Elsa Resano-Pérez, MD,^{(1)} Ignacio Méndez-Ramírez, PhD,^{(2)} Teresa Shamah-Levy, BSc,^{(3) }Juan A Rivera, MS, PhD,^{(3)} Jaime Sepúlveda-Amor, MD, DSc.^{(3) }

(1) Secretaría Académica, Instituto Nacional de Salud Pública. Cuernavaca, Morelos, México.

(2) Instituto de Investigaciones en Matemáticas Aplicadas, Universidad Nacional Autónoma de México. México, DF, México.

(3) Centro de Investigación en Nutrición y Salud, Instituto Nacional de Salud Pública. Cuernavaca, Morelos, México.

(1) Secretaría Académica, Instituto Nacional de Salud Pública. Cuernavaca, Morelos, México.

(2) Instituto de Investigaciones en Matemáticas Aplicadas, Universidad Nacional Autónoma de México. México, DF, México.

(3) Centro de Investigación en Nutrición y Salud, Instituto Nacional de Salud Pública. Cuernavaca, Morelos, México.

Objetivo. Describir la metodología y análisis de la Encuesta Nacional de Nutrición 1999 (ENN-99). Material y metódos. La ENN-99 es probabilística, con representatividad nacional de cuatro regiones y estratos urbano/rural. Las unidades últimas de muestras son los hogares seleccionados por muestreo estratificado y por conglomerados. La población de estudio fueron los niños menores de cinco años de edad, los niños 6-11 años (escolares) y las mujeres de 12-49 años, en quienes se obtuvieron mediciones antropométricas, niveles de hemoglobina, información sobre morbilidad y sus determinantes, datos socieconómicos y demográficos, y en una submuestra se determinaron micronutrimentos (hierro, zinc, vitamina A, ácido fólico, vitamina C, yoduria) y se obtuvo información sobre dieta. Resultados. La Encuesta obtuvo una tasa de respuesta de 82.3%, la no respuesta se asoció en 5.9% a no respuesta del informante y el resto fue ocasionado por no estar habitadas las viviendas seleccionadas. Conclusiones. La información obtenida a través de esta encuesta actualiza la información existente en nuestro país y es la base para la formulación de políticas y programas prioritarios.

# ABSTRACT

Objective. To describe the methods and analyses of the 1999 National Nutrition Survey (NNS-99). Material and Methods. The 1999 National Nutrition Survey (NNS-99) is a probabilistic survey with nationwide representativity. The NNS-99 included four regions and urban and rural areas of Mexico. The last sampling units were households, selected through stratified cluster sampling. The study population consisted of children under five years of age, schoolage children (6-11 years), and women of childbearing age (12-49 years). Data were collected on anthropometric measurements, hemoglobin levels, morbidity and its determinants, and socioeconomic and demographic characteristics. In addition, data on diet and micronutrients intakes (iron, zinc, vitamin A, folic acid, vitamin C, and iodine) were obtained in a sub-sample of subjects. Results. The response rate for the NNS-99 was 82.3%; the non-response rate was 5.9% and the remaining did not participate due to uninhabited houses. Conclusions. This survey updates the information on nutritional status in Mexico and should serve as the basis for food and nutrition policy-making and priority program design.

## Introducción

## Material y Métodos

## Bibliografía

### Enlaces refback

The first national probabilistic nutrition and diet survey in Mexico was carried out in 1988. The National Nutrition Survey 1988^{1} (NNS-1988) described the nutrition situation for the first time, both nationwide and by region (the country was divided into four regions for that purpose). Surveys previous to 1988 and other recent surveys had excluded either urban or rural zones and had not included representative samples, thus limiting their scope. Unlike those surveys, the NNS-1988 included both urban and rural zones. Also, its probabilistic sampling design made it representative at the national and regional level.

During the NNS-1988, data were collected from children <5 years of age, and women of reproductive age, the groups considered to be the most nutritionally vulnerable. Reliability of anthropometric measures was assured through collection by well-trained and standardized personnel. Height and weight measurements, together with age data, could be used for constructing indicators to differentiate between types of malnutrition. The NNS-1988 collected data from over 13 000 households, including 19 000 women aged 12-49 years and almost 7 500 children <5 years of age. Data were collected from women and children on anthropometric measures, dietary intake, sociodemographic characteristics, and health and illness indicators. Data on anemia were collected from women only.

A second NNS was carried out 10 years after the first one, to account for demographic and socioeconomic changes that were likely to be influencing the prevalence and distribution of malnutrition due to deficiency of nutrients and overweight. It was necessary to update the information on the nutritional status of he population, given the high prevalence of malnutritionand overweight found in the previous survey. This paper presents the methods of the 1999 National Nutrition Survey.

During the NNS-1988, data were collected from children <5 years of age, and women of reproductive age, the groups considered to be the most nutritionally vulnerable. Reliability of anthropometric measures was assured through collection by well-trained and standardized personnel. Height and weight measurements, together with age data, could be used for constructing indicators to differentiate between types of malnutrition. The NNS-1988 collected data from over 13 000 households, including 19 000 women aged 12-49 years and almost 7 500 children <5 years of age. Data were collected from women and children on anthropometric measures, dietary intake, sociodemographic characteristics, and health and illness indicators. Data on anemia were collected from women only.

A second NNS was carried out 10 years after the first one, to account for demographic and socioeconomic changes that were likely to be influencing the prevalence and distribution of malnutrition due to deficiency of nutrients and overweight. It was necessary to update the information on the nutritional status of he population, given the high prevalence of malnutritionand overweight found in the previous survey. This paper presents the methods of the 1999 National Nutrition Survey.

The study population consisted of Mexican subjects residing in their households at the time of the study, of the following age groups: children <5 years of age, school-age children (5-11 years), and women aged 12-49 years. The survey was conducted between October 1998 to March 1999. It included the entire country, aggregated by localities with < 2 500 inhabitants,* <15 000 inhabitants, and 15 000 or more inhabitants, as well as into four regions.*

* Instituto Nacional de Estadística Geografía e Informática establishes an operational definition of rural localities as those < 2 500 inhabitants and of urban localities as those > 2 500 inhabitants, according to the last National Household Sampling Frame of INEGI.

* North: Baja California, Baja California Sur, Coahuila, Chihuahua, Durango, Nuevo León, Sonora, Tamaulipas; Center: Aguascalientes, Colima, Guanajuato, Jalisco, México (excluding municipalities and localities adjoining México City), Michoacán, Morelos, Nayarit, Querétaro, San Luis Potosí, Sinaloa, Zacatecas; South: Campeche, Chiapas, Guerrero, Hidalgo, Oaxaca, Puebla, Quintana Roo, Tabasco, Tlaxcala, Veracruz, Yucatán and México City (México City and adjoining urban municipalities of the State of México).

The survey included diverse data collection strategies for the different age groups. Table I describes collection strategies, surveyed population, and type of data collected. Three different samples are readily apparent in Table I. The original sample included the entire study, with the following data: household characteristics, morbidity, anthropometry, capillary blood sample, and breastfeeding and complementary feeding of children <2 years.

A sub-sample was characterized by application of a food questionnaire and a women’s questionnaire. A sub-sample of the sub-sample selected for dietary assessment was utilized to obtain a venous blood sample, a urine sample, and table salt intake.

Sample design

The probabilistic nature of the sample allowed making inferences to the target population and ensured the quality of estimations, mainly presented as rates, means, and proportions.

The sampling frame corresponded to the National Household Sampling Frame of INEGI.^{2} This sampling frame was constructed by stratifying within each Mexican state to demarcate zone grouping localities.‡

‡ Zone I: Cities and metropolitan areas selected for the National Survey of Urban Employment (ENEU); Zone II: Other cities of 100 000 and more inhabitants or state capitals; Zone III: Localities with 20 000 to 99 999 inhabitants; Zone IV: Localities with 15 000 to 19 999 inhabitants Zone V: Localities with 2 500 to 14 999 inhabitants; Zone VI: Localities with <2 500 inhabitants.

Construction of sampling units

Primary sampling units. The primary sampling units (PSU) consisted of one or part of one Basic Geographical Statistical Area (BGSA). This is described as the geographical reference area for statistical information, in localities with 2 500 or more inhabitants it is made up of a number of household blocks delimited by streets and avenues (urban BGSA). Localities with <2 500 inhabitants form a rural BGSA within a geographseveral adjoining BGSAs when the number of households was less than the required number to form a PSU. Depending on the zone, they were conformed as follows: zone I PSU: A BGSA with at least 480 households –the combination of two or more adjoining BGSA from the same stratum, with at least 480 households–; zones II to VI PSU: A BGSA or the combination of two or more BGSA with at least 280 households in localities with 2 500 or more inhabitants, or 100 households in localities with <2 500 inhabitants.

* North: Baja California, Baja California Sur, Coahuila, Chihuahua, Durango, Nuevo León, Sonora, Tamaulipas; Center: Aguascalientes, Colima, Guanajuato, Jalisco, México (excluding municipalities and localities adjoining México City), Michoacán, Morelos, Nayarit, Querétaro, San Luis Potosí, Sinaloa, Zacatecas; South: Campeche, Chiapas, Guerrero, Hidalgo, Oaxaca, Puebla, Quintana Roo, Tabasco, Tlaxcala, Veracruz, Yucatán and México City (México City and adjoining urban municipalities of the State of México).

The survey included diverse data collection strategies for the different age groups. Table I describes collection strategies, surveyed population, and type of data collected. Three different samples are readily apparent in Table I. The original sample included the entire study, with the following data: household characteristics, morbidity, anthropometry, capillary blood sample, and breastfeeding and complementary feeding of children <2 years.

A sub-sample was characterized by application of a food questionnaire and a women’s questionnaire. A sub-sample of the sub-sample selected for dietary assessment was utilized to obtain a venous blood sample, a urine sample, and table salt intake.

Sample design

The probabilistic nature of the sample allowed making inferences to the target population and ensured the quality of estimations, mainly presented as rates, means, and proportions.

The sampling frame corresponded to the National Household Sampling Frame of INEGI.

‡ Zone I: Cities and metropolitan areas selected for the National Survey of Urban Employment (ENEU); Zone II: Other cities of 100 000 and more inhabitants or state capitals; Zone III: Localities with 20 000 to 99 999 inhabitants; Zone IV: Localities with 15 000 to 19 999 inhabitants Zone V: Localities with 2 500 to 14 999 inhabitants; Zone VI: Localities with <2 500 inhabitants.

Construction of sampling units

Primary sampling units. The primary sampling units (PSU) consisted of one or part of one Basic Geographical Statistical Area (BGSA). This is described as the geographical reference area for statistical information, in localities with 2 500 or more inhabitants it is made up of a number of household blocks delimited by streets and avenues (urban BGSA). Localities with <2 500 inhabitants form a rural BGSA within a geographseveral adjoining BGSAs when the number of households was less than the required number to form a PSU. Depending on the zone, they were conformed as follows: zone I PSU: A BGSA with at least 480 households –the combination of two or more adjoining BGSA from the same stratum, with at least 480 households–; zones II to VI PSU: A BGSA or the combination of two or more BGSA with at least 280 households in localities with 2 500 or more inhabitants, or 100 households in localities with <2 500 inhabitants.

Secondary sampling units. The secondary sampling unit (SSU) in zone I was composed of one or more complete and adjacent blocks with at least 40 inhabited households. In zones II through VI, the SSU consisted of private households.

Tertiary sampling units. The tertiary sampling units (TSU) were found only in zone I and consisted of private households.

Stratification

Sample size calculation

The sample size should be sufficient to ensure that estimations are of good statistical quality. In a multiplepurpose survey, the calculation of an adequate sample size for each single purpose may be quite complex due to the number of variables and relationships of interest. Multiple analyses should be approached by selecting the most important variables that have the lowest values, to obtain the minimum sample size required when the event of interest is uncommon in the total population, thus including by default other more common events. The sample size was calculated using the formula:^{3}

The sample size was n=2 000 households. The proportions utilized as reference to calculate the sample size came from the National Nutrition Survey 1988 (NNS-1988) and are presented in Table II. A basic variable to calculate the sample size was the proportion of children <2 years of age starting milk formula feeding in the first 4 to 5 months of life, since this is an uncommon event representing the smallest study group.

Table III shows the sample distribution by region and locality. Finally, the sub-sample for diet was applied in 4 200 households for each of the three intended populations, and the sub-sample for venous blood collection in 2 000 households for those three populations.

A multistage, stratified, cluster sampling selection was conducted. In each household, data were collected from all children <5 years of age, all school-age children (5-11 years), but only for one woman aged 12-49 years.

Households were selected independently in each state/zone. This process varied by zone.

Zone I. In the first stage ngh PSU were selected with probability proportional to size (number of households in the PSU in the sampling frame). In the second stage k SSU were selected with probability proportional to size (number of households in the SSU in the sampling frame) in each PSU selected in the first stage. In each SSU, five households were selected without replacement with equal probability. Thus, the probability of selecting a household was given by:

Where:

n_{gh}= PSU sample size;

m_{ghi}= number of households of the i-th PSU in the h-th

stratum of the g-th state;

m_{gh}= number of households in the h-th stratum of the gth

state;

m_{ghij}= number of households in the sampling frame in

the j-th SSU of the i-th PSU in the h-th stratum of

the g-th state;

k = SSU sample size, and

m*_{ghij}= number of households listed in field work for the

j-th SSU of the i-th PSU in the h-th stratum for the

g-th state

Zones II to V. In the first stage n_{gh} PSU were selected with probability proportional to size (number of households in the PSU in the sampling frame). In the second stage t households were selected without replacement with equal probability within each PSU selected in the first stage. Thus, the probability of household selection in the PSU_{i} is given by:

Where:

_{ghi}= total number of households listed in field work for

the i-th PSU within the h-th stratum of the g-th state.

Zone VI. In the first stage n_{gh} PSU were selected with probability proportional to size (number of households in the PSU in the sampling frame). In the second stage, there were two or four segments of 10 households on average in each PSU. Therefore, the probability of selecting a household in the PSU_{i} of the h stratum in the g state was given by:

Where:

n_{gh}= PSU sample size;

m_{ghi}= number of households for the i-th PSU within the

h-th stratum of the g-th state;

m_{gh}= number of households in the h-th stratum of the g-th state;

n_{seg}= sample size of segments, and

N_{seg}= total number of segments constructed in the PSU.

Selection of the sub-sample for dietary data. The sub-sample for diet was selected independently for each study population; one of every five of the 21 000 households was selected. In this way, the probability of selecting a household where dietary data were collected for each of the three target populations was given by:

Where:

PD(V_{ghij})= probability of selecting a household for the subsample

on diet, and

P(V_{ghij})= probability of selecting a household within the

j-th SSU within the i-th PSU in the h stratum of

the g state.

For the subgroup of women only one woman was selected from those living in a given household; thus, the probability of selecting a woman is given by:

Where:

PM= probability of selecting a woman, and

NMV= number of women in the household.

Selection of the sub-sample for venous blood data. For each target population, one of every three households was selected from those selected for the diet sub-sample. Venous blood specimens were collected from the target population.

The probability of selecting the household for the venous blood specimen collection was:

Where:

PS(V_{ghij})= probability of selecting a household for venous

blood collection.

Estimates

The estimate of the total number of characteristic X is given by:

Where:

F_{ghi}= expansion factor of the household within the i-th

PSU of the h stratum in the g state (inverse of the

probability of selecting a household), and

X_{ghik}= value of the characteristic of interest in the k-th interview

within the i-th PSU in the h stratum of the g

state.

The estimation of proportions, rates, and means

uses the combined rate estimator R, given by:

Tertiary sampling units. The tertiary sampling units (TSU) were found only in zone I and consisted of private households.

Stratification

Once constructed, the PSUs were stratified by an index of socioeconomic status for each locality and zone using the following variables from the Population and Household Census of 1995: percent of literate population aged 6 to 14 years, percent of literate population aged 15 years, percent of households with public sewage system, percent of households with potable water. For zone VI, the variable “percent of households with electricity” was added. The index was the first principal component obtained wi th the variables for each zone and locality.

Sample size calculation

The sample size should be sufficient to ensure that estimations are of good statistical quality. In a multiplepurpose survey, the calculation of an adequate sample size for each single purpose may be quite complex due to the number of variables and relationships of interest. Multiple analyses should be approached by selecting the most important variables that have the lowest values, to obtain the minimum sample size required when the event of interest is uncommon in the total population, thus including by default other more common events. The sample size was calculated using the formula:

where:

n= sample size in number of households;

P= proportion of event of interest;

Q= 1 – P;

r= maximum acceptable relative error;

k= preset value for appropriate statistical estimation;

DEFF= design effect from cluster sampling;

TNR= non-response rate, and

PHV= average number of inhabitants per household.

Substituting the expression yields:

p= .0575;

Q= 0.9425;

r= 0.165;

k= 1.645 (value of tables with a 90% confidence);

DEFF= 2.3 (from an average of DEFF of above variables from

NNS-1988);

TNR= 0.15, and

PHV= 0.21 children <2 years of age per household.

n= sample size in number of households;

P= proportion of event of interest;

Q= 1 – P;

r= maximum acceptable relative error;

k= preset value for appropriate statistical estimation;

DEFF= design effect from cluster sampling;

TNR= non-response rate, and

PHV= average number of inhabitants per household.

Substituting the expression yields:

p= .0575;

Q= 0.9425;

r= 0.165;

k= 1.645 (value of tables with a 90% confidence);

DEFF= 2.3 (from an average of DEFF of above variables from

NNS-1988);

TNR= 0.15, and

PHV= 0.21 children <2 years of age per household.

The sample size was n=2 000 households. The proportions utilized as reference to calculate the sample size came from the National Nutrition Survey 1988 (NNS-1988) and are presented in Table II. A basic variable to calculate the sample size was the proportion of children <2 years of age starting milk formula feeding in the first 4 to 5 months of life, since this is an uncommon event representing the smallest study group.

Households were selected independently in each state/zone. This process varied by zone.

Zone I. In the first stage ngh PSU were selected with probability proportional to size (number of households in the PSU in the sampling frame). In the second stage k SSU were selected with probability proportional to size (number of households in the SSU in the sampling frame) in each PSU selected in the first stage. In each SSU, five households were selected without replacement with equal probability. Thus, the probability of selecting a household was given by:

Where:

n

m

stratum of the g-th state;

m

state;

m

the j-th SSU of the i-th PSU in the h-th stratum of

the g-th state;

k = SSU sample size, and

m*

j-th SSU of the i-th PSU in the h-th stratum for the

g-th state

Where:

n_{gh}= PSU sample size;

m_{ghi}= number of households of the i-th PSU in the h-th

stratum of the g-th state;

m_{gh}= number of households within the h-th stratum of the g-th state, and

m

stratum of the g-th state;

m

m*

the i-th PSU within the h-th stratum of the g-th state.

Zone VI. In the first stage n

Where:

m

h-th stratum of the g-th state;

m

n

N

Selection of the sub-sample for dietary data. The sub-sample for diet was selected independently for each study population; one of every five of the 21 000 households was selected. In this way, the probability of selecting a household where dietary data were collected for each of the three target populations was given by:

Where:

PD(V

on diet, and

P(V

j-th SSU within the i-th PSU in the h stratum of

the g state.

For the subgroup of women only one woman was selected from those living in a given household; thus, the probability of selecting a woman is given by:

Where:

PM= probability of selecting a woman, and

NMV= number of women in the household.

Selection of the sub-sample for venous blood data. For each target population, one of every three households was selected from those selected for the diet sub-sample. Venous blood specimens were collected from the target population.

The probability of selecting the household for the venous blood specimen collection was:

Where:

PS(V

blood collection.

Estimates

The estimate of the total number of characteristic X is given by:

Where:

F

PSU of the h stratum in the g state (inverse of the

probability of selecting a household), and

X

within the i-th PSU in the h stratum of the g

state.

The estimation of proportions, rates, and means

uses the combined rate estimator R, given by:

The fitting of sample per region is shown in Table IV.

Data collection results

Table V shows that the response rate was 82% of households. The non-response in 4.46% of households was due to refusal or to the absence of a proper informant; 13.3% of households were uninhabited or just temporally inhabited. The response rate per region is also shown.

Data analysis. Data analysis included a socioeconomic index, modified after that proposed by Bronfman,^{4} by weighting different predictors using the principal components method. Predictor variables were: household flooring material, potable water, and ownership of electrical home appliances (washing machine, refrigerator, television, radio, and stove). This index explained 51.6 % of the variance of the set of variables, with a single component scoring poverty (SES –2.28797 to 1.54518).

1. Sepúlveda-Amor J, Lezana MA, Tapia-Conyer R, Valdespino JL, Madrigal H, Kumate J. Estado nutricional de prescolares y mujeres en México: resultados de una encuesta probabilística nacional. Gac Med Mex 1990;126(3):207-225.

2. Instituto Nacional de Estadística, Geografía, e Informática. Resultados del Conteo de Población y Vivienda de 1995. México, DF: INEGI, diciembre 1995.

3. Cochran W. Sampling techniques. Third edition. Otawa, Canada: John Wiley & Sons, 1977.

4. Bronfman M, Guiscafré H, Castro V, Castro R, Gutiérrez G. La medición de la desigualdad: una estrategia metodológica, análisis de las características socioeconómicas de la muestra. Arch Invest Med 1988;19:351-360.

2. Instituto Nacional de Estadística, Geografía, e Informática. Resultados del Conteo de Población y Vivienda de 1995. México, DF: INEGI, diciembre 1995.

3. Cochran W. Sampling techniques. Third edition. Otawa, Canada: John Wiley & Sons, 1977.

4. Bronfman M, Guiscafré H, Castro V, Castro R, Gutiérrez G. La medición de la desigualdad: una estrategia metodológica, análisis de las características socioeconómicas de la muestra. Arch Invest Med 1988;19:351-360.

- No hay ningún enlace refback.