ modified for the previous
7 days and a 3-day weighed FR were significant, but the HFFQ was administered
immediately after the weighed FR when parental awareness of food intake
would be expected to be high (142). In a urban Hispanic community, the
reported percentage of calories from fat did not differ between the bi-annual
administration of the HFFQ and quarterly 24HRs (148). Although the HFFQ
overestimated intake of total fat, saturated fat, and cholesterol, significant
associations were found between total serum cholesterol and LDL cholesterol
and classification of children into tertiles of total fat, saturated fat,
calorie adjusted fat, and total fat intake.
Other Questionnaires
FFQ instruments targeting specific foods or nutrients show promise. A
beverage FFQ showed high correlations (>.5) with intake of milk and
four other beverages reported on a 3-day estimated FR (98). Estimates
of total fat, saturated fat, and dietary cholesterol correlated well between
a 17-item fat FFQ and a 24HR and 3-day estimated FR in a rural Head Start
population (149). A calcium FFQ overestimated calcium intake by 18% but
correlated well with a 4-day estimated FR, and also classified 84% of
the children correctly into calcium intake quartiles (150).
|
