Methodology for the analysis of dietary data from the Mexican National Health and Nutrition Survey 2006.

Abstract

To describe the methodology for the analysis of dietary data from the Mexican National Health and Nutrition Survey 2006 (ENSANUT 2006) carried out in Mexico.
Dietary data from the population who participated in the ENSANUT 2006 were collected through a 7-day food-frequency questionnaire. Energy and nutrient intake of each food consumed and adequacy percentage by day were also estimated. Intakes and adequacy percentages > 5 SDs from the energy and nutrient general distribution and observations with energy adequacy percentages < 25% were excluded from the analysis.
Valid dietary data were obtained from 3552 children aged 1 to 4 years, 8716 children aged 5 to 11 years, 8442 adolescents, 15951 adults, and 3357 older adults.
It is important to detail the methodology for the analysis of dietary data to standardize data cleaning criteria and to be able to compare the results of different studies.

Full text

Page 1

S523
salud pública de méxico / vol. 51, suplemento 4 de 2009
Dietary methodology from ENSANUT 2006
Original article
Methodology for the analysis
of dietary data from the Mexican National
Health and Nutrition Survey 2006
Sonia Rodríguez-Ramírez, MSc, (1) Verónica Mundo-Rosas, MSc, (1)
Alejandra Jiménez-Aguilar, MSc, (1) Teresa Shamah-Levy, MSc. (1)
Rodríguez-Ramírez S, Mundo-Rosas V,
Jiménez-Aguilar A, Shamah-Levy T.
Metodología para el análisis de información dietética
de la Encuesta Nacional de Salud y Nutrición 2006.
Salud Publica Mex 2009;51 supl 4:S523-S529.
Resumen
Objetivo. Describir la metodología de análisis de la infor-
mación dietética de la Encuesta Nacional de Salud y Nutrición
de México (ENSANUT 2006). Material y métodos. Se
recolectó de la ENSANUT 2006 información dietética de la
población mediante un cuestionario de frecuencia de con-mediante un cuestionario de frecuencia de con-
sumo de siete días. Se estimó la ingesta y porcentaje de ade-
cuación de energía y nutrimentos. Se excluyeron del análisis
porcentajes de adecuación > 5 DE de la distribución general
de energía y nutrimentos y observaciones con porcentaje de
energía < 25%. Resultados. Se obtuvo información dietética
válida de 3552 niños de entre 1 a 4 años de edad, 8716
niños de 5 a 11 años de edad, 8 442 adolescentes, 15 951
adultos y 3357 adultos mayores. Conclusiones. Detallar la
metodología de análisis de la información dietética sirve para
estandarizar los criterios de limpieza de la información y hacer
comparables los resultados entre diferentes estudios.
Palabras clave: análisis de ingestion de energía y alimentos;
encuestas de nutrición; México
Rodríguez-Ramírez S, Mundo-Rosas V,
Jiménez-Aguilar A, Shamah-Levy T.
Methodology for the analysis of dietary data from
the Mexican National Health and Nutrition Survey 2006.
Salud Publica Mex 2009;51 suppl 4:S523-S529.
Abstract
Objective. To describe the methodology for the analysis of
dietary data from the Mexican National Health and Nutrition
Survey 2006 (ENSANUT 2006) carried out in Mexico. Mate-
rial and Methods. Dietary data from the population who
participated in the ENSANUT 2006 were collected through
a 7-day food-frequency questionnaire. Energy and nutrient
intake of each food consumed and adequacy percentage by
day were also estimated. Intakes and adequacy percentages
> 5 SDs from the energy and nutrient general distribution
and observations with energy adequacy percentages < 25%
were excluded from the analysis. Results. Valid dietary data
were obtained from 3552 children aged 1 to 4 years, 8716
children aged 5 to 11 years, 8442 adolescents, 15951 adults,
and 3357 older adults. Conclusions. It is important to
detail the methodology for the analysis of dietary data to
standardize data cleaning criteria and to be able to compare
the results of different studies.
Key words: energy, nutrient intake analysis; nutrition surveys;
Mexico
(1) Centro de Investigación en Nutrición y Salud, Instituto Nacional de Salud Pública. Cuernavaca, Morelos, México.
Received on: April 11, 2008 • Accepted on: November 19, 2008
Address reprint requests to: Mtra. Verónica Mundo. Instituto Nacional de Salud Pública, Av. Universidad 655,
col. Santa María Ahuacatitlán. 62100 Cuernavaca, Morelos, México.
E-mail: vmundo@insp.mx

Page 2

Original article
S524
salud pública de méxico / vol. 51, suplemento 4 de 2009
Rodríguez-Ramírez S et al.
O
evaluation of the population nutrition status1 and to be
able to provide information about the average dietary
intake of a specific population group.2 The use of dietary
data can be broadly classified into three categories:
monitoring and nutrition surveillance, analysis of the
relationship between diet and disease, and evalua-
tion of the impact of policies and programs related to
nutrition.3,4
Using methods like 24-hour recall or food frequency
allows the identification of dietary indicators that can be
employed to compare the intake of nutrient-rich food
and its association with risk factors. They also help to
establish food-intake patterns in the population.1
Care in dietary data processing (i.e. collection,
input of data, cleaning and analysis) must be taken
into consideration when designing national surveys to
ensure data will be accurate and valid, which will help
to adequately estimate energy and nutrient intakes.
Dietary data from several national nutrition sur-
veys are available in Mexico. Dietary information from
children and women of reproductive age was obtained
in the 1988 survey,5 whereas dietary information from
pre-school and school children, as well as from women
of reproductive age, was collected in the 1999 national
survey. The most recent available data come from the
Mexican National Health and Nutrition Survey 2006
(ENSANUT 2006), in which dietary information was col-
lected from pre-school and school children, adolescents,
adults, and older adults, for both sexes. Data from this
survey are representative at the national, regional, and
local urban and rural levels.6,7
It has become necessary to timely document the
tools and processes created to analyze, in a confident
manner, Mexican population dietary data taken from
probabilistic national samples. This paper aims to de-
scribe the methodology used for the analysis of dietary
data from the ENSANUT 2006.
ne of the main reasons for collecting dietary data
through national surveys is to have an integral
Material and Methods
Study population: We included data from pre-school
children aged 1 to 4 years, school children aged 5 to 11
years, adolescents aged 12 to 19 years, adults aged 20 to
59 years, and older adults aged 60 years and over.
Sample selection: The ENSANUT 2006 was a nationally
representative probabilistic survey stratified by clusters.
It was carried out from October 2005 to May 2006. Data
from 48 304 households were obtained. Collected infor-
mation included general data about households, health,
anthropometry, biochemical tests, and dietary quality
and quantity. For determining sample size it was consid-
ered that the minor importance proportion (minimum
interest prevalence) should have a value of 8.1%. It was
also considered that state estimators obtained through
the survey should have a 25% maximum relative error, a
95% confidence interval, a 20% nonresponse rate, and a
design effect of 1.7, determining a sample size of at least
1 476 households by state. A subsample corresponding
to a third of the whole sample was selected for dietary
information.8
Informed consent was obtained from each subject
or subject’s parent or guardian for their participation
in the study. The survey protocol was approved by the
Ethics Committee of the National Institute of Public
Health, Mexico.
Dietary data collection instrument: Dietary information
was collected with an adapted version of the semi-
quantitative food-frequency questionnaire found in the
Procedure Handbook for Nutrition Projects, published
by the National Institute of Public Health (INSP).9 The
questionnaire included 101 foods classified into 14
groups. For each food, number of days of intake per
week, times a day, portion size, and number of por-
tions consumed within the seven days before the date
of interview were asked.
The food table included in the questionnaire, as well
as the portion size (very small, small, medium, large,
and very large) and mean weight of portion, for each
population group, were estimated by a group of INSP
researchers from the analysis of the most-consumed-
foods data obtained from the National Nutrition Survey
1999. It is worth mentioning that portion sizes were
different for each age group studied.
Personnel training: The food-frequency questionnaire
was administered by personnel trained and standard-
ized on collection and input of data. Details on the data
collection method are available in the above-mentioned
procedure handbook.9 Data were entered into HP Com-
paq nx 6120 and Dell Latitude D510 laptop computers.
Personnel from the INSP developed and validated the
electronic questionnaires for input of data into the Fox
Pro program version 7. Input was backed up and data
were sent daily to the INSP.
Quantity of food consumed: Per-week and per-day quan-
tities of each food consumed was obtained from the
number of days per-week, times a day, portion size
(weight) and number of portions.
Cleaning of the quantity of each food consumed
by setting the maximum values that corresponded to
a plausible intake for each age, sex and physiological
status group was carried out. Food-frequency tables
for every food analyzed were generated to establish
plausible intake of each food in each population group
(pre-school children, school children, adolescents, adults

Page 3

S525
salud pública de méxico / vol. 51, suplemento 4 de 2009
Dietary methodology from ENSANUT 2006
Original article
and older adults), with the goal of observing maximum
and minimum number of portions consumed per-week
of each food included in the questionnaire. Maximum
values considered non-biologically plausible were ex-
cluded, although exclusion should not go beyond 1%
of accumulated percentage. No minimum value was
excluded and all quantities below the maximum cutoff
points for each food were considered as valid.
Cleaning of maximum and minimum values of corn flour
and wheat flour tortillas: Because tortilla is a staple for the
Mexican population, an additional cleaning procedure
was performed considering the weight of both corn
and wheat tortillas, with a valid interval of 10 g mini-
mum value and 500 g maximum value. Those values
came from the mean range observed on tortilla weight
reported by the same population. On the basis of that
range, mean weight of both corn and wheat tortillas was
calculated for each state. Mean weight was also imputed
to those cases outside the valid interval.
Energy, macronutrient and micronutrient intakes: Total
energy and nutrients contained in each food consumed
and in total diet for each individual were estimated
through the nutrient composition database compiled by
the INSP.* Energy, carbohydrate, protein, fat, fiber, iron
(total, heme, and nonheme), zinc, vitamin C, vitamin
A, folate and calcium intakes were calculated. Because
information was organized by week of intake, total
quantities of energy and nutrients were divided by 7 to
obtain an average intake per day. With respect to vitamin
C, loss during cooking was considered, so the quantity
of vitamin C was adjusted depending on type of food
and cooking method commonly used.10-13
Some of the items on the food-frequency table were
made up of several ingredients, so an average estimation
of their nutritional value was performed. That value was
calculated by adding the nutrients of each ingredient or
food (as specified in the nutrient composition database)
employed in Mexican dishes according to the propor-
tions indicated in the recipes. Food was considered in
raw condition. Although the estimation is imprecise, not
having come from a bromatological analysis, it is a good
approximation of the quantity of nutrients contained
in some dishes, as it takes into account loss factors and
gram conversion, such as densities, edible portion and
cooked-raw/raw-cooked factor.
Energy and nutrient adequacy percentage: Nutrient ad-
equacy percentage for children, adolescents and adults
was calculated from intake data using the reference val-
ues proposed by the United States Institute of Medicine
(IOM). The estimation of protein, iron, zinc, vitamin C,
retinol and folate adequacy was done using the Esti-
mated Average Requirement.14-19 For calcium, Adequate
Intake value was used as reference since the Estimated
Average Requirement value has not yet been established
because of lack of information for such calculations.20
For carbohydrates and fats, 50 and 30%, respectively,
of the energy derived from those macronutrients were
used as adequacy values. Reference values for the
estimation of protein and micronutrient adequacy are
shown in Table I.
Regarding older adults over 60 years of age, ad-
equacy percentages were calculated in accordance with
the guide of nutrition needs for older adults published
by the WHO.21
Energy adequacy for each individual was calculated
employing the Estimated Energy Requirements as the
reference.22 Therefore, it was necessary to make use
of several data included in the calculation equations;
those data were weight, height, age, and physiological
status of each individual. Energy requirements were
calculated for those individuals with a body mass index
(BMI) within the normal limits for their age and sex. For
adults, BMI limits were ≥ 18.5 and < 25; for ≥ 2-years-old
children and for adolescents those limits were based on
data distribution, considering as minimum value above
1% of the distribution, and as maximum value the cutoff
points proposed by Cole et al.23
For those individuals with a BMI outside the
normality range, weight corresponding to the BMI
limit value, considering the height of the individual,
was imputed. For instance, in adults with BMI ≥ 25 the
weight corresponding to a BMI of 24.9 was imputed. The
same was done for the lower limit value, BMI < 18.5, by
imputing the weight corresponding to a BMI of 18.5.
For those individuals without available height and
weight data, median value of height and weight for the
population with same age, sex, and physiological status,
and with data availability, was imputed.
On the other hand, as it is well known, equations
require a specific physical activity factor. There was no
data on physical activity for this analysis; therefore, we
used data from the 1999 National Nutrition Survey that
showed that women led a sedentary life.24 Likewise,
children were found to have low physical activity, in
accordance with a Torun et al. study.25 Regarding men,
they also showed low physical activity.
Cleaning of data on intake and percentages of energy and
nutrient adequacy: As suggested in dietary analyses,6, 7
intake and adequacy observations greater than 5 SDs
* INSP. Bases de datos del valor nutritivo de los alimentos. Compi-
lación del Instituto Nacional de Salud Pública. 2004 (Unpublished
document).

Page 4

Original article
S526
salud pública de méxico / vol. 51, suplemento 4 de 2009
Rodríguez-Ramírez S et al.
Table I
RefeRence values of estimated aveRage RequiRements of micRonutRients and pRoteins foR the estimation
of diet adequacy in mexican population. mexico, ensanu t 2006
Group

Age
(Years)
Vitamins
Vit. C (mg)
Minerals
Zinc (mg)Protein (g) Folate (μg)Retinol (RE)* (μg)Iron (mg)Calcium‡ (mg)
Boys and girls§



1 - 3
4 - 8
10.44
15.2
120
160
13.0
22.0
210
275
3.0
4.1
2.5
4.0
500
800
Boys§


9 - 13
14 - 18
27.36
44.53
250
330
39.0
63.0
445
630
5.9
7.7
7.0
8.5
1 300
1 300
Girls§


9 - 13
14 - 18
27.36
38.34
250
330
39.0
56.0
420
485
5.7
7.9
7.0
7.3
1 300
1 300
Women§



19 - 30
31 - 50
51 - 60
38.0
38.0
38.0
320
320
320
60.0
60.0
60.0
500
500
500
8.1
8.1
5.0
6.8
6.8
6.8
1 000
1 000
1 200
Pregnant§



14 - 18
19 - 30
31 - 50
50.1
50.1
50.1
520
520
520
66.0
70.0
70.0
530
550
550
23.0
22.0
22.0
10.0
9.5
9.5
1 300
1 000
1 000
Breastfeeding§



14 - 18
19 - 30
31 - 50
59.9
59.9
59.9
450
450
450
96.0
100.0
100.0
885
900
900
7.0
6.5
6.5
10.9
10.4
10.4
1 300
1 000
1 000
Men§


19 - 30
31 - 60
47.0
47.0
1 g/kg of
weight by day
320
320
75.0
75.0
625
625
6.0
6.0
9.4
9.4
1 000
1 000
Older adults#
≥60 years
40080.0 65010.0
7.0 (men)
4.9 (women)
1 000
* Retinol equivalents
‡ Adequate intake, as data of estimated average requirements are unavailable
§ Requirement proposed by the United States Institute of Medicine
# As recommended by the World Health Organization
from the general distribution of energy, macronutrients,
iron, zinc, vitamin C, vitamin A, folates and calcium
were excluded from the analysis. Further on, data dis-
tribution was graphically analyzed, and those observa-
tions within 5 SDs but graphically separated from the
conglomerate of the majority of data were eliminated as
well. Concerning lower values, energy adequacy values
less than 25% were eliminated for all age groups, as
adequacy percentages less than that could not represent
an intake compatible with life.

process, we calculated new expansion factors.
Because dietary data distribution is usually biased
toward high values, data are presented as medians, 25
and 75 percentiles of the distribution. The process to
obtain nutrient vectors from data was done in Access
(Microsoft Office, 2003) using the Visual Basic code and
SQL consults. Data cleaning was performed with Stata,
version 9.2 (Stata Corporation, TX, USA) and SPSS, ver-
sion 15.0 (SPSS Inc., Chicago, IL, USA).
Due to the data elimination during the cleaning

Page 5

S527
salud pública de méxico / vol. 51, suplemento 4 de 2009
Dietary methodology from ENSANUT 2006
Original article
Results
We analyzed dietary information of 39 998 individuals.
Data lost by the cleaning process were between 3.3 and
11.9% among the different groups of population.
Pre-school children: Dietary data from 3959 children less
than 5 years old were obtained. Because 461 observa-
tions lacked data on weight and height, it was necessary
to impute data from the same study population in ac-
cordance with age and sex so as to be able to calculate
the percentage of energy adequacy.
After cleaning, 407 observations were excluded
(10.3%). In 203 of those, there was low energy (less
than 25%) or energy adequacy greater than 5 SDs from
the adequacy mean. The other 204 observations had
adequacies above 5 SDs from the adequacy of several
nutrients. Final sample size after cleaning of all the
main variables (energy, macronutrients, total iron,
zinc, vitamin C, retinol equivalents, folate and calcium)
was 3 552 observations, which represented a total of 7
836 674 children aged 1 to 4 years countrywide.
School children: Dietary data were collected from 9 383
children between 5 and 11 years of age. In 396 observa-
tions it was necessary to impute height and weight of
the distribution median according to age and sex since
that data was not available.
Because 667 observations (7.1%) had adequacy
percentages above 5 SDs for one or more of the nutrients
studied, or an energy adequacy percentage less than
25%, they had to be excluded.
In total, 8 716 valid complete data were analyzed,
representing 15 494 558 children aged 5 to 11 years
(Table II).
Adolescents: Data from 8 768 adolescents aged 12 to 19
years were obtained. It was necessary to impute height
and weight of the distribution median in 532 and 545
observations, respectively, in accordance with age and
sex, as those observations lacked that information.

eliminated because of energy and/or nutrient adequacy
percentages above 5 SDs or energy adequacy percentage
less than 25%. Therefore, 8 442 observations with valid
dietary information were considered, which represented
18 276 531 adolescents at a national level.
Adults aged 20 to 59 years: Data from 16 494 adults were
obtained, and 543 (3.3%) observations were eliminated
as they presented intakes and adequacy percentages
above 5 SDs or energy intakes less than 25%. Thus,
valid information corresponded to 15 951 individuals
aged 20 to 59 years, who represented a population of
47 946 764 nationwide.
Older adults: Data from 3 812 adults between 60 to 99
years old were collected, and 455 observations were
eliminated (11.9%) owing to adequacies above 5 SDs
from the energy and nutrient adequacy mean, or energy
adequacies less than 25% of those recommended.
In total, valid information was obtained from 3 357
adults who represented 11 084 070 older adults aged 60
years and over nationwide.
Table II shows original sample size and total
number of feasible data, by age group and with valid
dietary information, obtained through the different
cleaning stages.
After cleaning, 326 observations (3.7%) were
Discussion
The contribution of this paper is that it shows in detail
how the methodology for processing and analyzing
dietary data from the ENSANUT 2006 in Mexico was
established, from its planning to the presentation of
results. This methodology has been used in the presen-
tation of descriptive dietary information of different
groups from the population in the ENSANUT 2006.26-29
This paper can be useful to standardize data cleaning
in different population groups and in different studies,
making it possible to compare results and to handle
data easily, as well as to establish criteria and to validate
Table II
sample size of dietaRy data analysis. mexico, ensanu t 2006
Study group Total sample size Data loss due to cleaning
Sample size with feasible
dietary data
Sample
expansion factor
Pre-school children (1- 4 years old)
School children (5-11 years old)
Adolescents (12-19 years old)
Adults (20 -59 years old)
Older adults (60 years and over)
3 959
9 383
8 768
16 494
3 812
407
667
326
543
455
3552
8716
8442
15 951
3 357
7 836 674
15 494 558
18 276 531
47 946 764
11 084 070