|
Nutritional Surveillance Project, Bangladesh
Since its inception in 1990,
the Nutritional Surveillance Project (NSP) of Helen Keller International (HKI)
and the Institute of Public Health Nutrition (IPHN), Government of
Bangladesh, has been an independent source of high-quality data on the
health and nutrition of children and women in Bangladesh. The NSP was
originally designed to monitor the impact of disasters and the
effectiveness of relief and rehabilitation programs in disaster-prone
areas of Bangladesh. While this role has continued, information generated
by the NSP over the last 16
years has helped inform policymakers, program managers and donor
organizations on many development concerns in the country, including
health, nutrition, food security, homestead food production, gender
disparities, and rural and urban poverty.
The NSP collects information
from children aged less than five years, their mothers and their households
throughout rural Bangladesh and in the urban slums of the six largest cities
in the country. These data are collected by IPHN and
NGO partners with
supervision, quality control and data analysis by HKI. Data collection takes
place every two months to capture seasonal changes in nutrition and health.
This allows the impact of disasters, programs and policies to be
distinguished from seasonal effects. It also makes the NSP better prepared
to assess the impact of a disaster or other crisis event because data are
collected a maximum of two months before and after an event and therefore
the magnitude of any change can be more accurately assessed. NSP data
collected on over
930,000 households
in the last 16 years can be used to monitor long-term trends, and to measure
the impact of programs and policies against any underlying secular trends.
|
NSP data collection
|
|
NSP data collection sites
The
NSP currently collects data in 28 rural sub-districts (upazila), four in
each of the six divisions as well as in the
Chittagong Hill Tracts, and in urban slums in Barisal, Chittagong,
Dhaka, Khulna, Rajshahi and Sylhet. See the map of the
NSP data collection
sites. |
|
Frequency and timing of data
collection
Since its inception in 1990, the NSP has conducted
‘rounds’ of data collection every two months in February, April, June,
August, October and December. The rounds are timed to coincide with the
six seasons in Bangladesh. Each round of data collection takes seven to
eight weeks to complete. The rounds are numbered consecutively,
beginning with the first round in April 1990. The months of data
collection and the corresponding season for the rounds in 2005
are given in Table 1. |
|
|
|
Table 1. Months of data
collection and the corresponding season for the six rounds of NSP data
collection in 2005 |
|
|
Round of data collection
|
Months of data collection
|
Name of season
|
|
|
90
|
February - March
|
Spring (Bashonto)
|
|
|
91
|
April - May
|
Summer (Grishmo)
|
|
|
92
|
June - July
|
Rainy season (Borsha)
|
|
|
93
|
August - September
|
Pre-Autumn (Sharot)
|
|
|
94
|
October - November
|
Late Autumn (Hemonto)
|
|
|
95
|
December - January
|
Winter (Sheeth)
|
|
|
Sampling
For each
round of data collection, a new sample of 10,500 households in rural
Bangladesh and 1,500 households in the urban slum sites is selected. A
household is eligible for inclusion if it contains at least one
physically abled child aged less than 5 years and if the mother is
present. Sampling strategies differ for the rural and urban slum
samples.
Rural Bangladesh
The current stratified
multi-stage cluster sampling design was introduced in February 1998 and
provides data that are statistically representative at the divisional
and national levels. Data are collected from four sub-districts in each
of the six divisions of the country as well as the Chittagong Hill
Tracts (CHT). The divisional sub-districts were randomly selected when
the sampling design was introduced, the CHT sub-districts were newly
selected in 2003. NSP sub-districts remain the same from round to round.
For each round of data collection, 15 mauza (administrative unit
within a sub-district) are randomly selected from each of the four
sub-districts. One village is randomly selected from each mauza,
and 25 households are systematically sampled from each village.
Therefore the total number of households in each round of data
collection is:
28 sub-districts * 15 mauza * 25 hh = 10,500 households.
Urban slums
Households are
selected from NGO working areas in the urban slums of six cities:
Barisal (Wards 2, 6, 24), Chittagong (Wards 2, 11, 18, 28, 39, Wards 21
& 28 during Feb-May (rounds 90, 91), Wards 7 & 19 from June onwards
rounds 92-95), Dhaka (Wards 1, 12, 16, 22, 27, 31, 34, 43, 48, 51, 58,
65), Khulna (Wards 3, 16, 27), Rajshahi (Wards 2, 19, 23) and Sylhet
(Wards 8, 14, 27). Households were selected using a two-stage cluster
sampling design. Slums were selected from each urban slum site using
simple random sampling, and households were systematically sampled from
each slum. Overall, 1500 households were selected from the six
divisional cities: Barisal (150), Chittagong (300), Dhaka (600) Khulna
(150), Rajshahi (150) and Sylhet (150). |
|
Data collected
In each selected household, the weight and
height/length of one mother and all her children below 5 years are
measured. A precoded questionnaire is used to record the anthropometric
measurements, together with other information on the health and
nutrition of the mother and her children, household demography and
socio-economic status, homestead food production and household food
consumption.
The rural and urban slum questionnaires and
codebooks provide further information on how the data that are shared on
this CD-ROM were collected, including how the questions were asked and
how the responses were coded. |
| |
|
Rural
questionnaire |
Urban slum questionnaire |
|
Rural
codebook |
Urban
slum codebook |
|
Field team
The NSP works with a well-established local
NGO in each rural sub-district and urban slum site. The NGO employs the
field teams, which are responsible for selecting households according to
the sampling design and for conducting the interviews. Each field team
comprises two Data Collection Officers, one of whom is designated the
Field Team Leader, and at least one of whom is a woman. The NGO also
employs a Field Coordinator, who supervises data collection and provides
administrative and logistic support to the field teams. |
|
Field supervision and
quality control
The NSP has a system of
field supervision and quality control to ensure that good quality data
are collected. Field Monitoring Officers supervise the field teams
during each round of data collection, and provide support to the NGO
partners in discussions with local officials. Quality Control Officers
re-visit 5-10% of households without prior notice on the day following
the data collection by the field teams and recollect data on selected
indicators, including anthropometric measurements. Data collected by the
Quality Control Officers are later compared with the data collected by
the field teams to assess the accuracy of the data. Field teams who
perform unsatisfactorily are either re-trained or dismissed, depending
on any mitigating reasons and the seriousness of the situation.
|
|
Data entry and management
Each partner NGO employs
at least one Data Entry Operator who enters the raw data from the
questionnaires using a standard data entry package designed by HKI. The
data files from all 24 rural sub-districts and six urban slum sites are
merged at the HKI/NSP Data Management Unit in Dhaka. A Microsoft FoxPro®
program is run to perform a series of checks for invalid data, which are
then checked against the questionnaires before editing. The
questionnaires are stored at HKI in Dhaka for at least 5 years. |
|
Training
All field teams, Data
Entry Operators, Field Monitoring Officers and Quality Control Officers
are carefully selected and trained before they conduct their work. HKI
provides two weeks of initial basic training to all these staff and
refresher training for 2-3 days before each round of data collection.
The refresher training allows the NGO staff to interact with HKI staff,
share their experiences, and discuss problems encountered in the field.
During these sessions any problems with data quality are shared and,
where necessary, the sources of problems are identified and resolved.
Special training is given when new questions are added or the
questionnaire is modified. Through this interactive training process,
HKI helps to develop the field staff’s understanding of the surveillance
system. Field teams have to demonstrate good performance during
reliability tests that are performed by each field staff at least once
per year.
Exploring the rural data
|
Guidelines for analyzing the data
The data collected by the NSP in rural
Bangladesh in 2005 are provided in two file types, an SPSS data file
and a DBF data
file. The SPSS data file needs to be used with SPSS for Windows
software. The DBF data file can be read by most statistical
packages, including EPI-INFO, which can be downloaded free of charge
from the internet (www.cdc.gov/epiinfo/). |
|
The data
file contains data collected on all children aged less than five
years, their mothers and their households in rural Bangladesh by the
NSP during six rounds of data collection between February 2005 and
January 2006. For information on how the data were collected go to
NSP data collection.
Each row
in the data file provides all the data collected on a single child,
together with the data collected on his/her mother and his/her
household. Each child is therefore a single case in the data file.
The cases are sorted by round of data collection [ROUND], division
[DIVISION], upazila [UPAZILA], mauza [MAUZA],
household number [HHNO] and the identification number of each child
[CID]. Each child is identified by a unique set of values for these
six variables.
Each
column in the data file provides all the data collected on a single
variable. Most of the variables in the data file represent a single
question in the questionnaire. For example, the variable [SDW]
represents all data collected on the question ‘Where does your
household get drinking water?’ Some variables have been computed
from other variables. For example, the variable for child age [CAGE]
was determined by finding the difference in months between a child’s
date of birth and the date of visit to the household.
See the
rural codebook for information on the variables, including the
variable names, variable labels, categories, missing values and
notes on how the data were collected. |
The NSP sampling design (see NSP data
collection) provides statistics that are representative at the
divisional and national level for rural Bangladesh.
A weighting factor
is applied to the data to produce national rural statistics for all
children aged less than five years, mothers or households. This
weighting factor is needed because the NSP samples the same number of
households from each division in Bangladesh but each division has a
different number of rural households. The rural dataset contains the
variable [DWEIGHT] which is used to calculate national rural statistics.
The weighting factor was calculated using data on the number of rural
dwelling households by division taken from the 1991 population census
(see Table 1). A more recent population census was conducted in 2001 but
this data has not yet been published. Some statistical programs,
including SPSS, have an option to weight data during statistical
analysis. If this option is not available, the weighted statistics can
be calculated manually using the values provided in Table 2. No
weighting factors are needed to calculate divisional rural statistics.
Table 1. The number of rural dwelling
households by division in Bangladesh adjusted for undercounting.
|
|
Number of rural
dwelling households
|
|
Barisal
|
|
|
Chittagong
|
2,711,051
|
|
Chittagong Hill
Tracts |
128,924
|
|
Dhaka |
|
|
Khulna
|
1,910,033
|
|
Rajshahi
|
4,366,406
|
|
Sylhet
|
1,043,762
|
|
Total |
15,870,391
|
Source:
1991 population census, BBS.
Table 2. Design weights for division by
round of NSP data collection in 2005.
|
|
Barisal |
Chittagong |
Chittagong Hill Tracts |
Dhaka |
Khulna |
Rajshahi |
Sylhet |
|
Feb 2005 |
0.53970 |
1.19566 |
0.05686 |
1.97868 |
0.84238 |
1.92572 |
0.46064 |
|
Apr 2005 |
0.53970 |
1.19566 |
0.05686 |
1.97868 |
0.84238 |
1.92572 |
0.46064 |
|
Jun 2005 |
0.53975 |
1.19577 |
0.05686 |
1.97887 |
0.84246 |
1.92590 |
0.46038 |
|
Aug 2005 |
0.53975 |
1.19577 |
0.05686 |
1.97887 |
0.84246 |
1.92590 |
0.46038 |
|
Oct 2005 |
0.53975 |
1.19577 |
0.05686 |
1.97887 |
0.84246 |
1.92590 |
0.46038 |
|
Dec 2005 |
0.53975 |
1.19577 |
0.05686 |
1.97887 |
0.84246 |
1.92590 |
0.46038 |
|
The NSP collects data on all children
aged less than 5 years belonging to one mother in the selected
households. If you wish to analyze data from only one child per
mother/household, select cases that satisfy the condition [CID]=1.
Note that this child is the mother’s oldest child aged less than 60
months.
Indicators of a child’s nutritional
status
Key indicators of a child’s nutritional
status that can be determined using the NSP data provided on this
CD-ROM are described below.
Malnutrition: |
|
|
The Z-scores of
weight-for-age, height-for-age and weight-for-height have been
calculated using NCHS growth references. The World Health
Organization (WHO) recommends the following indicators of child
malnutrition, based on these Z-scores: |
| |
|
|
Underweight
(weight for age Z-score)
|
|
Underweight
|
[CWAZ] <-2 SD
|
|
Moderate
underweight |
[CWAZ] -3 to <-2
SD |
|
Severe
underweight |
[CWAZ] <-3 SD
|
|
|
|
|
Stunting (height
for age z-score)
|
|
Stunting
|
[CHAZ] <-2 SD
|
|
Moderate
stunting |
[CHAZ] -3 to <-2
SD |
|
Severe stunting
|
[CHAZ] <-3 SD
|
|
|
|
|
Wasting (weight
for height z-score)
|
|
Wasting
|
[CWHZ] <-2 SD
|
|
Moderate
wasting |
[CWHZ] -3 to <-2
SD |
|
Severe wasting
|
[CWHZ] <-3 SD
|
|
Source:
WHO (1995). Physical Status: the Use and Interpretation of
Anthropometry. WHO Technical Report Series 854. WHO, Geneva. |
|
Mothers who have more than one child
aged less than five years occur more than once in the dataset. When
analyzing data collected from mothers it is important that each
mother is selected only once using [CID]=1.
Indicators
of a mother’s nutritional status
Key indicators of a mother’s
nutritional status that can be determined using the NSP data
provided on this CD-ROM are described below. For all of these
indicators, it is important to consider the pregnancy status of the
mothers. Non-pregnant mothers can be selected for analysis using [MPREG]=0
and pregnant mothers can be selected using [MPREG]>0.The variable [MBMI]
identifies maternal nutritional status according to internationally
agreed BMI-based categories.
Chronic
Energy Deficiency (CED):
The indicator
for CED in non-pregnant women is body mass index (BMI), which is
calculated by dividing a mother’s weight in kilograms by the square
of a mother’s height in meters: [MWT]/[MHT/100]2. The
different grades of mother’s CED are defined as follows: |
|
Mild
CED |
BMI 17.0 - <18.5 kg/m2 |
|
Moderate CED |
BMI 16.0 - <17.0 kg/m2 |
|
Severe CED |
BMI
<16.0 kg/m2
|
|
Source: WHO (1995). Physical
Status: the Use and Interpretation of Anthropometry. WHO
Technical Report Series 854. WHO, Geneva. |
|
A household occurs more than once in
the dataset if the selected mother has more than one child aged less
than five years. When analyzing data collected at the household
level it is important that each household is selected only once
using [CID]=1. |
|
The data collected by the NSP in urban
slums in Bangladesh in 2005 are provided in two file types, an SPSS
data file and a
DBF data file. The SPSS data file needs to be used with SPSS for
Windows software. The DBF data file can be read by most statistical
packages, including EPI-INFO, which can be downloaded free of charge
from the internet (www.cdc.gov/epiinfo/).
|
|
The
data file contains data collected on all children aged less than 5
years, their mothers and their households in the urban slum sites in
Barisal, Dhaka, Chittagong, Khulna, Rajshahi and Sylhet by the NSP
during six rounds of data collection between February 2005 and
January 2006. For information on how the data were collected go to
NSP data collection.
Each row in the data file provides all
the data collected on a single child, together with the data
collected on his/her mother and his/her household. Each child is
therefore a single case in the data file. The cases are sorted by
round of data collection [ROUND], urban slum site [CITY], urban slum
[SLUM], household number [HHNO] and the identification number of
each child [CID]. Each child is identified by a unique set of values
for these five variables.
Each column in the data file provides
all the data collected on a single variable. Most of the variables
in the data file represent a single question in the questionnaire.
For example, the variable [SDW] represents all data collected on the
question ‘Where does your household get drinking water?’ Some
variables have been computed from other variables. For example, the
variable for child age [CAGE] was determined by finding the
difference in months between a child’s date of birth and the date of
visit to the household.
See the urban slum codebook for
information on the variables, including the variable names, variable
labels, categories, missing values and notes on how the data were
collected. |
|
The six urban slum sites
in the NSP urban sample are not randomly selected from all urban
slums in Bangladesh (see
NSP data collection
for the sampling design) and so the data
are not nationally representative of all urban slums in Bangladesh.
As there are considerable differences in living conditions, health
and nutrition between urban slums in Bangladesh, the data collected
from each urban slum site should be analyzed separately and not
aggregated with the data from the other urban slum sites. |
|
The NSP collects data on all children
aged less than five years belonging to one mother in the selected
households. If you wish to analyze data from only one child per
mother/household, select cases that satisfy the condition [CID]=1.
Note that this child is the mother’s oldest child aged less than 60
months.
Indicators of a child’s nutritional
status
Key indicators of a child’s nutritional
status that can be determined using the NSP data provided on this
CD-ROM are described below.
Malnutrition: |
|
|
The Z-scores of
weight-for-age, height-for-age and weight-for-height have been
calculated using NCHS growth references. The World Health
Organization (WHO) recommends the following indicators of child
malnutrition, based on these Z-scores: |
| |
|
|
Underweight
(weight for age Z-score)
|
|
Underweight
|
[CWAZ] <-2 SD
|
|
Moderate
underweight |
[CWAZ] -3 to <-2
SD |
|
Severe
underweight |
[CWAZ] <-3 SD
|
|
|
|
|
Stunting (height
for age z-score)
|
|
Stunting
|
[CHAZ] <-2 SD
|
|
Moderate
stunting |
[CHAZ] -3 to <-2
SD |
|
Severe stunting
|
[CHAZ] <-3 SD
|
|
|
|
|
Wasting (weight
for height z-score)
|
|
Wasting
|
[CWHZ] <-2 SD
|
|
Moderate
wasting |
[CWHZ] -3 to <-2
SD |
|
Severe wasting
|
[CWHZ] <-3 SD
|
|
Source:
WHO (1995). Physical Status: the Use and Interpretation of
Anthropometry. WHO Technical Report Series 854. WHO, Geneva. |
|
Mothers who have more than one child
aged less than 5 years occur more than once in the dataset. When
analyzing data collected from mothers it is important that each
mother is selected only once using [CID]=1.
Indicators of a mother’s nutritional
status
Key indicators of a mother’s nutritional
status that can be determined using the NSP data provided on this
CD-ROM are described below. For all of these indicators, it is
important to consider the pregnancy status of the mothers.
Non-pregnant mothers can be selected for analysis using [MPREG]=0
and pregnant mothers can be selected using [MPREG]>0.The variable [MBMI]
identifies maternal nutritional status according to internationally
agreed BMI-based categories.
Chronic
Energy Deficiency (CED):
The indicator
for CED in non-pregnant women is body mass index (BMI), which is
calculated by dividing a mother’s weight in kilograms by the square
of a mother’s height in meters: [MWT]/[MHT/100]2. The
different grades of mother’s CED are defined as follows: |
|
Mild
CED |
BMI 17.0 - <18.5 kg/m2 |
|
Moderate CED |
BMI 16.0 - <17.0 kg/m2 |
|
Severe CED |
BMI
<16.0 kg/m2
|
|
Source: WHO (1995). Physical
Status: the Use and Interpretation of Anthropometry. WHO
Technical Report Series 854. WHO, Geneva. |
|
A household occurs more than once in the
dataset if the selected mother has more than one child aged less
than five years. When analyzing data collected at the household
level it is important that each household is selected only once
using [CID]=1.
|
|
|
