Transcript Slide 1

Implementing
Bangladesh Demographic and Health Survey
S.N. Mitra
Md. Shahidul Islam
International Workshop on Large Scale National Surveys
October 18-19, 2012
Department of Statistics, Rajshahi University
1. Introduction
The Bangladesh Demographic and Health Survey, popularly
known as BDHS, is a household based periodical survey
conducted 3-4 years apart
 Until now, six BDHSs have been implemented. The latest
BDHS was done in 2011.
BDHS is sponsored by the Ministry of Health and Family
Welfare, implemented under the authority of NIPORT with
technical support provided by ICF Macro International
Inc/USA, and is funded by USAID.
Mitra and Associates implemented all the six surveys under
contracts obtained from the ICF Macro International Inc/USA
through competitive biddings.
2. Size
BDHS is a large survey operation. It covers the whole of Bangladesh and
employs a sample of more than 10,000 households.
It uses five questionnaires: a household questionnaire, a women's
questionnaire and a men's questionnaire, a community questionnaire and
two verbal autopsy questionnaires—one for death at 0-28days, and one for
death at 29 days-59 months).
It requires about an hour to conduct an interview with a respondent.
 It is also a complex survey requiring administering of the birth history
table, contraceptive table, child immunization table, and contraceptive
calendar table.
Besides, the survey involves:
a. Measuring height and weight of under-five children and their
mothers,
b. Obtaining blood pressure measurements, anaemia testing and
glucose testing for specific age groups, as well as
c. Arsenic testing of household drinking water.
3. Objectives
For comparability, major objectives of BDHS are kept unchanged between
its different rounds. However, in every round, some new data are collected
while excluding the old data considered being no more as important. The usual
objectives of BDHS include the following, among others:
To provide up-to-date data on demographic rates, particularly fertility and
infant mortality rates at the national and subnational level;
To measure the level of contraceptive use of currently married women;
To assess the nutritional status of children under age 5, women and men
by means of anthropometric measurements (weight and height) , and to
assess infant and child feeding practices;
To provide data on maternal and child health, including antenatal care,
assistance at delivery, breastfeeding, immunizations, etc.
And to provide measures of the factors that determine the level of and
trends in the above parameters
4. Sample Design
BDHS sample is a nationally representative, probability sample of
households.
It is drawn in two stages, stratifying the country by rural and urban areas
and the divisions of the country.
A Census Enumeration Area (EA), created to have an average of 120
households, is used as the Primary Sampling Unit (PSU).
In latest BDHS (namely 2011 BDHS), 600 EAs were selected in the first
stage as the sample clusters, with 207 included from urban areas and 393
from rural areas.
EAs from each stratum are selected independently with Probability
Proportional to the EA size (PPS method).
 In order to have a sampling frame for selection of households at the
second stage, households in every selected EA (cluster) are listed, making
house-to-house visits.
Thirty households are systematically selected from a cluster to yield the
total sample for a survey.
Thus, a total sample of 18,000 households were selected in the 2011
BDHS, including 11,790 from rural areas and 6,210 from urban areas.
(NIPORT et al., 2012).
In order to provide separate estimates for both the rural and urban areas
as well as for each of the divisions, the small divisions are over sampled
compared to the other divisions, as are the urban areas compared to the
rural areas.
Thus the BDHS sample is not self-weighting. Weighting of the sample is
therefore needed to generate aggregate level estimates.
5. Respondents
Two categories of respondents are interviewed in the BDHS.
One category consists of ever-married women age 10-49, and one
category of ever-married men age 15-54.
While all ever-married women in every household included in the sample
are considered as eligible respondents to be interviewed in the survey,
interviews with ever-married men are conducted in only a sub-sample of
the sample households.
In 2011 BDHS, for example, ever-married men were interviewed in a
sub-sample of one-third of the sample households.
6. Preparation of Questionnaires
There are a number of steps involved in the preparation of the BDHS'
questionnaires.
The questionnaires are first drafted by using the MEASURE DHS Model
Questionnaires.
The drafts are then modified by the Technical Work Group (TWG) to make
the questionnaires appropriate to the Bangladesh's situations, by adding
and deleting questions as needed.
The questionnaires are prepared in English.
The implementing agency (Mitra and Associates) translates them into
Bengali.
Translations back to English are done to verify the accuracy of the
Bengali translations.
The draft questionnaires pre-tested in two urban areas and two rural
areas.
The purpose of the pretest is to check the translation, consistency and
integrity of the questionnaires.
A minimum of 200 households is interviewed in the pretest, with 50
households covered from each pretest- area.
 Based on the pretest results, the questionnaires are finalized, making the
necessary modifications by TWG, IFC Macro and Mitra and Associates.
7. Household Listing Operation
Household listing operation is carried out usually in four phases, with a
phase usually spanning over three weeks.
For example, in 2011 BDHS, initially 19 teams of listers (later reduced to
15 teams) were deployed for four months for the listing operation. A listing
team consisted of two listers.
In addition, 6 listing supervisors were deployed to check and verify the
work of the listing teams.
8. Data Collection Operation
In the BDHS, data collection work is carried out in phases, deploying
several interviewing teams.
In the 2011 BDHS, 16 interviewing teams were deployed to complete the
data collection work in five phases, with a phase spanning over three
weeks.
An interviewing team is composed of a male supervisor, a female
supervisor/editor, four female interviewers, two male interviewers, and a
logistical assistant.
Four quality control teams are employed for quality control checking of
the data. A quality control team includes a male quality control officer and a
female quality control officer.
9. Monitoring of Fieldwork
Monitoring of fieldwork is a key tool of BDHS to ensure collection of quality
data
As soon as completed questionnaires for a phase are received and
registered in the tracking register, some key information from the
questionnaire is entered into the computer.
The key information usually chosen includes such variables as household
result code, number of eligible respondents, individual result code,
respondent's age, marital status, number of children born in the last five
years, current contraceptive use status, knowledge of contraceptive
method, ever use status, etc.
The 'key information' entry is completed in 3/4 days.
The key information is then tabulated by rural-urban area, by division, by
team, and by interviewer.
The findings are shared with the interviewing teams in the debriefing
session held one/two days before the teams are sent out for the next
phase.
In the debriefing session, the results of the 'key information' tables are discussed
in detail, by team and by interviewer.
Interviewers having relatively high non-response rates are asked to explain why
they had more non-response cases than the others. The explanations are also
sought from their supervisors. The quality control officers who had checked the work
of the interviewers are also asked to give their views regarding the interviewers'
questionable performances.
Similarly, the results for every other key variable are discussed in the debriefing
session.
In the process, the survey management team identifies the supervisors and
interviewers who require further training/instructions for improvement of their
performances.
If anyone is found to have been dishonest, he/she is terminated from the service.
Benefit of the monitoring of interviewing performances is enormous. Every
interviewer, supervisor and quality control officer would know that there is a way of
checking every body's work.
Because of this awareness, most people engaged in the survey are expected to
work diligently to obtain accurate data.
10. Training
BDHS’s field personnel (listers, interviewers, supervisors and quality
control officers) are provided adequate training to prepare them for the survey
work.
Training on household listing operation for listers is conducted for 7
working days, while training on data collection operation for the field
personnel (interviewers, supervisors and quality control officers) is
conducted for three weeks.
Training consists of lectures, classroom practices, group discussions and
role-playing. At the conclusion of the training, a test is taken to see if the
trainees have achieved the knowledge and skills required to work in the
survey.
Those who cannot pass the test are not employed.
11. Data Management/Report Writing
This task is carried out in several steps.
As soon as questionnaires from the field are received at the office, they
are registered in the registration book. One registration officer is employed
to do the registration work, and store and maintain the questionnaires. He
also supplies the questionnaires for use by the data processing staff and
receives these records back when their use is over.
Data are entered and processed using the computer facility of Mitra and
Associates and by employing the statistical package program, CSPro,
supplied by ICF Macro.
After the data are cleaned, Mitra and Associates produces the tables for
the report. The clean data set and the tables are then submitted to ICF
Macro for the preparation of the report.
12. Quality of Fieldwork
High response rates are a key indicator of BDHS’s excellent field work,
and hence, of its data quality.
As shown in Table 1, out of 17,511 occupied households in the 2011
BDHS sample, 17,141 were successfully interviewed, achieving a
household response rate of 98% nationally.
Individual response rates were as high for women at 98% nationally; they
were also over 90% nationally for men, who are relatively less available at
home than women.
There were little variations in the response rates between rural and urban
areas.
Response rates were also high in the other BDHSs, upholding the
excellent fieldwork as a salient characteristic of the BDHS.
Table 1: Response Rates in 2011 BDHS
Interviews
Household interviews
Rural Areas
Urban Areas National
97.2 (11,476)
98.2 (6,035)
97.9
(17,511)
Interviews with Ever-married
women age 12-49
97.0
(6,390)
98.4
(11,832)
97.9
(18,222)
Interviews with Ever-married
men age 12-49
90.6
(1,586)
92.9
(2,757)
92.0
(4,343)
13. Reliability of Estimates
Estimates from the BDHS were generally found to be statistically reliable.
For example, as shown in Table 2, in the 2007 BDHS, the TFR (Total
Fertility Rate) per women was estimated as 2.710 within error margins of
+/- 0.122 at the 95% confidence level in the national sample;
as 2.805 within error margins of +/- 0.154 in the rural sample and as
2.395 within error margins of +/- 0.134 in the urban sample;
the CPR (Contraceptive Prevalence Rate) per 100 currently married
women was estimated as 55.8 within error margins of +/- 1.60 at the 95%
confidence level in the national sample;
as 54.0 within error margins of +/- 2.00 in the rural sample and as 62.0
within error margins of +/-2.40 in the urban sample;
Similar evidence of stability was apparent for the estimates from the other
BDHS surveys.
Table 2: Sampling Errors of the Total Fertility Rate (TFR) and
Contraceptive Prevalence Rate, in BDHS 2007
Parameter
Estimate
Standard
Error (SE)
95% Confidence Limits
Lower Bound Upper Bound
TFR
National
Rural
Urban
2.710
2.805
2.395
0.061
0.077
0.067
2.589
2.650
2.260
2.832
2.959
2.530
CPR
National
Rural
Urban
0.558
0.540
0.620
0.008
0.010
0.012
0.542
0.520
0.596
0.574
0.560
0.643
Another evidence of reliability of BDHS estimates was apparent in the
comparison of the population age distribution between the two consecutive
surveys.
 With the fertility and mortality declining, the age distribution changes, but
the changes, being gradual and small, are unlikely to make the distribution
discernibly different between the two consecutive surveys done 3-4 years
apart.
Thus, the similarity of the age distribution obtained from the 2005 BDHS
survey with that from the 2009 BDHS survey, shown in Figure1, confirms
the reliability of the BDHS estimates.
Figure 1: Comparison of the Population Age distribution between
BDHS 2004 and BDHS 2007
14
12.8 12.8
12.6
11.9
11.7
12
10
11.3
10.8
9.1
8.7
8
7.9
7.5
6
6.8
6.4
6.5
5.9
7.1
6.6
5.3
5.1
4
2
4.5
4.4
3.7
3.0
2.4
0
<5
5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60+
BDHS 2004
BDHS 2007
Reliability of BDHS estimates was also apparent in the comparison of the
TFR and CPR estimates at the divisional level;
As shown in Figure, in the 2011 BDHS, among the divisions of the
country, the lowest TFR at 2.0-2.4 was associated with the highest CPR per
100 women at 63-66 in the Rajshahi and Khulna divisions;
and the highest TFR at 3.7 associated with the lowest CPR at 32 in the
Sylhet division;
Again, the Barisal and Dhaka divisions, both, had the same TFR at 2.8,
with both having the same CPR at 56;
Contraceptive use is a major determinant of fertility in Bangladesh. Thus,
the fertility and contraceptive data were found internally consistent in the
2011 BDHS showing higher TFR in a division where CPR was lower or vice
versa;
Attesting to the credibility of the data, such internal consistencies were
also notable in the other BDHS surveys.
Figure 2: Comparison of Total Fertility Rate and Contraceptive
Prevalence Rate by Division in the BDHS 2007
80
60
63.1
65.9
56.3
56.4
43.9
40
20
20.0
24.0
28.0
28.0
32.0
37.0
31.5
0
Khulna
Rajshahi
Total Fertility Rate
Barisal
Dhaka
Chittagong
Sylhet
Contraceptive Prevalence Rate
13. Conclusion
Professionally implemented with utmost care, BDHSs have emerged as
credible sources of population and health data for policy makers and
program managers (MOHFW, 2011).
If BDHS’s experiences are further refined and used, survey taking
capabilities would doubtless improve in the country.
Thanks