PHC  MANAGEMENT  ADVANCEMENT PROGRAMME



ASSESSING  COMMUNITY  HEALTH  NEEDS
AND  COVERAGE




MODULE  2:
USER'S GUIDE




USER'S  GUIDE




Dr. Jack Reynolds

ISBN:  1-882839-01-3
Library of Congress Catalog Number: 92-75462

Monitoring the health of children, such as this child 
in Portugal, helps to identify health needs and, later, assess
coverage 
of children with priority services
Photo by Jean-Luc Ray for AKF



Dedicated toDr. Duane L. Smith (1939-1992),Dr. William B. Steeler
(1948-1992)and all other health leaders, managers and workers who
follow their example in the effort to bring quality health care
to all in need.
Assessing the awareness of women about good nutrition 
and the treatment of common problems, such as
diarrhoea,  is important
Photo by Jean-Luc Ray for AKFAn overview of PHC MAP
 The main purpose of the Primary Health Care Management
Advancement Programme (PHC MAP)  is to help PHC management teams
collect, process and analyse useful management information<P9BI>. 

Initiated by the Aga Khan Foundation, PHC MAP is a collaborative
programme of the Aga Khan Health Network1 and PRICOR.2  An
experienced design team and equally experienced PHC practitioner
teams in several countries, including Bangladesh, Chile,
Colombia, the Dominican Republic, Guatemala, Haiti, India,
Indonesia, Kenya, Pakistan, Senegal, Thailand and Zaire, have
worked together to develop, test and refine the  PHC MAP
materials to make sure that they are understandable, easy to use
and helpful.
PHC MAP includes nine units called modules. These modules focus
on essential information that is needed in the traditional
management cycle of planning-doing-evaluating.  The relationship
between the modules and this cycle is illustrated below.

PHC MAP modules and the planning-evaluation cycle

PHC MAP MODULES
1. Information needs
2. Community needs
3. Work planning
4. Surveillance
5. Monitoring indicators
6. Service quality
7. Management  quality
8. Cost analysis
9. Sustainability



1 The Aga Khan Health Network  includes the Aga Khan Foundation, 
the Aga Khan Health    Services, and the Aga Khan University, all
of which are involved in the strengthening of primary health care

2 Primary Health Care Operations Research is a worldwide project 
of the Center for Human Services, funded by  the United States
Agency for International Development
Managers can easily adapt these tools to fit local conditions. 
Both new and experienced programmers can use them. Government and
NGO managers, management teams, and communities can all use the
modules to gather information that fits their needs.  Each module
explains how to collect, process and interpret information that
managers can use to improve planning and monitoring.  The modules
include User's guides, sample data collecting and data processing
instruments, optional computer programs, and Facilitator's
guides, for those who want to hold training workshops.
The health and management services included in PHC MAP are listed
below. 

OTHER HEALTH CARE
Water supply, hygiene and 
  sanitation
School health
Childhood disabilities 
Accidents and injuries 
Sexually transmitted diseases
HIV/AIDS
Malaria
Tuberculosis
Treatment of minor ailments
Chronic, non-communicable diseases

HEALTH SERVICES

Planning
Personnel management
Training
Supervision
Financial management
Logistics management
Information management
Community  organisation

MANAGEMENT      SERVICES
Health and management services

GENERAL        
PHC household visits          
Health education      
          
MATERNAL CARE       
Antenatal care      
Safe delivery       
Postnatal care      
Family planning     
               

CHILD CARE
Breastfeeding      
Growth monitoring   
Nutrition education 
Immunization        
Acute respiratory infection   
Diarrhoeal disease control     
Oral rehydration therapy

Several Manager's guides supplement these modules.  These are:
Better Management: 100 Tips, a helpful hints book that describes
effective ways to help managers improve what they do;
Problem-solving, a guide to help managers deal with common
problems; Computers, a guidebook providing useful  hints  on
buying  and operating computers, printers, other hardware and
software; and The computerised  PRICOR thesaurus, a compendium of
PHC indicators.  
Women carry the burden of most family chores and of 
promoting health and preventing disease within their
 families and communities
Photo by Jean Luc Ray for AKFThe Primary Health Care Management Advancement Programme has been
funded by the Aga Khan Foundation Canada, the Aga Khan Foundation
USA, the Aga Khan Foundation's head office in Geneva, the
Rockefeller Foundation, the Canadian International Development
Agency, Alberta Aid, and the United States Agency for
International Development under two matching grants to AKF USA. 
The first of these grants was "Strengthening the Management,
Monitoring and Evaluation of PHC Programmes in Selected Countries
of Asia and Africa" (co-operative agreement no.
OTR-0158-A-00-8161-00, 1988-1991); and the second was
"Strengthening the Effectiveness, Management and Sustainability
of PHC/Mother and Child Survival Programmes in Asia and Africa"
(co-operative agreement no. PCD-0158-A-00-1102-00, 1991-1994). 
The development of Modules 6 and 7 was partially funded through
in-kind contributions from the Primary Health Care Operations
Research project (PRICOR) of the Centre for Human Services under
its co-operative agreement with USAID (DSPE-6920-A-00-1048-00).

This support is gratefully acknowledged.  The views and opinions
expressed in the PHC MAP materials are those of the authors and
do not necessarily reflect those of the donors.

All PHC MAP material (written and computer files) is in the
public domain and may be freely copied and distributed to others.
Contents
QUICK START1
INTRODUCTION3
 What are "rapid community surveys"3
 Some limitations of rapid surveys 6
 How you can use this module10
 How the module is organized11
RAPID SURVEY PROCEDURES15
 Step 1: Specify the objectives of the rapid survey15
 Step 2:Decide what indicators to use19
 Step 3:Develop an outline for the survey report 21
 Step 4:Design the data collection instruments23
 Step 5:Develop the sampling procedures30
 Step 6:Schedule the survey41
 Step 7: Collect the data41
 Step 8:Enter, verify, and tabulate the data43
 Step 9:Analyse, interpret, and report the findings 47
 Step 10:Develop an action plan 50
APPENDICES: TEMPLATES, TOOLS, GUIDELINES, AND COMPUTER PROGRAMS53
 A.How to use Epi Info for rapid surveys  55
 B.Questionnaire design guidelines 81
 C.Rapid survey instruments 87
 D.Cluster survey registers147
 E.Guidelines for training and supervising interviewers183
 F.Cluster sampling programmes cluster identification
worksheet187
 G.Other sampling tools191
 G.1 Estimates of target group sizes191
 G.2 Sample size estimation for WHO two-stage cluster      
survey193
 G.3 Random number table195
 G.4 Random sampling procedures197
 G.5 Estimating fertility and child mortality rates and ratios199
 H.  Survey management forms203
 I.   Tabulation and analysis templates207
 I.1. Analysis plan207
 I.2. Data entry and analysis templates211
 I.3. Confidence interval estimation templates221
REFERENCES AND BIBLIOGRAPHY223
ACRONYMS AND ABBREVIATIONS 225
GLOSSARY226

Acknowledgements
This module draws heavily on concepts and procedures used by the
World Health Organization in its rapid assessments of
immunization coverage, the recent work of Ralph R. Frerichs of
the UCLA School of Public Health on rapid surveys, and Anthony
Bennett's work on mini-surveys.  Bennett and Peerasit
Kamnuansilpa (both in Thailand) have been instrumental in early
field tests of prototypes of rapid surveys in PHC, and their
experience has been incorporated into this module.  Early
versions of the rapid survey instruments were reviewed by members
of the Aga Khan Health Network at a workshop in Dhaka,
Bangladesh, in May 1990.  Subsequent feedback from Khatidja
Husein of the Aga Khan University (Pakistan) and Esther Sempebwa
of the Kisumu Primary Health Care Project (Kenya) were especially
helpful.
Thanks are due also to Pierre Clauquin of AKF for his suggestions
on computer programs, and to Paul Richardson of URC/CHS for his
advice on sampling.
The computerised version of this module is based on, Epi Info,
Version 5: A Word Processing, Database and Statistics System for
Epidemiology on Microcomputers, by A.G. Dean, J.A. Dean, A.H.
Burton, and R.C. Dickers. Epi Info is a joint project of the
Centers for Disease Control (CDC) and the World Health
Organization (WHO).  The Epi Info manual and computer program are
in the public domain and may be freely copied.  The program is
"bundled" with this module to enable managers to use the computer
version of the module.

Reviewers:
Donald  Belcher Veterans' Administration, Seattle, Washington USA
William Reinke Johns Hopkins School of Hygiene and Public Health, 
Johns Hopkins University, Baltimore, Maryland USA
Gilbert M. Burnham Johns Hopkins School of Hygiene and Public
Health, Johns Hopkins University, Baltimore, Maryland, USA
David H. Peters Johns Hopkins School of Hygiene and Public
Health, Johns Hopkins University, Baltimore, Maryland, USA
Paul Zeitz. 
Johns Hopkins School of Hygiene and Public Health, Johns Hopkins
University, Baltimore, Maryland USA
Al Henn
AMREF, Nairobi, Kenya
Asif Aslam, Melvyn Lobo, Inaam-ul-Haq, Syed Meboob Ali Shah,
Khatidja Husein Aga Khan University, Karachi
Barkat-e-Khuda URC, Bangladesh


Field tests: , Countries,, participating organisations,, field
test facilitators
Pakistan:, Aga Khan University,, Karachi; Facilitator:  Khatidja
Husein,, Aga Khan University
Thailand:, Ministry of Public Health,, Srisaket; Somboon
Vachrotai Foundation (SVF); ASEAN Institute for Health
Development (AIHD); Health and Population Research Company
(HPRC); Facilitators: Pearasit Kamnuansilpa,, HPRC,, Thongchai
Sapanuchart,, SVF,, Butsabar Subongkot,, HPRC
India:, Junagadh PHC Project,, Jonpur; Sidhpur PHC Project,,
Ahmedabad; Aga Khan Health Service,, India; URMUL Trust PHC
Project
Kenya:, Mombasa PHC Programme; Kisumu PHC Programme;
Facilitators: Paul Richardson,, URC,, Esther Sempebwa,, Mombasa
PHC Programme
Columbia:, Fundacion Santa Fe de Bogota; Facilitator: Jorge E.
Medina

Quick start
How to use the prototype questionnaires - rapid surveys
 Select a questionnaire
If you already know something about community surveys, you can
use the questionnaires in this manual to carry out a PHC survey. 
Go to Appendix C and select a questionnaire.  Modify it to fit
your particular needs and situation.  If you do not need to make
any changes to the questionnaire, simply print out as many copies
as you need, one per interview.
 Draw your sample
You also need to draw your sample.  The User's guide includes
instructions for drawing a cluster sample (30 clusters by seven
respondents each).  This is normally a large enough sample for a
rapid survey.  These instructions show you how to do this
manually. You do not need a computer.  However, if you have a
computer, see Appendix F for instructions.  It describes how to
use the enclosed computer program to draw your sample quickly.
 Collect the data
We are assuming that you will interview seven eligible mothers in
each of 30 clusters.  It is best to send a two-person team to
each cluster - one to conduct the interviews, the other to find
the respondents and check the completed interviews.  You only
need to complete seven interviews in each cluster.  See Step 7 in
the User's guide if you need more information about selecting
households.
 Summarise the data
Assemble all of your questionnaires. Then calculate the totals
for each question and enter them into a blank questionnaire. this
will give you a summary of the entire survey on one sheet.
 Analyse the data
Compute percentages for each "yes" answer, and summarise the
results.  Write out the major findings for each question.  See
the analysis section in the User's guide for analysis
instructions.
Computerised surveys
If you would like to use a computer to conduct your survey, turn
to Appendix A, which describes how to use Epi Info to design,
collect, and analyse your survey data.  A copy of the Epi Info
manual and the Epi Info computer program is included in this
module for your use.


In Bangladesh, the empowerment of women for social 
and economic development has yielded substantial improvements in
the health status of their families
Photo by Jean-Luc Ray for AKFPHC MAP MANAGEMENT COMMITTEE
Dr. Ronald Wilson Aga Khan Foundation, Switzerland (Co-Chair)
Dr. Jack Bryant Aga Khan University, Pakistan (Co-Chair)
Dr. William Steeler Secretariat of His Highness the Aga Khan,
France (Co-Chair)
Dr. Jack Reynolds Center for Human Services, USA (PHC MAP
Director)
Dr. David Nicholas Center for Human Services, USA
Dr. Duane Smith Aga Khan Foundation, Switzerland
Dr. Pierre Claquin Aga Khan Foundation, Switzerland
Mr. Aziz Currimbhoy Aga Khan Health Service, Pakistan
Mr. Kabir Mitha Aga Khan Health Service, India
Dr. Nizar Verjee Aga Khan Health Service, Kenya
Ms. Khatidja Husein Aga Khan University, Pakistan
Dr. Sadia Chowdhury Aga Khan Community Health Programme,
Bangladesh
Dr. Mizan Siddiqi Aga Khan Community Health Programme, Bangladesh
Dr. Krasae Chanawongse ASEAN Institute for Health Development,
Thailand
Dr. Yawarat Porapakkham ASEAN Institute for Health Development,
Thailand
Dr. Jumroon Mikhanorn Somboon Vacharotai Foundation, Thailand
Dr. Nirmala Murthy Foundation for Research in Health Systems,
India
PHC MAP TECHNICAL ADVISORY COMMITTEE
Dr. Nirmala Murthy 
Foundation for Research in Health Systems, India (Chair)
Dr. Krasae Chanawongse ASEAN Institute for Health Development,
Thailand
Dr. Al Henn African Medical and Research Foundation (AMREF),
formerly of the Harvard Institute for International Development
Dr. Siraj-ul Haque Mahmud Ministry of Planning, Pakistan
Dr. Peter Tugwell Faculty of Medicine, University of Ottawa,
Canada
Dr. Dan Kaseje  Christian Medical Commission, Switzerland,
formerly of the  University of Nairobi, Kenya
KEY PHC MAP STAFF AT THE CENTER FOR HUMAN SERVICES
Dr. Jack Reynolds (PHC MAP Director)Dr. Neeraj Kak
Dr. Paul RichardsonMs. Lori DiPrete Brown
Dr. David NicholasMs. Pam Homan
Dr. Wayne StinsonDr. Lynne Miller Franco
Ms. Maria FranciscoMs. Mary Millar

Introduction
This module introduces and shows you how to use rapid community
surveys to assess community PHC needs and to evaluate programme
effects on PHC coverage.
 What are "rapid community surveys"?
PHC managers need timely and useful information about the health
status of their target populations so that they can do a better
job of planning and monitoring PHC services.  The traditional way
to collect health information is through large surveys, done
infrequently and generally on a national scale.  This information
is of little use to local managers.
"Local programme managers often totally lack data upon which to
assess or evaluate the health problems in their area. It is
usually not possible to interpolate the results of large general
population surveys (for) local estimates."

<$FSmith, G. S. "Development of rapid epidemiological assessment
methods to evaluate health status and delivery of health
services." International journal of epidemiology 18 (4
[Supplement] ) S4, 1989.>

The "rapid survey" is a new tool for getting this kind of
information quickly and inexpensively.  It is especially useful
for local PHC managers who need information about their local
populations.
The "rapid survey" is an alternative to traditional large-scale
sample surveys. It was originally developed to assess
immunization coverage

These surveys are designed to help PHC managers collect
population-based information on health status, behaviour, and
knowledge. The typical rapid survey can be carried out in two to
three weeks, from design to final report. It involves 200-300
household interviews, drawn from 30 clusters of seven to ten
respondents each. The interview schedule is short (20-30 items),
and the questions are phrased in "yes"/"no" terms to permit
statistical tests of significance. The surveys are often
pre-coded so that the data can be entered into a local computer
and immediately analysed. The analysis is simple.

Example
When they are done well, rapid surveys are very impressive. The
author participated in his first rapid survey in  Thailand in
late 1987. Twenty two participants from three developing
countries attended a one-week workshop where they learned the
principles of rapid surveys in the first three days and designed
questionnaires and developed the samples in the next two days for
two simultaneous surveys - one on antenatal care, and one on
family planning. In the second week they collected the data
(three days), processed it (one day), and presented a report (one
day) on the findings to the provincial management team. Excerpts
from the presentation are shown below.

<$FHenderson, R. H. and T. Sunaresan, "Cluster sampling to assess
immunization coverage: A review of experience with a simplified
method."  Bulletin of the World Health Organization 60 (2):
253-260, 1982.> and has been adapted to other epidemiological
areas by Frerichs.

<$FFrerichs, R. R. and Tar Tar, K. "Computer-assisted rapid
surveys in developing countries." Public health reports 104:
14-23, 1989.> Recently, some exploratory work has been undertaken
to adapt the methodology to family planning and primary health
care, with special emphasis on using rapid surveys to help
managers improve planning and monitoring of services.

<$FFrerichs, R.R., et al. "Family planning survey and antenatal
survey, Srisaket, Thailand," December, 1987 (unpublished paper,
URC/CHS); "Institutionalizing the use of rapid surveys for family
planning decision-making," an Operations Research Proposal, Gadja
Mada University and University Research Corporation, November
1989; and "Primary Health Care Management Advancement Programme
1989-1992," a proposal of the Aga Khan Health Network and PRICOR,
November, 1989.>

Table 1: Type of attendant at last delivery,, N = 206, +, +
Type attendant             Number       Percent
Government health worker    114          55.3
Traditional birth attendant, 89          43.2
Unattended,                   3           1.5
Total                       206         100.0



Table 2: Received tetanus toxoid last pregnancy,, N = 209, +, 
+
Received TT, Number, Percent
Yes, 145, 69.4
No, 52, 24.9
DK/NR, 12, 5.7
Total, 209, 100.0

Figure 1: Received tetanus toxoid last  pregnancy, N= 
209

The entire survey, including training, was conducted in only two
weeks. This was not a survey of a small population. Six districts
with an estimated combined population of 410,891 were sampled. 
The antenatal care (ANC) survey targeted married women currently
living in these six districts who had had a pregnancy outcome,
live birth, stillbirth, miscarriage, or abortion, within the past
24 months.  This population was estimated to be 24,653 women.  
 The total number of people interviewed in a rapid survey is
typically 210, i.e., 30 clusters x 7 respondents each.   Although
this number appears small, it has been used extensively in the
Expanded Programme for Immunization (EPI) and has been shown to
produce unbiased estimates within the desired level of precision,
that is, plus or minus 10 percent.
A number of improvements can be made in this methodology to
reduce sampling error, increase the level of confidence, and
gather more detailed information.  These are described in this
guide.
 
Some limitations of rapid surveys
Although rapid surveys are attractive alternatives to traditional
large-scale sample surveys, there are trade-offs.  The most
obvious is that the number of questions must be limited.  The
manager cannot expect detailed findings or analysis.  Another is
that these surveys are designed to assess levels of health or
service coverage, not to identify the determinants or causes.  A
rapid survey can tell you the proportion of women who breast feed
their children, but it usually cannot tell you the differences
between women who do and do not breast feed. This is because the
sample is very small and provides a single estimate.  You cannot
break the sample into sub-samples to see, for example, if younger
women breast feed longer than older women, or if rural women are
different from urban woman, or if Moslem women are different from
Christian women.  If you want to have that kind of information,
you need to do a separate rapid survey for each group.
Questions that are going to be tested for statistical
significance must be tabulated in a "yes"/"no" format, as certain
statistical rules apply to this type of sample.  For this reason,
questions that are going to be analysed statistically should not
be multiple choice or open-ended.  However, it is possible to do
statistical tests of multiple choice questions by recoding the
questions into a "yes"/"no" format.  (This is explained in the
text.)  There is some value in using multiple choice and
open-ended questions, even if their results cannot be analysed
statistically. This is because they may produce useful
descriptive information that can give you suggestions of possible
explanations.
Rapid surveys often rely on CHW's, health staff, or community
members to conduct the interviews rather than on professional
interviewers. Even with careful training, non-professionals are
often tempted to take shortcuts, thus producing biased results.
Typical problems include not contacting certain households or
women because: they are too far away; the interviewer "knows"
that some women are not eligible and therefore do not need to be
contacted; and the interviewer knows the respondents so well that
he or she answers the questions for them.
These problems are not unique to rapid surveys however, and can
usually be controlled through training, close supervision, and
spot checks, or re-interviews with a small sample of respondents. 
Appendix E provides some guidelines for dealing with this
problem.
Sampling can be a problem if the WHO/EPI cluster sampling
procedure isn't appropriate for a given survey.  This technique
seems to work better for EPI than some other PHC interventions.
Immunizations are often carried out as campaigns where all of the
children in a target village or urban block are immunized at one
time. Thus, the villages tend to be "homogeneous," i.e., either
most of the children are immunized, or most are not. This means
that it is not overly important who is interviewed in a given
cluster, since most children have the same characteristic (they
are immunized or not). This is not so for some of the other PHC
interventions, such as antenatal care. In such cases, this
technique may produce biased results if households are skipped
because eligible respondents are not at home when the survey is
conducted.
Given the distances that might have to be travelled to get to a
cluster, the standard WHO/EPI procedure is to interview only
those women who are at home.  This can produce significant bias,
particularly during planting and harvest periods when many
able-bodied women are away in the fields.  This problem can be
largely avoided, by scheduling the survey during seasons when
respondents are likely to be home, by visiting villages early in
the morning or in the evening, by arranging the visit to coincide
with a special event, or by making call-back visits.  The
trade-off is that this can increase the costs of and the time
needed to conduct the survey.
Rapid survey results do not produce exact estimates of values. 
Rather, they produce confidence intervals, usually plus or minus
10 percent.  For example, the ANC survey described previously
showed that 69.4 percent of the respondents received tetanus
toxoid immunization during their last pregnancy.  It is more
accurate to say that we are 95 percent confident that the true
percentage lies somewhere between 59.4 and 79.4 percent.  Another
way to state this is that the estimate is 69.4 percent, plus or
minus 10 percentage points.
This is a large interval and may not be of much use to some
managers, especially if they want to plot trends over time.  All
surveys have a potential sampling error, but it is usually much
lower, around three to five percentage points.  This problem can
be dealt with, at least partially, in several ways:  by
increasing the sample size; by accepting a lower confidence
interval (e.g., 90 percent, 80 percent); and by conducting a
post-enumeration survey to validate and adjust the results, if
necessary.
You can survey several target groups at the same time, such as
children under two years of age (for immunization), pregnant
women (for antenatal care), and children under age five (for
ORT).  However, the overall sample for this type of study will
usually be larger than 210 because you need 210 respondents from
each target group.  But many respondents fall into two or more
categories.  A typical example is a pregnant woman with a child
under age five.  She can respond to questions on antenatal care,
ORT, immunization, growth monitoring, nutrition, ARI, water and
sanitation, and other topics.

Figure 2: Proportion of married women who received TT
, 90%, 95%, 99%

High, 0.774, 0.79, 0.821
Mean , 0.694, 0.694, 0.694
Low, 0.613, 0.597, 0.567


Proportion
Confidence intervals
0.774
0.79
0.821
0.694
0.613
0.597
0.567

Surveys of mortality and morbidity require much larger samples
because the events are so rare.  For example, if you want to
assess infant mortality, you may need a sample of 2,000 mothers
to identify enough infant deaths.  For child and maternal
mortality rates you may need 7,000 interviews.  If the maternal
mortality rate (MMR) is 250/100,000 live births, you would
normally need to identify 1,000 women who had been pregnant just
to find two to three maternal deaths.
Some indirect estimation techniques have been developed to get
estimates with smaller samples.  These are described in Appendix
G.5.  However, the samples are still relatively large and the
analytical techniques are complex.  You should probably get
expert advice before trying to measure mortality rates.
Identifying broad disease patterns and health problems is
probably more useful to PHC managers. This can be done through
the Vital Events questionnaire provided in this module.   It may
also be important to determine the causes of some of the more
serious events, such as a maternal death or a case of polio. 
Module 4 (Surveillance) describes how to do that.
Finally, these surveys are completed most rapidly if computers
(laptops or PCs) are used. The survey programs, including those
provided in this module, can help you to calculate your sample in
a matter of minutes. Data entry and analysis can also be done
very quickly. Surveys for which computers are not used will take
longer. But they can be done, and this User's guide describes
how.
 How you can use this module<$&common gaps>
This module is designed to provide PHC managers, consultants, and
researchers with simple and inexpensive tools that they can
readily adapt to assess quickly community health needs and/or PHC
programme effects on health knowledge, behaviour, and status.
The most common information gaps that managers have fall into
these two areas. First, many managers have no way of determining
what the real PHC needs of their target population are.  What do
people know about immunizable diseases?  What do they do about
diarrhoea?  What is the health condition of their infants?  And,
second, they have no way of assessing the effects that their PHC
programme activities are having on those needs: What have mothers
learned about nutrition? What are they doing about sanitation?
What improvements have been made in immunization coverage?
Module 2 is designed to help managers collect and analyse this
kind of information and to do it quickly, simply, and
inexpensively.  It can be used by established and new PHC
programmes.  Public and private programmes can use it, as well as
single (categorical) and comprehensive (integrated) programmes.
Although this module is part of a series, it can be used
independently. You do not need to use any other module before or
after this one. We hope that you will, however.  That's the
purpose of having a series, after all.
Module 2 is linked to Modules 3 (Work Planning), 4
(Surveillance), and 6 (Service Quality) in particular.  Module 2
can provide broad findings which the other modules can be used to
examine in more detail. In addition, the instruments in Module 2
can be used in Module 4 for surveillance of morbidity and
mortality.
 How the module is organised
This User's guide includes:
 Basic instructions: how to design and conduct a rapid survey by
hand (without a computer); ten easy steps described in pages
14-49.
 Epi Info to design and carry out a rapid survey (with a
computer):  See  Appendix A,  26 pages of instructions, and use
the complete computer files.
 Sample questionnaires and "cluster registers" that you can adapt
to fit your own needs.  There is one for each PHC topic,
pre-coded and ready to use or adapt to fit local needs in
Appendices C and D.
 A simple computer program for drawing cluster samples quickly
and accurately (Appendix F), and other tools for estimating
sample sizes and drawing random samples (Appendix G).
 Other guidelines and tools that you may find useful: suggestions
for constructing questionnaires (Appendix B); guidelines for
training and supervising interviewers (Appendix E); survey
management forms (Appendix H); and tabulation and analysis
templates (Appendix I).
 
Sample rapid survey questionnaires have been included for each of
the principal PHC services:
 
 Health education
 Antenatal care, safe delivery and postnatal care
 Family planning
 Breast feeding, growth monitoring, nutrition education
 Acute respiratory infections
 Breast feeding
 Diarrhoeal disease control/oral rehydration therapy
 Childhood disabilities
 Child immunization
 Growth monitoring/nutrition education
 Water supply, environmental hygiene and sanitation
 Accidents and injuries
 Chronic, non-communicable diseases
 Malaria
 Tuberculosis
 Sexually-transmitted diseases, HIV/AIDS
 
In addition, one questionnaire has been prepared for assessing
multiple health needs/services and three others for morbidity and
mortality.
 
 Community assessment of PHC services
 Vital events and health status
 Child morbidity and mortality
 Adult morbidity and mortality
 
There is also a Facilitator's guide that you can use if you wish
to conduct a workshop or seminar on rapid surveys.  Computer
programs are also included on the disks that are included.  One
disk contains the complete Epi Info manual and program files.
 Where to begin
If you already know something about surveys and computers, you
may want to skip to Appendix A: How to use Epi Info to conduct
rapid surveys.  If not, you should start with the basic
guidelines in the next section: Rapid survey procedures. These
include worksheets that you can fill out as you follow the steps.
You may also want to have a facilitator help you, especially if a
group is going to learn how to do rapid surveys.
The facilitator should have experience in surveys and computers. 
He or she need not be an expert trainer.  The Facilitator's guide
includes session plans and provides charts that can be
photocopied for handouts or transparencies.

 Adaptations
All of the tools, checklists, guidelines, computer programmes,
and other material included in this module are illustrative.  It
is expected that they will need to be revised to fit local
conditions, and you are encouraged to do so.
You may want to change the wording of questions in the prototype
questionnaires; you may want to eliminate certain questions and
add others; you may want to redesign the format, design your own
instrument, use another computer programme (SPSS or dBase), draw
your sample from your computerised household registration
database, create other kinds of graphs, and so on. We want to
make it very clear that this is encouraged. These tools are not
presented as "standards" that should be used in all PHC
programmes.  Rather, they are designed to encourage and help
managers to carry out community surveys by providing them with a
starting point, general tools, and guidelines that they can adapt
to fit their own situations.
Don't forget to include the target communities and your CHWs in
the design, execution and analysis of the survey.  There are
several advantages to this:

 It increases awareness and knowledge about PHC in general;
 It increases awareness in the communities of their own health
problems and needs; and 
 It increases "ownership" of the survey, its results and
recommendations.

Community representatives can also help you to phrase questions
in local terms so that they are better understood.  They can help
you to identify common health problems to be  included in your
survey, help you to find eligible respondents, and help you to
understand responses and hidden meanings.
The elderly need to be targeted with health messages 
so that their authority in the family helps support good health 
practices
Photo by Jean-Luc Ray for AKFRapid survey procedures
The steps in the design and conduct of rapid surveys are no
different from those of larger, more traditional sample surveys. 
If a team is going to design the survey, field experience has
shown that it may be faster to divide the tasks and have one
group develop the data collection instrument (Steps 1-4) and
another the sampling (Steps 5-6). 

Step 1:, Specify the objectives of the rapid survey
Step 2:, Decide what indicators to use 
Step 3:, Develop an outline for the survey report
Step 4:, Design the data collection instruments
Step 5:, Develop the sampling procedures
Step 6:, Schedule the survey
Step 7:, Collect the data
Step 8:, Enter,, verify and tabulate the data
Step 9:, Analyse,, interpret and report the findings
Step 10:, Develop an action plan

 Step 1: Specify the objectives of the rapid survey
If you have gone through Module 1, you may already have
determined the information you want about the health needs and
PHC coverage of your target groups.
Module 1 asked you to summarise your programme's major health
goals, target groups, the PHC services offered to each target
group, and indicators of coverage for each PHC service.  It also
suggested that you include additional target groups or health
services if you think your programme should be expanded.  The
point of going through these steps was to determine if you have
enough information about coverage, and if not, the information
that you need to collect.
A summary worksheet from Module 1 is reproduced below.
If you haven't already filled out a worksheet like this, it would
be helpful to do so now. This should give you a summary of the
coverage information you need to collect.

Target group (yrs), Health services , Coverage indicators
 5, Child immunization, % << 24 mos. fully immunized
, Growth monitoring, % << 5 yrs. weighed
, Oral rehydration, % using ORT last episode diarrhoea
, Nutrition education, % << 2 yrs low weight-for-age 
Married women 15-49, Antenatal care & TT, % pregnant women
enrolled in ANC
, Family planning,    No. new acceptors

Now you can be specific and state the objectives of your rapid
survey.  The following worksheet can help you to summarise them.
First, you need to identify the user or users of the information. 
This is critical, since each could have different interests.  If
there are multiple users (yourself, a donor, an evaluation team),
you should find out exactly what each one wants.
You also need to clarify precisely why this information is
needed.  The two most common purposes are to assess health needs
and to evaluate programme performance.  These are quite
different, and although Module 2 can address both, the design and
analysis of your rapid survey will be slightly different
depending on which purpose you have.  For one thing, if you are
interested in assessing performance, you would limit the survey
to current PHC services.  But if you want to assess needs, you
would not.

WORKSHEET FOR SPECIFYING RAPID SURVEY OBJECTIVES, +
User, Target groups
     ___Manager, ___Children
___Board, ___Women
___Donor, ___Other:______________
___Community , 
___Other:______________ , 
Purpose, PHC service(s)
___Planning,  _____________________
___Health status/needs,  _____________________
___Service status/needs, _____________________
___Evaluation, _____________________
___Service coverage/effects, _____________________
___Health status/impact,  Geographic area: _______________
Start Date:______________, End Date:  _______________
, 
Next, you should decide which target population and which PHC
component the user(s) wants to study.  The summary worksheet from
Module 1 is a place to start.  But you may have listed several
target groups and services.  Do you want to study all of them or
just one?  You can study as many as you want, but keep in mind
that you will have to draw a separate sample for each target
group that you include.  So if you decide to study ANC,
immunization, and sanitation, you will have to do three rapid
surveys: one of pregnant women, one of children 12-23 months, and
one of households.  These can all be done at one time, but each
will require 30 clusters of seven respondents.   In many cases,
the clusters and respondents can be the same for each survey.  So
if you plan to do several rapid surveys, it would be economical
to combine them.
This module describes how to do this in a relatively simple way,
so do not be intimidated.  If you need information on several
target groups and services, then note who and what they are in
the worksheet.
You also need to specify the geographic scope of the survey. 
Will it cover the entire catchment area or a part of it?  Keep in
mind that the survey results will only represent the area
covered. If you decide to limit the survey to five districts
because they are close by, then the outlying areas will not be
represented in the study.  That means you cannot apply the
results to the entire project, only to the five districts.
Lastly, it is helpful to determine immediately when the study
will start and especially when the results will be needed.  It
does no good to provide rapid survey results after the deadline
for making key decisions.

Measuring  effects and impact, +
Module 2 can be used to measure:, +
Programme effects on target groups, +
, Knowledge
, Attitudes/motivation
, Behaviour/coverage
Programme impact on target groups, +
, Morbidity
, Mortality
, Disability
, Fertility
 
You can use Module 2 to attribute changes in knowledge, attitude,
and practice (KAP) to your programme.  To do this you should:
conduct a baseline survey of the current KAP of your target
groups; carry out your programme for a specified period, e.g.,
1-2 years; and conduct a follow-up survey sometime later.
Use the computer program in Appendix G-2  (Hypothesis testing -
two samples of equal size) to determine the sample sizes needed
for a "before-after" comparison.  Estimate the current level of
your principal effect or impact measure, such as coverage. Enter
this in the line that reads "Est. proportion with the attribute
FIRST (or BEFORE) population." Decide the amount of change you
wish to detect and enter that in the next line "Est. the
proportion with the attribute in the SECOND (or AFTER)
population."  For example, if you want to measure changes in
infant mortality, you would enter the current figure (say .08) as
the BEFORE proportion and .06 as the AFTER proportion.  Thus you
will be able to detect a drop in the IMR from 80/1,000 to
60/1,000. Enter the other information requested, and the
programme will compute the sample sizes you will need for the
baseline survey and the follow-up survey.
Your evaluation will be even better if you include a control area
which is similar to your project area but which does not get PHC
services. Conduct baseline and follow-up surveys there at the
same time. The design of this kind of study is summarised below,
in which O indicates an observation (baseline and follow-up
surveys) and X indicates the PHC programme intervention.


, Study design, +, +
, Baseline, Intervention, Follow-up
Programme area, O1, X, O2
Control area, O1, , O2

Note: Impact evaluations are easy to conceptualise, but difficult
to carry out.  Seek professional advice before you start.

Step 2: Decide what indicators to use<$FSee Module 1 for a
discussion of types of indicators and Module 5 for lists of
indicators.>
This module focuses on outcome indicators, especially coverage.
These are the best indicators for assessing health needs and the
effects of the PHC programme on health.
Other modules deal with input and process indicators: the
performance of health workers (Module 3), short-term assessment
of PHC activities (Module 5), the quality of PHC services (Module
6), and the effectiveness and efficiency of PHC management
(Module 7). You can include any of these indicators in a rapid
survey, of course, as long as they are community-based.  Rapid
surveys gather information about populations, not about health
centres or staff.
Thus, the manager (or team) needs to decide which outcome
indicators to examine.  In most cases these will deal with
coverage.  In general, coverage indicators tell you the
proportion of the target population that is protected by your
programme.  Examples of coverage indicators (taken from Module 1)
are shown below.

Examples of PHC coverage indicators , +
Service, Coverage indicator
Antenatal care, % made 3 or more ANC visits
Tetanus toxoid immunization, % received TT immunization
Safe delivery, % delivered by trained attendant
Family planning, % current users of FP services
Breast feeding, % breast feeding to 18 months
Growth monitoring, %  2 yrs weighed last quarter
Child immunization, % 12-23 mos. fully immunized 
Acute respiratory infection, % ARI cases treated
Diarrhoea disease control, No. children << 5 yrs with
diarrhoea/1,000
Oral rehydration therapy, % used ORT last episode diarrhoea
Water,, sanitation,, hygiene, % households with safe
water/latrine
Vitamin A, % 6-12 mo. received Vit. A
Tuberculosis, % cases followed to cure
Malaria, % cases treated
Sexually transmitted diseases/HIV, % target group
infected/treated
Disability, % 5 yrs disabled
Health education at home, % schools receiving or participating in
health education activities
Drug supply, % communities with adequate supplies
Accidents and injuries,  No. accidents + injuries/1,000
population
Chronic,, non-communicable diseases, % target group with
hypertension,, 
chronic heart disease,, anaemia,, diabetes
Treatment of minor ailments, % episodes treated
Nutrition education, % low weight-for-age

You may also be interested in finding out what people know about
a service or health problem, the skills they have in diagnosing
and treating health problems, or people's opinions about your
health workers and programme.  You may also want to assess health
status and the impact of the programme on health.  These are all
legitimate interests and can be included in a rapid survey.
Module 5 includes lists of recommended indicators for each PHC
and management service.  It also includes separate lists of
morbidity and mortality indicators.  These lists can be helpful
guides in designing a questionnaire.
Make a list of the key indicators you want to measure.  It
shouldn't be a long list. Start with the desired outcome.  Select
one or two indicators for that. Then work backwards (as described
in the "If-then" sequence and Worksheet A in Module 1) to
determine what needs to happen for this outcome to be achieved. 
Select one or more key indicators for that outcome, and so on.
This is a good time to involve the staff, community
representatives, and others who are part of the programme. They
can articulate the perspectives of the community, and they often
have valuable insights as to what is important, and feasible, to
measure.
In addition to performance indicators, three other types of data
are usually needed in a rapid survey:
 Descriptors: respondent name, village name, address of
household, etc;
 Characteristics of respondent: age, sex, parity, education,
literacy, income, caste, race, ethnic group, etc; and
 Survey management data: interviewer name, supervisor, date of
interview, etc.
The descriptors and management data are needed to identify the
respondent and, in particular, to make a call-back visit and to
correct possible errors.
The characteristic used to describe the surveyed group.  If
several surveys are done, then the characteristics can be useful
for making comparisons between one kind of respondent and
another.  The manager may want, for example, to know if there is
any difference between the health behaviour of people in one
district and another, or between one ethnic group and another. 
Separate surveys would be required to do this, of course. 
Remember that each rapid survey is designed only to provide
information about a single population.
 Step 3: Develop an outline for the survey report
Most researchers don't do this, but it will be very helpful at
this point to outline a report so that you and your colleagues
know what information will be produced. That will make
construction of the questionnaire, data analysis, and especially
interpretation much easier.  If this is not done, you may get a
lot of information that you don't want, such as 10 pages of
tables describing the characteristics of the survey population.
Step 10 includes a generic outline for getting started.
This is also a good time to check with outside users to make sure
that the report will meet their expectations.
Most managers aren't interested in background, methodology,
qualifications, descriptions of the survey process, descriptions
of the survey population, etc.  They want results. So write your
report outline to suit the user's needs. The other information
should be presented, but not necessarily first. Many managers
want the "bottom line" first and will ask about the sample, data
collection procedures, and other methodological issues later.
Construct a list of "dummy tables" (blank tables -- without data)
and identify the kinds of frequency distributions and
cross-tabulations the user(s) wants. Don't forget to include
graphs (pie charts, histograms).  Appendix I includes an
illustrative list of frequency distributions and
cross-tabulations.  This type of list will be needed so that the
analyst knows which tables and graphs to produce.  Obviously, it
can be modified later, but it can help your planning to start it
now.
List questions and issues that you think should be addressed in
the report.  Typical questions are:
 What percentage of the target group is covered?
 What are the major reasons some people aren't covered?
 Where do people go for services?
 What are the major reasons they don't utilise available
services?
Finally, make an outline, and set a page limit.  Put more
emphasis on visual presentation through handouts, transparencies,
slides, etc. These attract managers' attention more than written
text.  Remember KISS - "Keep It Straightforward and Simple".

Step 4: Design the data collection instruments
The following checklist summarises your main options and the
substeps that you will need to follow.

Checklist for designing questionnaires, +
Type survey instrument:, Types of questions/fields:
__Register, ___Multiple choice
___Other, ___Open-ended
, ___Dates
, ___Ranges (e.g.,, 1-4 years)
, 
Target group(s):, Coding:
, 
___Children, ___Uncoded
___Women, ___Pre-coded
___Other, ___Numerical
, ___Alphabetical
, ___NA: Not applicable
, ___DK: Don't know
, ___NR: No response

 Questionnaire or register
First you need to decide if you want to use a questionnaire or a
"cluster register."
Questionnaires usually provide more information, including
instructions for the interviewer, the exact phrasing of each
question, and pre-coded responses.  But you have to have one
questionnaire for each of the 210 respondents.  That is, a
minimum of 210 pages, 420 if its a two-page questionnaire.
Cluster registers allow the interviewer to record the responses
of all seven or eight respondents from one cluster on the same
page. This means you will need only 30 pages, one for each
cluster.
But the number of items ("questions") is limited by the size of
the paper.  Instructions and question phrasing have to be
provided elsewhere, or learned beforehand.  Excerpts of both are
shown on the next pages.


Exhibit 1: Excerpt from rapid survey questionnaire on antenatal
care, safe delivery, and postnatal care
Complete for all women currently living in the household 
who have had a pregnancy outcome during the past 24 months. The
outcome may be a live birth, stillbirth, or abortion. If the
woman has had more than one pregnancy, ask about the most recent
pregnancy outcome.                                                
                                     
IDENTIFICATION, +, +, +, +, +, +, +, +
1., Study no, , 2., Province no., , 3., Cluster no, 
4., Interviewer no, , 5., Respondent no., , 6., Date of
interview,   /  /
7., Respondent age, , 8., Respondent sex, , , , 

NAME OF RESPONDENT________________________________, +, +, +, +,
+, +, +, +

9.   How many live births have you had so far? Number 
of live births:____ <R>
     ___(99) DK/NR

10.  Did you receive antenatal care during your last 
pregnancy?
___(1) Yes    ___(0) No, go to Q14    ___(9) DK/NR, 
go to Q14

11.  How many times did you get antenatal care? _____times

 DK/NR, enter 99)

12.  How many months had you been pregnant before 
you got antenatal care?
____(1) 3 mo. (first trimester)
____(2) 4-6 mo. (second trimester)
____(3) 7-9 mo. (third trimester)
____(9) DK/NR

13.  Which is the principal place you received antenatal 
care?
____(1) Hospital
____(2) Health centre/clinic
____(3) Private hospital/clinic
____(4) Local TBA/healer
____(5) Other site of care specify: _______________
____(9) DK/NR

14.  Did anyone advise you to get antenatal care? ___ 
(0) No

Yes:
____(1) Physician, nurse           ____(5) Mother, relative
____(2) Community nurse/midwife         ____(6) Friend, 
neighbour
____(3) CHW/volunteer              ____(7) Other:_________
____(4) Traditional birth attendant          ____(9) DK/NR
15.  Did you receive a tetanus vaccination during 
your last pregnancy?
___(1) Yes    ___(0) No, go to Q17    ___(9) DK/NR, 
go to Q17
16.  How many vaccinations did you receive?
___(1) One   ___(2) Two   ___(3) Three or more   ___(9)DK/NR, go 
to Q17

Questionnaire continues for four months


Exhibit 2: Excerpt from cluster form on antenatal 
care, safe delivery and postnatal care
(1), Study No.  , N, , +, +, +, +, +, +, +, +, +
(2), Province No., A, ^, ^, ^, ^, ^, ^, ^, ^, ^, ^
(3), Cluster No. , M, ^, ^, ^, ^, ^, ^, ^, ^, ^, ^
(4), Interviewer, E, ^, ^, ^, ^, ^, ^, ^, ^, ^, ^
(5), Respondent No. , , 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
(6), Date     /     /     , , , , , , , , , , , 
(7), Age, , , , , , , , , , , 
(8), Sex, , , , , , , , , , , 
(9), How many live births have you had so far?, , , , , , , , , ,
, 
(10), Did you receive antenatal care during your last pregnancy?,
, , , , , , , , , , 
(11), How many times did you get antenatal care?, , , , , , , , ,
, , 
(12), How many months had you been pregnant before you got
antenatal 
care?, , , , , , , , , , , 
(13), Which is the principal place you received antenatal care?,
, , , , , , , , , , 
(14), Did anyone advise you to get antenatal care?, , , , , , , ,
, , , 
(15), Did you receive a tetanus vaccination during your last
pregnancy?, , , , , , , , , , , 
(16), How many vaccinations did you receive?, , , , , , , , , , ,
(17), During your pregnancy,, did you take iron pills to keep you
strong?, , , , , , , , , , , 
(18), What was the outcome of your most recent pregnancy?, , , ,
, , , , , , , 
(19), Where did the delivery take place?, , , , , , , , , , , 
(20), Who was the main person attending the delivery?, , , , , ,
, , , , , 
(21), What is the name of the local CHW?, , , , , , , , , , , 
(22), Has the CHW visited/contacted you during the last three
months?, , , , , , , , , , , 
Key (12), +, (18), , , , , , , , , , 
1, three mo. (first trimester), 1, live birth, 
+, +, +, +, +, +, +, +, +

2, 4-6 mo. (second trimester), 2, stillbirth, 
+, +, +, +, +, +, +, +, +
3, 7-9 mo. (third trimester), 3, abortion/miscarriage, 
+, +, +, +, +, +, +, +, +
9, DK/NR, 4, DK/NR, +, +, +, +, +, +, +, +, +
, , , , , , , , , , , , 
, , , , , , , , , , , , 
, , , , , , , , , , , , 
, , , , , , , , , , , , 
, , , , , , , , , , , , 


Procedures
For larger clusters, 15 or more respondents, you may want to
switch the rows and columns.  That is, put the questions in the
top row and the respondents in the left column.  Here's a partial
example.  See Appendix I for an example of how to set up a
register like this on a spreadsheet so that all 210 respondents
can be recorded and analysed on one (long) page or spreadsheet.

Respondent no. name house, Age, Rec'd
ANC, No.visits, Source of service, +, +, Rec'd
TT, No. doses, Pregnancy outcome, +
^, ^, ^, ^, Hosp, Hlth.
Cntr., TBA, ^, ^, Live, Still

1, , , , , , , , , , 
2, , , , , , , , , , 
3, , , , , , , , , , 
4, , , , , , , , , , 
5, , , , , , , , , , 
6, , , , , , , , , , 
7, , , , , , , , , , 
8, , , , , , , , , , 



One target group or several
Many managers will want to include several topics and target
groups in the same survey.  This will not be a problem as far as
instrument design is concerned.  Generally, you have three
choices. First, you could use the sample Community assessment of
PHC questionnaire in Appendix C. This instrument includes 4-5 key
questions on each of several PHC topics: 
Availability of health care
Antenatal care and childbirth
Family planning/child spacing
Breast feeding and growth monitoring
Immunization
Diarrhoea and ORT
Water and environmental sanitation
Malaria
Child disabilities
Tuberculosis
Sexually-transmitted diseases
Second, you could combine several questionnaires from Appendix C
into a single instrument.  Third, you could pick out questions
you are interested in from other questionnaires and construct
your own instrument.
However, be aware that combined questionnaires usually require
information about different target groups.  If that is the case,
you will have to draw separate samples for each target group.
This will be explained more fully in the discussion of the
sampling step.
When doing a multi-target group survey, it is best to prepare the
questionnaire in modular form.  Combine all of the questions for
one target group and put them in one section.  Develop a separate
section for each target group.  Be sure to include descriptor
information in each module so you can identify the respondent.
 Design the data collection instrument(s)
The sample instruments in Appendices C (questionnaires) and D
(cluster registers) can be photocopied and used as is. They can
also be revised, new ones can be compiled, etc. The questions
should be designed to collect the indicators you identified in
Step 2.
The sample instruments show how most questions can be structured
in a "Yes/No," precoded format.  This is important for those
questions that you want to analyse statistically.  When multiple
choice and continuous variable questions are needed, make sure
they can be recoded later to be analysed as "yes"/"no"
(dichotomous) questions.  Some examples:

Multiple choice, Recoded as dichotomous (Yes/No)
Received care from :, Received care from:
, 
1 Provincial hospital , 1 Provincial hospital
2 District hospital, 2 Other
3 Health centre, 
5 Traditional birth att., 
6 Private clinic, 
7 Other (name:          )   , 
9 Don't know/No response<P12M>, 9 Don't know/No response
, 
 
You can repeat this procedure for each response.  That is, you
can recode the question as District vs. Other; Health centre vs.
Other, and so forth.  You can recode the question for a different
age group: under 25 years vs. 25 years and over.
Continuous variables can be recoded as shown in the following
example.

Continuous , Recoded as dichotomous (Yes/No)
Age of respondent_ _ yrs., Age of respondent
(Enter completed years, 1 ___ Under 30 years
or 99 if DK/NR), 2 ___ 30 years and over
, 9 ___ Don't know/no response
 
Most rapid survey questions allow only one response per question.
If you include questions that allow multiple responses, you can
do one of two things.  First, you can ask the respondent for the
most important (principal, major, etc.) single response, and code
only this response.  For example:
 
Q.8. Where did you first hear about ORT? (Check the first source
mentioned)
 (1)  Friend or relative 
 (2)  CHW 
 (3)  Doctor/nurse 
 (4)  Television 
 (5)  Radio 

Or, you can record each response as a separate question.
 
Where did you hear about ORT? (List all mentioned)
 Q.8  Friend or relative (Y/N)
 Q.9  CHW (Y/N)
 Q.10 Doctor/nurse (Y/N)
 Q.11 Television (Y/N)
 Q.12 Radio (Y/N)
 
Appendix B includes some other suggestions for designing
questionnaires.

 Pre-test the instrument
If the questionnaire has to be translated into a local language,
this should be done after it is drafted, but before it is
pre-tested.  Do not leave it to each interviewer to do his or her
own translation.  This can cause confusion and misinterpretation. 
It is best to translate the instrument into the local language
and then have someone else translate it back to make sure the
questions are clear.
When the instrument has been drafted, it should be pre-tested
with a small sample (five to ten women) to make sure that the
questions are understandable, the pre-coded responses are
realistic, and the sequence of questions is logical. Then the
instrument should be revised as appropriate.
Ideally, the interviewers should be involved in the pre-testing,
perhaps as part of their training.  Their feedback can be very
helpful in making revisions. It is also important that they
follow the sampling procedures for selecting households and
respondents.  They may have questions and suggestions about those
procedures, as well.
Pre-testing is very important and should be done carefully with
the intended target group as respondents.  Pre-tests with office
colleagues or PHC staff are of very limited value, since they do
not represent the target groups.  A worksheet should be devised
for summarising feedback.  Or notes could be written on the
instruments to identify problems with comprehension (interviewer
and respondent), coding, and logical sequence.
 Estimate the data collection requirements
The pre-tests should give you an idea of the amount of time it
will take to find a respondent, and complete an interview.  With
this information you can estimate the number of interviewers you
will need and the number of days it will take to complete data
collection.
Usually you will want a two-person team to complete at least one
cluster per day, seven or eight interviews.  You will also need
several supervisors, the number of which will depend on the
number of interviewer teams you have and the distance the
supervisors will need to travel between clusters.
If the questionnaire is short, the respondents are easy to find,
and distances between clusters are short, then a team  should be
able to complete two or three clusters each day.  Use the
following formula to make a quick estimate:

over {complete~~2~~clusters~/~day}> = 15 "team days"

If you have five teams (ten persons) the data collection could be
completed in three days.  Or three teams would require five days. 
Fifteen teams could finish in one day!
 Develop a code book
This can be very useful, especially when the questionnaire is
long or in modular form.  Code books can be important during
analysis, especially if tables are produced with codes instead of
labels.
Code books generally include the following: variable number,
name, label, value codes, and  value labels.  An example is
included in Appendix B.

Step 5: Develop the sampling procedures<$&step 5[v]>
 Target groups, attributes and sample sizes
These instructions are for manually drawing a cluster sample of
30 clusters with seven respondents each.<$FThis module emphasizes
cluster samples.  If you can draw a simple random sample, you are
encouraged to do so.  See Appendix G for instructions. 
Stratified samples can be even more effective, but they are also
more complex.  You should seek expert advice for this kind of
sample.  Stratified samples are not described in this module.>
 You may skip this step if:
 You have a complete household listing of your survey area so
that you can draw a random sample.  See Appendix G
 You plan to use a computer to draw your sample. See Appendix F
which describes how to use a spreadsheet program (Cluster
Identification Worksheet) to draw a cluster sample
First we will describe the basic procedures for drawing a sample
of 30 clusters.  Then we will discuss some variations of those
procedures that you may have to apply.  The selection of the
seven respondents from each cluster is described in Step 7.
 Determine the size of the clusters
Although most people think of clusters as natural groupings of
people (villages, census tracts, urban blocks), clusters have a
different meaning in sampling. In cluster sampling you will
divide your total survey population into 30 equal groups.  Each
of those groups will be a "cluster."  Then you will identify
seven respondents in each of those clusters.  This second step
will be discussed in Step 7:  Collect the data.
You need to know the total population of your survey population
to determine the size of each cluster.  Simply divide the total
population by 30.  For example, if your catchment area has 45,000
people, each cluster will include 1,500 people (45,000/30 =
1,500).
It doesn't matter if there are fewer or more than 30 villages or
districts, since you will define the clusters by dividing the
total population into 30 groups of equal size.
It also doesn't matter if the population is scattered over a
large area.  People can even be on islands or in remote areas and
still be included in the sample. If you want your sample to
represent all of your target population, then do not leave any
"natural clusters" out. However, if it is not feasible to include
some areas, then leave them out. BUT, remember, your sample will
not represent those who are left out. If you limit your sample to
people within one kilometre of a health centre, for example, then
that is all it represents.
You must have population size estimates of the sub-units of your
sample. List these sub-units, villages, census tracts, voting
precincts, towns, in Columns A and B of a Cluster identification
worksheet(see Exhibit 3).


Exhibit 3: Cluster Identification Worksheet            
Cluster.WQ1
Number of Clusters, +, +, 30, , (Enter Number Desired), 
+, +, +
Sample Population Size, +, +, 29,481, , (Enter total from Column
C), +, +, +

Cluster Size (sampling interval), +, +, 983, ^, , +, +, +
Random Start Number, +, +, 491, ^, ^, ^, ^, ^
INPUT DATA, +, +, , ^, OUTPUT DATA, +, , Sample Sites
Enter, Preset, Enter, Computer, ^, Preset, Computer, ^, 
A, B, C, D, ^, E, F, ^, G
Community, Community, Estimated, Cumulative, ^, Selected, 
Start, ^, Community
Name, Number, Population, Population, , Cluster, Number, , Name
Pagai, 1, 548, 548, ^, 1, 491, ^, Pagai
Santai, 2, 730, 1,278, ^, 2, 1,474, ^, Serina
Serina, 3, 686, 1,964, ^, 3, 2,457, ^, Fanta
Mulrose, 4, 280, 2,244, ^, 4, 3,440, ^, Fanta
Fanta, 5, 1,256, 3,500, ^, 5, 4,423, ^, Rostan
Bagia, 6, 684, 4,184, ^, 6, 5,406, ^, Mt. Sil
Rostan, 7, 919, 5,103, ^, 7, 6,389, ^, Mt. Sil
Mt. Sil, 8, 1,374, 6,477, ^, 8, 7,372, ^, Livton
Livton, 9, 1,133, 7,610, ^, 9, 8,355, ^, Pulau
Farry, 10, 544, 8,154, ^, 10, 9,338, ^, Pingra
Tunis, 11, 193, 8,347, ^, 11, 10,321, ^, Pingra
Pulau, 12, 375, 8,722, ^, 12, 11,304, ^, Pingra
Sasarota, 13, 333, 9,055, ^, 13, 12,287, ^, Pingra
Pingra, 14, 3,504, 12,559, ^, 14, 13,270, ^, Srivish
Kanata, 15, 336, 12,895, ^, 15, 14,253, ^, Srivish
Sirvish, 16, 2,115, 15,010, ^, 16, 15,236, ^, Balding
Balding, 17, 258, 15,268, ^, 17, 16,219, ^, Manalopa
Rescuut, 18, 678, 15,946, ^, 18, 17,202, ^, Manalooa
Krista, 19, 207, 16,153, ^, 19, 18,185, ^, Masraf
Manalopa, 20, 1,162, 17,315, ^, 20, 19,168, ^, Abrama
Garafa, 21, 408, 17,723, ^, 21, 20,151, ^, Singri
Spiltar, 22, 455, 18,178, ^, 22, 21,134, ^, Chitoma
Masraf, 23, 978, 19,156, ^, 23, 22,117, ^, St. Kitt

Abrama, 24, 335, 19,491, ^, 24, 23,100, ^, Nevis
Junagadh, 25, 541, 20,032, ^, 25, 24,083, ^, Mt. Carans
Singri, 26, 725, 20,757, ^, 26, 25,066, ^, Charga
Kalarata, 27, 355, 21,112, ^, 27, 26,049, ^, Fosterville
Ichitoma, 28, 498, 21,610, ^, 28, 27,032, ^, Maryoak
Chaplar, 29, 347, 21,957, ^, 29, 28,015, ^, Slipfern
St. Kitt, 30, 186, 22,143, ^, 30, 28,998, ^, Punjak
Nevis, 31, 1,341, 23,484, ^, , , ^, 
Mt. Carans, 32, 670, 24,154, ^, , , ^, 
Betul, 33, 321, 24,475, ^, , , ^, 
Charga, 34, 672, 25,147, ^, , , ^, 
Rio Negra, 35, 705, 25,852, ^, , , ^, 
Fostervill, 36, 444, 26,296, ^, , , ^, 
Maryoak, 37, 781, 27,077, ^, , , ^, 
Slipfern, 38, 959, 28,036, ^, , , ^, 
Tinggi, 39, 305, 28,341, ^, , , ^, 
Punjak, 40, 763, 28,104, ^, , , , 
Capital, 41, 377, 29,481, , , , , 
TOTAL, , , 29,481, , , 28,998, , 
To add more rows see Appendix I.2,, "Adjusting the Spreadsheet."

Press <<CTRL + W>> to widen columns., +, +, +, +, +, +, , +

Record the population of each sub-unit in Column C. This figure
does not have to be exact. The relative size of each sub-units is
what is important. Thus, you can even use data which are a few
years old. In Column D, add the cumulative population.  For
example, add the population of the second district (730) to that
of the first (548) to get the cumulative population of the first
two districts (1,278). Add the population of the third district
(686) to that total (1,278) to get the next cumulative figure
(1,964).
The total size of this population is 29,481.  If your addition is
correct, this should be the figure you end up with at the bottom
of Columns C and D.  Divide that total by the number of clusters
(30) to get the cluster size (both numbers are entered at the top
of the form). 
 
29,481/30 = 982.7,  rounded off to 983
 
This figure is also called the sampling interval, which means the
interval between one cluster and the next.
You are now going to select randomly a number in the first
cluster.  That will be a number between 1 and 983.  Theoretically
this number represents the first person in the cluster that you
will interview.
Use the random number table in Appendix G or take a three-digit
number from any currency note.  In our example, the random number
turned out to be 491.  Enter this at the top of the form. This is
the "start number" for your first cluster.  Now add 983 to that
number to get the start number of the second cluster (491+983 =
1,474). Add 983 to that total to get the third start number
(2,456), and so on.  Repeat this process until you have 30 start
numbers.  These are listed in the table in Column F.
The last step is to identify the communities where those start
numbers are located.  Compare the first number (491) with the
cumulative figures.  Find the first number in Column D that is
greater than 491.  This is 548 (Pagai).  Thus, the first start
number is in Pagai.  Put a 1 in Column A to identify Pagai as the
location of the start number for the first cluster or list the
names of the selected sites in Column G.  Do the same thing with
the second start number in Column E (1,474).  The first number in
Column D that is greater than 1,474 is 1,964 (Serina).  Put a 2
in Column E to identify Serina or write the name in Column G. 
Continue until you have identified all 30 communities where your
start numbers are located.
Communities that have large populations (such as #14 and #16,
Pingra and Srivish) are likely to have more than one start
number.  This is because their populations are two or three times
larger than the cluster size (983). Therefore, two or three of
your clusters may be in one community.  
Now that you know where you have to go, you should look at the
data collection estimate that you made at the end of Step 4. Will
travel time to the clusters increase or decrease the amount of
time it will take to collect the data?  Do you need to revise
your estimates?  Do not draw another sample to make data
collection easier!
This procedure may seem complicated at first, but try it a few
times and you will quickly get the hang of it.  You can also run
the computerized version of this worksheet, which is even easier. 
See Appendix F.
 Variations
 If your sample population is very large, say 300,000, you can
still follow this procedure by doing it twice.  First, list the
large sub-unitss, districts for example.  Follow this same
procedure to identify the districts where the 30 clusters are
located.  Then list the smaller sub-units in each of those
districts.  Make the computations again to find the village where
each start number occurs.  As you will learn, it is better to
start with a large number of sub-unitss. If you only have 30-35
sub-unitss, almost all will be selected and you will have to do a
lot of listing for the next selection.  It would be better to
start with 100-200 sub-units.

 If your population is very small, say 15,000 or less, you should
make sure that there will be enough respondents in each cluster
to interview.  For example, WHO estimates that the target group
for EPI surveys of children 12-23 months of age, averages three
percent of the population in developing countries.   Thus, you
would need clusters of at least 250 people to find seven children
in this age group.  To take account of those who are away,
ineligible, etc.,  WHO suggests doubling that number to 500.  For
30 clusters, therefore, you would need a minimum of 15,000 people
to do a survey of that target group.

 If there are fewer than 30 "natural clusters," then what? For
example, what should you do if there are only 24 villages, or 13
districts.  Remember that these administrative units are not the
sampling clusters.  Follow the same procedures described above to
fill in the Cluster identification worksheet (Exhibit 3).  You
will simply select more clusters per administrative unit.  Try
this with Exhibit 3.  Take the first 20 communities listed and
follow the same procedures.  The total population will be 17,315
and the cluster size (sampling interval) will be 577.  After you
draw a random start number you will add 577 to it to select the
first cluster; then add 577 again to select the second, and so
on.  You will still get 30 clusters and 30 start numbers, but
there will be several communities that will have more than one
cluster.  If you can, it would be best to divide your communities
into smaller administrative units first, say into
sub-communities.  In this way you can avoid having several
sampling clusters in the same community.
 If you are looking for a rare event, say an infant death or a
pregnant woman who received two TT shots, then you may have to
have much larger clusters in order to find them.  To estimate the
cluster size for rare events, you need to know:  the percentage
of the target group in the sample population (e.g., women who
were pregnant last year);  and the percentage of that target
group that has the attribute you are looking for (e.g., two TT
shots).
@SPASI-HALF = 
 An attribute is similar to an outcome indicator. For each PHC
component there are usually one to three important attributes
that a manager is likely to want to measure.  These would have
been identified in Step 2 and would be included in the data
collection instrument.  Some examples are listed below as well as
included in the next worksheet:
Example:  Let us assume that the survey is of antenatal care
(ANC) and that the target group consists of women who had a
pregnancy outcome within the past 12 months.  Local figures may
show that approximately 4 percent of the total population is
pregnant in a year. The survey wants to determine how many of
those women:  1) received at least one tetanus toxoid
immunization during their pregnancy; and,  2)  were delivered by
a trained attendant.  The rough estimates, which might be based
on service records or prior experience, show that 33 percent
probably received TT, and 15 percent were delivered by a trained
attendant.
For the first indicator the survey needs to be able to find seven
women in each cluster who were pregnant in the last 12 months. 
That is estimated as 4 percent of the population.  Thus, the
minimal cluster and sample size would be:

<$E{7~~respondents} over .04> = 175 cluster size * 30 clusters<R> 
                        = 5,250 population

However, this is a minimum size. You would probably want to
double it to make sure that there would be at least 15 people in
each cluster from which to draw. Thus, the clusters should be at
least 350 population each.
The second indicator requires finding seven women in each cluster
who were pregnant and were delivered by a trained attendant. 
Only 15 percent of the women who were pregnant were delivered by
a trained attendant.  Thus, the minimal cluster size would be

<$E{7~~respondents} over {.04~*~.15}> = 1,167 cluster size * 30
clusters <R>                          = 35,000 population

If doubled to be safe, that would require clusters of 2,334 and a
total sample population of 70,000.  
See Appendix G for a computer program (Target.WK1) that will help
you make the estimates of various target groups.
Increasing the number of clusters and the number of respondents
per cluster may help you get more accurate results.  Of course,
that will also increase the cost.  


WORKSHEET FOR IDENTIFYING ATTRIBUTES AND ESTIMATES, +, +, +, 
+, +, +
PHC component, Target population, Percent of total pop., Minimum
cluster size
No. eligible in cluster, +, Size of sample population <R>
(# respondents * # clusters), +
, , , 7, 15, 7 * 30, <%-2>15 * 30<%0>

Antenatal, Pregnant last 12 months
Delivered by trained attend., 4%
15%, 175
1,167, 375
2,334, 5,250
35,000, 11,250
70,000
Family plan., Married women 15-44, 20%, 35, 75, 1,050,   2,250
Growth mon., Child << 2, 8%, 88, 188, 2,640, 5,640
ORT, Child << 5, 14%, 50, 107, 1,050, 3,210
Immunisation, Child 12-23 mo., 3%, 233, 500, 6,990, 15,000
Water/sanit., Household, 100%, 7, 15, 21, 450


Whether it is worth  doing depends partly on how homogeneous the
clusters are.  Homogeneous means that all of the respondents in a
cluster are the same with respect to the attribute you are
studying.  As mentioned previously, this is often the case in
immunization programmes.  In any given cluster either most of the
children have been immunized or most have not been. 
Heterogeneous is the opposite.  It means that the respondents are
different.  For example, some children would be fully immunized,
some would have had one shot, others two, others three, and some
none at all.  
The rule of thumb is: if the clusters are homogeneous, reduce
sampling error by increasing the number of clusters (since all of
the respondents are similar, increasing the number of respondents
will not help); and if the clusters are heterogeneous, reduce
sampling error by increasing the number of respondents per
cluster.
The number of clusters should not be reduced below 30.  This is
the minimum number that is required to produce relatively valid
results.  Do not succumb to the temptation to double the number
of respondents in order to cut the clusters in half.  You can
increase the number of clusters, but there must be at least 30.
Generally, it is easier and less costly to increase the number of
respondents in a cluster, since the interviewers are already
there, than to add more clusters, which means travelling to
another site.  If you can reduce the sampling error from plus or
minus 10 percent to plus or minus 6 percent by increasing the
number of respondents in a cluster, it may be well worth the
extra cost.
See Appendix G for a computer program that can help you estimate
the required sizes for clusters and respondents.
 Multiple target groups
In Step 4 we noted that many managers will want to include
several PHC topics and target groups in the rapid survey. If the
target groups for all of the topics are the same, then there will
be no effect on the sample.  For example, if the survey covers
growth monitoring, immunization, and use of ORT, and the target
group for all three services is children under two years old,
then only one sample needs to be drawn.
However, if the target group for immunization is children 12-23
months of age, then you will not be able to get information on
those under 12 months of age for the immunization portion of your
survey.  Thus, you will need to contact an additional number of
people to find seven eligible respondents.
The previous worksheet can help you to determine the cluster
sizes you will need for each target group. From this you can also
estimate the number of households you will probably have to visit
to find your quotas for each part of the survey.
"For the same overall total sample size, however, a survey in
which a large number of clusters is selected and a few households
visited in each, will give more precise results than a survey in
which a larger number of households is visited in each of a
smaller number of clusters."
A rough rule of thumb is: find the target group in your survey
with the lowest percentage of the target population, and
calculate the minimal cluster size for that group.  Use that as
the minimal size for your survey. Then examine the ratios between
the size of that cluster and the size of the clusters needed for
your other target groups. That ratio will tell you roughly how
many more households you must contact to complete the survey. For
example

PHC service, Target group, Minimum cluster size
ORT, Children << 5 years, 107
Immunization, Children 12-23 months, 500
, , Ratio 500:107 = 4.7
 
You will have to contact about five times as many households to
complete your immunization questions as your ORT questions.  If
you get all your ORT data from the first 10 households contacted,
you will probably have to contact 40 more to complete the
immunization questions.This assumes that you follow the EPI "next
nearest front door" approach. If you have a complete household
listing, you will be able to identify seven eligible respondents
for each part of your survey from this listing.
Again, if you want to find only 20-30 people with the attribute,
you will only have to find one or two per cluster, not all seven.
Again, see the TARGET program in Appendix G for help in making
these estimates.
The last step (selecting the households) will be described in
Step 7.
 Sampling for mortality estimates
If you want to measure infant, child, or maternal mortality rates
or ratios, you will have to modify these procedures.  As
mentioned earlier, you will probably need samples of 2,000
eligible respondents for an infant mortality survey and 7,000 for
child and maternal mortality surveys.  You can conduct a
mortality survey at the same time as a conventional cluster
survey, but you will have to interview all eligible women in each
household, and you will have to visit an additional number of
households to find the 2,000 to 7,000 eligible women.  Use the
SIZE.WK1 worksheet in Appendix G to estimate the size of the
sample you will need.  Use the Vital Events questionnaire in
Appendix C to collect your mortality data.
A recent UNICEF publication describes how to conduct childhood
mortality surveys.<$FDavid, P. H., et al.  Measuring childhood
mortality:  A guide for simple surveys.  Unicef, Regional Office
of the Middle East and North Africa, Amman, Jordan, 1990.> This
handbook contains complete information for formulating
questionnaires, drawing samples, collecting and analysing data,
and preparing reports.  One of the questionnaires from this
handbook is included in Appendix C, Childhood Mortality.  The
instructions for drawing a sample for this questionnaire are in
Appendix G.  The calculation and analysis procedures for direct
estimations are straightforward, but the indirect estimation
techniques described in the handbook are complicated.  You should
call on a trained demographer if you wish to make indirect
estimates.
To measure maternal mortality you will need to find even more
eligible respondents.  One approach, called the sisterhood
method,<$FGraham, W. et. al. Estimating maternal mortality; The
sisterhood method. Studies in Family Planning, Vol. 20, No. 3,
May/June, 1989, pp. 125-135.> involves interviewing all adults in
the household þ  or even in the village or block þ  to identify
everyone who had an adult sister who had been pregnant.  This
method produces indirect estimates of the probability of dying. 
It is relatively simple, and since all adults are interviewed,
you may find enough eligible respondents in 1,000 to 2,000
households.  But, this is still a large number of households to
contact.  In addition, the method is controversial.<$FTrussell,
J. & Rodriguez, G.  A note on the sisterhood estimator of
maternal mortality.  Studies in Family Planning, Volume 21,
Number 6, November/December, 1990, pp 344-346.>
Again, we recommend that you get expert advice before designing a
mortality survey.

Step 6: Schedule the survey
The three most important things to do at this point are: finalise
the data collection schedule; prepare the survey management
forms; and recruit and train the interviewers.
At the end of Step 5 you made some estimates of the "data
collection requirements," meaning the number of interviewers and
supervisors you would need and the number of days required to
collect the data.  You should finalise the data collection
schedule now and prepare an overall schedule for all other
aspects of the study.  This includes the recruitment and training
of data collection staff; the production of the questionnaires;
logistical arrangements to get the interviewers to the clusters;
procedures for checking and verifying the completed
questionnaires, data entry, and analysis; and report preparation.
Survey management forms are important for keeping track of the
numbers of households contacted, the number of call-back visits
made, interviews completed, and so forth.  Appendix H includes
illustrative sample survey management forms for single and
multiple target groups.
Appendix E includes some guidelines for training and supervising
field interviewers.  You should prepare written instructions for
the interviewers that describe exactly how they should select
households, how to identify eligible respondents, which
respondents to interview, when and how often to make call-back
visits, how to check the completed form before leaving the
household, and what to do if a mistake is made.

Step 7: Collect the data
 Selecting households
The selection of the starting household must be made from within
the cluster.  The first house must be selected at random. It
would be best to select all seven households at random.
This may be possible if there is an up-to-date household listing
of the community.  The listing must be up to date, otherwise the
sample will be biased toward old-timers.  Use currency notes or
the random number table in Appendix G to select your seven
households.  To be safe, select ten, just in case there are
refusals, ineligible respondents, or people have recently moved
away.
If there is no list but the community is small, it would be best
to do a quick enumeration of all households and then select the
sample at random.
If that is not possible, the next best approach is the EPI
method.  The typical approach that WHO uses is to select the
first household at random in each cluster, interview an eligible
woman, if there is one at home, and then go to the next nearest
household to find the next respondent.  Respondents who are not
home are skipped, even if they are eligible.  This search
continues until the required number of interviews, usually seven,
has been completed.
The starting household is usually selected by choosing some
central point in the community, such as a market; spinning a
bottle to select a direction at random; walking in that
direction, counting, mapping, and numbering the households you
pass as you walk from the central point to the edge of the
community; and finally selecting one of these houses at random.
This house is the starting point.
There are several improvements on this approach that have been
suggested:

"It would be better to choose, say, the fifth nearest
household..."
 
"In large communities it would be a good idea to spread the
sample around by having more than one starting point in different
parts of the community."
 
"Any method which achieves a random or near-random selection of
households, preferably spread wide apart over the community,
would be acceptable as long as it is clear and unambiguous, and
does not give the field worker the opportunity to make personal
choices which may introduce bias."<$F Bennett, et al. op cit.>
 Multiple target groups
If you draw a random sample, remember to draw a separate one for
each target group.  If a household that you draw includes
respondents for two or more of your target groups, it is quite
alright to collect data from that household for as many of your
survey modules as possible.
If you follow the EPI approach, or the variations suggested
above, you would just keep going from house to house until you
completed all parts of your questionnaire.
 Call-backs
The EPI method replaces respondents who aren't at home with the
next available respondent. It is much better to try at least two
revisits to collect data from the household originally selected.
This maintains the integrity of the sample, whereas replacement
introduces bias.  People who stay at home may be very different
from those who are away at work.<%0>

Step 8:  Enter, verify, and tabulate the data

You may skip this step if:
 You are going to use the Epi Info computer program<R>Appendix A
contains complete instructions for entering, tabulating, and
producing tables and graphs
 You plan to use another computer program for data entry,
verification, and tabulation
This step describes manual procedures for: 1) summarising the
data that have been collected; and 2) producing some simple
tables. Suggestions for using standard spreadsheet programs are
described in Appendix I.
 Manual data entry
If you have been collecting data in cluster registers, you only
need to take the summary tabulations from each of the 30
registers and compile them on a summary form.  In the following
example, the sample cluster register shown in Step 4 is
reproduced with some illustrative data. The interviewer (or
supervisor) will tally the totals in the right column, as shown. 
For simple "yes"/"no" questions count the number of "yes"
responses. For example, five of the seven women received ANC
during their last pregnancy. One received care from a hospital,
two from a health centre, and two from TBAs.
For continuous variables, age, number of visits, add the numbers
and also show the number of respondents.  For example, the ages
of all seven respondents totals 190 years.  The total column
shows 190/7. Later, when the data from all of the clusters are
summarised, these figures can be used to compute average age.
Averages should not be calculated  for each cluster. The number
of ANC visits made by the five women who received ANC is shown as
10/5.
The summaries from each cluster can then be transferred to a
master sheet, such as that shown below. This form would have 30
columns, one (1) for the data from each cluster, and a column to
summarise the totals.
The data from the above form for cluster No. 1 are shown in the
second column from the left marked "1." For example, age is shown
as 190/7, Rec'd ANC is 5, and so forth. The data from each
cluster register would be entered this way and then the totals
produced at the end.
Data collected on questionnaires can be summarised in a similar
manner. This can be done in two steps or one. The two-step
process would be the same as that shown above.  Each interviewer
or supervisor would summarise the data from a cluster on a form
similar to the cluster register. That would then be summarised on
a cluster summary form and tabulated.
The one-step alternative is to develop a large summary form with
210 columns, or as many as there are interviews.
This is especially easy to do on a spreadsheet. Appendix I
includes two computerised forms that are designed for this
purpose.
 
Verifying and cleaning the data
The data need to be "verified" to make sure that no mistakes were
made in summarising the totals and transferring them to the
summary sheets.
This can be done by having two separate teams independently
summarise and transfer all of the data to summary tables. The
results are then compared.  Discrepancies can be checked and
corrected fairly easily this way.
Cleaning the data involves correcting mistakes in the original
interview forms and summary sheets.  Although the supervisors
probably checked each questionnaire and cluster register,
mistakes can still happen.  Typical mistakes include using the
wrong code, leaving a question blank, misinterpreting a written
code (e.g., 0 for 8), skipping to the wrong question, and
entering an answer in the wrong space.
Some of these mistakes will be caught by the supervisor, others
by the verification process, and some won't be noticed until the
preliminary analysis is done.  To find the source of the error,
you will usually have to go step-by-step back through the data
entry process: first to the summary sheets, then the cluster
forms, then the original questionnaires or registers.

Tabulation
Manual tabulation will usually be limited to summarising counts,
computing a few averages, and preparing some frequency
distributions. If you are using a computer, you can do much more.
See Appendix I for examples of simple computerised tabulation
procedures that use a spreadsheet program.
For manual tabulation, you will already have totals summarised in
the Summary Form.  Use these and the report outline that you
prepared in Step 3 to decide what to prepare. At that time you
also prepared some dummy tables. Prepare the data needed to fill
them in. Appendix I includes an illustrative list of frequency
distributions and cross-tabulations.

Interviewer name: B. Rangka, +, +, Date : 03/04/92, +, Book no.
1, +, +, +, +
Cluster: 1, +, +, Area: So. Bajju, +, Supervisor: Sonia Barang,
+, +, +, +
Respondent no., 1, 2, 3, 4, 5, 6, 7, 8, Total
Name, JB, KH, TD, MJ, NKT, TR, JN, , 
House no., 13, 25, 37, 38, 42, 54, 65, , 
Age, 23, 25, 22, 34, 42, 18, 26, , 190/7

Rec'd ANC last pregnancy, Y, Y, N, Y, Y, N, Y, , 5
No. ANC visits, 2, 2, , 1, 3, , 2, , 10/5
Source of service:, , , , , , , , , 
Hospital, Y, , , , , , , , 1
Health centre, , Y, , Y, , , , , 2
Local TBA, , , , , Y, , Y, , 2
Other:, , , , , , , , , 


Summary cluster form, +, +, +, +, +, +, +, +, +
Date of first interviewer: 03/04/92, +, +, +, +, Date of last
interview: 
03/21/92, +, +, +, +
Area: Bajju, Prepared by : Sonia Barang, +, +, +, Checked by:
Marcus 
Stefensen, +, +, +, +
Cluster No., 1, 2, 3, 4, 5, , 29, 30, Total
Age, 190/7, 185/7, 210/8, 175/7, 197/7, , 180/8, 180/7, 5715/214
Received ANC last pregnancy, 5, 4, 6, 3, 4, , 5, 5, 156
No. ANC visits, 10/5, 10/4, 13/6, 9/3, 10/4, , 12/5, 10/5,
314/156
Source of service:, , , , , , , , , 
Hospital, 1, , 1, , , , 1, , 15
Health Centre, 2, 1, 2, 1, , , 1, 2, 47
Local TBA, 2, 3, 3, 2, 4, , 3, 3, 94
Other:, , , , , , , , , 
Averages: Let's start with computing averages.  All of the data
that are continuous variables were entered in the summary table
as two numbers: the total years, visits, and events divided by
the total number of those who responded to that item. Examples
are age and number of ANC visits.  Just perform this division to
compute the average (mean): 

5715/214 = 26.7 years (the average age of your sample)
No. ANC Visits 314/156 = 2.01 visits (the average number of
visits made by women who received ANC during their last
pregnancy)

Please note that the denominators (the number of respondents) is
different in these examples.  A common mistake is to use the
wrong denominator, for example, dividing the total number of ANC
visits by 214, the total sample.  Be careful to use the correct
denominator.

Coverage percentages:  Your most important data will probably be
coverage data. This is computed by counting the number of people
covered and dividing it by the number of eligible respondents.
Examples from the Summary Form are the number who received ANC. 
The sheet shows that 156 women out of 214 received ANC.  To
compute the coverage percentage, divide 156/214 and multiply by
100.

Received ANC last pregnancy: 156/214 = .72897 * 100 = 72.9% (the
percent of eligible women covered)

Frequency distributions:  This information will also be important
for determining the numbers and percentages of people who use
different services, use different providers, have different
reasons for accepting a service, and so forth.  The Summary Form
shows that there were three sources of service for the 156 women
who received ANC.  To compute the frequency distribution, divide
the counts of each of these by 156 and multiply by 100.
You should go through your Summary Form and compute these three
types of statistics for your report. It is possible to do more
sophisticated tables by hand, such as cross-tabulations of ANC
use by age and computation of tests of significance, but these
are tedious when done manually.  If you need this information, it
would be best to use a computer programme or find someone who has
a computer and can do it for you.  They may want to use the
programmes included in Appendix I for tabulation and statistical
analysis.  Appendix I.3 also includes examples of a programme
developed by Ralph Frerichs for computing confidence intervals
for selected indicators (ANC.WK1).

Step 9: Analyse, interpret and report the findings
Consult your report outline again to make sure that you know the
user's most important indicators, questions, and issues. You have
already produced some of these in Step 8: coverage data and
frequency distributions of key variables, for example.
 Analysis and interpretation
When you have your tables completed, examine them to make sure
that you understand what they mean. Think in terms of different
kinds of interpretations:
 Descriptions: This is the most basic level of analysis.  Simply
present the facts: X number of women were served; Y percent
covered. 
 Performance: This requires comparing the descriptive data with
performance expectations or standards. Is 73 percent coverage
adequate, excellent, below expectations? Where is the programme
performing well? Where is it not?<%0>
 Explanations: Some of your frequency distributions can provide
explanations. Where did women go for service?  What reasons did
they give for not coming back? Rapid surveys usually do not get
into much detail in this category, but the results can stimulate
staff discussion and result in insightful explanations. For
example, why don't women come to the hospital? Why are so many
going to TBAs? Modules 6 and 7 can help you get some explanatory
data.
 Implications: This involves going beyond the data to think of
implications for the future.  If coverage is low, what does that
mean for the future? Can it be increased?  Is it feasible? How
could it be done? Should the programme try to get more women to
come to hospitals and health centres for ANC? Or should we train
TBAs to provide better services?
 Issues needing further study: Most studies raise questions, as
well as answer them. Rapid surveys are no different. Identify
questions that cannot be answered with the available data. Some
of those might be investigated with a second survey or through
one of the other modules that look at the quality of services
(Module 6), short-term monitoring of an activity (Module 5),
identifying high-risk women (Module 3), etc.
Remember that you can only do an analysis of the entire sample.
You cannot divide the sample into sub-sample<%0>s to compare
groups. This requires separate rapid surveys for each group.

Source of service, Number served, Percentage
Hospital, 15,   (15/156)*100 =   9.6
Health centre, 47, (47/156)*100 = 30.1%
TBA, 94, (94/156)*100 = 60.3%
Total, 156,                 100.0%


 Reporting
The easiest way to prepare a report is to present the findings
from each question in the sequence followed in the questionnaire. 
Another is to present the major findings first, since this is
what most managers want to know.  Most rapid survey reports are
presented orally at first and include a few tables and graphs of
key findings. The final written report may include additional
tables at the manager's request.
A typical formal outline of a research report is shown below:<%0>
 Title, authors, date, acknowledgements, table of contents
 Executive summary (key findings, implications, brief description
of the study design)
 Statement of the research problem (the background and problems
to be investigated)
 Study objectives (purpose of the study, expected outcomes)
 Methodology (brief description of the key indicators, sample,
instruments used, analysis procedures, timetable)
 Findings (summary of findings, divided into sections, formatted
to address questions and issues the user wants answered þ  should
include tables and graphs of key points)
 Discussion (interpretation of the findings, discussion of
implications for the future, identification of other issues
needing study or analysis)
Recommendations (suggested courses of action to take for policy,
planning, services, management, further research)<%0>
 Appendices (detailed data tables, questionnaire or register
used, reference materials)
The formal report does not have to be long.  Some of the sections
listed above can be covered in a paragraph, others in a page. 
The longest sections are likely to be the findings.  Keep the
audience in mind when preparing the report.  Managers, in
particular, are not likely to read long reports.  That's why the
executive summary is so important.  That may be all that they
read.

Step 10: Develop an action plan
The Implications and Recommendations sections of the formal
report may not necessarily lead to action. That is why it is
important to make development of an action plan a separate step
and to introduce it as part of the study objectives. The managers
and other users should be expecting to take action on the study
results. Make that a stated expectation from the beginning and
reinforce it throughout the study.
The best time to begin preparing an action plan is during the
oral presentation of the study results. The plan does not have to
be detailed, but it should include:
 WHAT: the action(s) to be taken should be specified (e.g.,
provide ANC training to TBAs, or brainstorm what can be done to
enrol high-risk women in ANC).
 WHO: the specific people who will be responsible for each action
should be identified (by name or position).
 WHEN: the dates for starting and/or completing the actions.
In some cases it may be important to include WHERE, to specify
the sites or locations where the actions will take place, HOW, to
outline the procedures that will be followed, and the RESOURCES
that will be made available to carry out the actions.
Specific details may need to be worked out later, and even some
of the above elements may have to wait until the formal report is
ready and can be studied more carefully.  If so, then they should
be incorporated into the preliminary action plan.


Worksheet for developing action plans, +, +, +
ACTION TO TAKE
(What), RESPONSIBLE
(Who), DATES
(When), OTHER (Where)
Appendices: Templates, tools, guidelines, and computer programs
A. How to use Epi Info to conduct rapid surveys
B. Questionnaire design guidelines
C. Rapid survey instruments
 Community assessment of PHC (overall)
 Health education
 Antenatal care, safe delivery and postnatal care
 Family planning
 Acute respiratory infection
 Breast feeding
 Diarrhoeal disease control/oral rehydration therapy
 Childhood disabilities
 Child immunization
 Growth monitoring/nutrition education
 Water supply, hygiene and sanitation
 Accidents and injuries
 Chronic, non-communicable diseases
 Tuberculosis
 Malaria
 Sexually-transmitted diseases, HIV/AIDS
 Vital events and health status
 Child morbidity and mortality assessment
 Adult morbidity and mortality assessment
 
D. Cluster survey registers
 Health education
 Antenatal care, safe delivery and postnatal care
 Family planning
 Acute respiratory infection
 Breast feeding
 Diarrhoeal disease control/oral rehydration therapy
 Childhood disabilities
 Child immunization
 Growth monitoring/nutrition education
 Water supply, environmental hygiene and sanitation
 Accidents and injuries
 Sexually-transmitted diseases, HIV/AIDS
 Malaria
 Tuberculosis
 Chronic, non-communicable diseases
 Vital events and health status
 Child morbidity and mortality assessment
 Adult morbidity and mortality assessment
 
E. Guidelines for training and supervising interviewers
F.Cluster sampling programmes
 Cluster identification worksheet
G.  Other sampling tools
G.1  Estimates of target group sizes (TARGET.WK1)
 G.2 Sample size estimation for WHO two-stage cluster survey 
 (SIZE.WK1)
 G.3 Random number table
 (RANDOM.WK1)
 G.4 Random sampling procedures
 H.  Survey management forms
 Form 1: Household enumeration
 Form 2: Respondent disposition
 Form 3: Multiple-target group management form
I. Tabulation and analysis templates
 I.1 Data analysis plan
 I.2 Rapid survey analysis template (RAPID_ANC.WK1)
 I.3 Cluster summary template (MINI_GM.WK1)
 I.4 Confidence interval estimation templates (TT.WK1; ANC.WK1)
Appendix A: How to use Epi info for rapid<R>surveys
What is Epi Info?
Epi Info is a simple computer program for surveys.  You can
design a questionnaire right on the computer screen or import one
from an unformatted (ASCII) file.  You can edit your questions
right on the screen þ  add, delete, move, and insert.
It allows you to insert the codes, skip patterns, range checks,
and directly apply other data entry instructions to the
questionnaire, as well as enter the data from your completed
interviews onto the questionnaire.  You can also import data from
a spreadsheet or dBase file for analysis.
The program includes simple commands for analysing the data, and
you can produce frequency distributions, tables, cross-tabs, and
more sophisticated statistical analyses right on the screen. You
can even produce bar and pie charts, and print the results.
In short, Epi Info is an all-in-one package for practically any
survey your project needs. You can set up the questionnaires,
data entry instructions, even the analysis instructions here and
give them to your colleagues in the field.  All they will have to
do is print out and duplicate the questionnaire, collect and
enter the data, and run the analysis program to get the results.
This document provides instructions for using Epi Info and
includes a prototype survey questionnaire that has been set up to
conduct a family planning contraceptive prevalence survey.
 What this appendix includes
 Instructions that describe how to:
 Install Epi Info and use the tutorials
 Develop a questionnaire in Epi Info
 Include data entry instructions in the questionnaire
 Enter data into Epi Info
 Analyse the data in Epi Info
 Produce tables and graphs for a report
@PGBRK = 
 Sample files that you can use to illustrate the<R>instructions:
 RAPIDFP.QES (a family planning questionnaire)
 RAPIDFP.REC (a data entry file)
 RAPIDFP.CHK (a file for including range checks and other
instructions for data entry)
 RAPIDFP.PGM (a file of analysis instructions to produce tables
and graphs)
 FPTEST.WK1 (a Lotus 1-2-3 spreadsheet file with sample data)
 FPTEST1.REC (a data entry file constructed from FPTEST.WK1)
@SPACI KECIL = 
Use the sample files that come with these instructions in
conjunction with the Epi Info computer program to learn how to
use Epi Info in conducting rapid surveys in family planning and
primary health care.  You can also read and print out sections or
all of the Epi Info User's guide from the enclosed diskette.

Installing EPI info and using the tutorials
Installation:  Epi Info is very easy to install. Follow the
instructions in the README file on the first PHC MAP diskette. It
will tell you how to extract Epi Info and other files.  When you
have extracted the Epi Info files and put them on separate
diskettes you can install Epi Info onto your computer.  Place
Disk 1 of the Epi Info System in one of your floppy drives, say
drive A.  At the A prompt type INSTALL.  Then just follow the
directions on the screen. See Chapter 4 of the Epi Info User's
guide for more details.

Notations:  These instructions use the same notations as the Epi
Info User's guide (see p. 7 of the guide).  Keys on the
computer's keyboard are indicated by <<   >>. Examples : << F1 >>
(means press the F1 key), << Ctrl + S >> (means hold down the
Control key and press the S key).  The material that you should
type is shown in boldface. Example: type EPI and press <<Enter>>.
Tutorials:  Epi Info includes a tutorial program in a file called
EPIAID.  First get into the EPI5 directory.  Then type EPI to
load EPI5.  The main menu should appear.  Move the cursor to EPED
and press <<Enter>> to load it. Then press <<F3>>.  Next move the
cursor to EPIAID and press <<Enter>>. Choose one of the tutorial
programs þ Word processing, Make Epi Info Questionnaire þ  and
press <<Enter>>.
There are also two analysis tutorials. Load EPI5 as before. Move
the cursor to ANALYSIS on the menu and press <<Enter>>. Then type
RUN TUTOR1 and <<Enter>> for basic analysis procedures or RUN
TUTOR2 for developing an analysis program.
The User's guide also includes Tutorial Instructions at the
beginning of Chapters 4 (Installation), 5 (Starting Epi Info), 8
(Entering Data), and 10 (the CHECK program). It also includes a
number of tutorials for advanced work with Epi Info.
 
The following summarises the basic tutorial programs included in
Epi Info:
 
 Tutorials  Source
 1. Installation    Chapter 4 (page 1)
 2. Starting        Epi Info Chapter 5
 3. Word processing EPIAID: Word processing
 4. Designing questionnaires EPIAID: Make EPI info questionnaire
 5. Data entry    Chapter 8
 6. Data entry instructions Chapter 10
 7. Analysis  ANALYSIS: RUN TUTOR1
 8. Analysis programs ANALYSIS: RUN TUTOR2

These instructions are based on Epi Info, Version 5: A
wordprocessing, database, and statistics system for epidemiology
on microcomputers, by A.G. Dean, J.A. Dean, A.H. Burton, and R.C.
Dickers.  Epi Info is a joint project of the Centers for Disease
Control (CDC) and the World Health Organization (WHO).
The User's guide and the computer programs are in the public
domain and may be freely copied, as can these instructions.  The
program and User's guide can be purchased for $35 from USD
Incorporated, 2075A West Park Place, Stone Mountain, GA 30087. 
The User's guide is also available from several countries, often
through Departments of Epidemiology in local schools of public
health.
 How to develop a questionnaire in Epi Info
You may construct a questionnaire directly in Epi Info or import
one that you developed on a word processor.  We recommend that
you develop (or revise/edit) your questionnaire on your word
processor first.  It will be easier for you to work on a word
processor with which you are already familiar. Epi Info uses a
simple word processor based on WordStar.  It may take some time
to get used to Epi Info's wordprocessing commands. We strongly
recommend that you read Chapters 6 and 7 of the Epi Info User's
guide for instructions.  Also, run the EPIAID tutorials on Word
processing and make the Epi Info Questionnaire.
The following instructions first tell you how to prepare a
questionnaire on your word processor (A). This is followed by
instructions for preparing one within Epi Info (B).

How to prepare a questionnaire that will be imported to Epi Info
1. Prepare your questionnaire on your word processor.
Use one of the draft questionnaires in Appendix C as a guide.
Also see the sample Family Planning questionnaire that follows. 
You may load these questionnaires into your word processor and
revise them by adding, deleting, or editing the questions.  List
the most likely responses and give each a code number. Make sure
to include 9 or 99 for "Don't Know/No" response (DK/NR). Example:

8. Are you using a method now? Yes (1),  No (0), DK/NR (9)

When you are finished, save the questionnaire twice.  Save it
first as a regular document, for example RAPIDFP.DOC.  You will
use this version for your interviewers.
Save it a second time as an unformatted file (ASCII).  You will
use this version to set up a data entry programme in Epi Info. 
When you save it, add QES as the extension.  Example: RAPIDFP.QES

2. Enter the Fields.
Now you will enter "fields" in the *.QES version of the
questionnaire.  You should still be working in your word
processor.

Types of fields.  A "field" is an area in the questionnaire where
you will enter data.  After each question you will insert symbols
for an appropriate field. These are examples of questionnaire
items followed by different kinds of fields:
 Today's date: <</mm/yy>>  Interviewer name ______
 2. How old are you? ##   3. Are you married? <<Y>>

The first is a "date" field. You would enter two numerals each
for the day, month, and year. The second, to the right, is called
a "string" field.  You can enter up to 80 characters in a string
field.  The third, following, "How old are you/," is a "numeric"
field.  The # symbol represents a numeral, in this case, a
two-digit number.  The last is a field for entering "yes"/"no"
responses.

Symbols. The following list summarises the major symbols you can
used for your fields.  Usually, each question will have one field
for the response, and you will decide which type of field is
appropriate and where to place it.  The numeric and string fields
require a symbol for each digit or character in the response. 
For example, if you record the respondent's age, you need to
allow space for two digits, so enter ##.  If you want to record
the person's name, you might allow space for 20 or 30 characters. 
Type one underline character for each space : ________________ . 
(This is a 20-space string.)

#, For numbers:  ## (26)     ###.# (26.4)     .###(.264)
<<Y>>, For Yes/No responses: Y = Yes,, N = No,<R>(Accepts Y,, N,,
and space (missing value. Does not allow DK/NR codes.)
<<dd/mm/yy>> , For dates:  23/04/92
_______, For written responses (especially for open-ended
questions) maximum length is 80 characters.
<<IDNUM>>,  For the case number:  0231

Instructions. Type in the appropriate symbols in the second
(*.QES) version of your questionnaire.  Put them where you want
the data to be entered. Later you will use this questionnaire to
construct a file for entering data into the computer.  The
computer program will display this questionnaire, and the cursor
will skip from one field to the next to allow you to enter data.  
Thus, it is important where you place the symbols. In the sample
questionnaire, we have placed all of the symbols in the right
margin.  You may also place them at the end of the questions,
below them, inside them, wherever you wish.  For example:
 
Are you using a method now? # Yes  (1), No  (0) DK/NR  (9)
Are you using a method now?  Yes  (1), No  (0) DK/NR  <T>(9) #
# Are you using a method now?<T>Yes  (1), No  (0) DK/NR  <T>(9)


Remove any symbols that you don't want to be read as fields by
the computer.  Underlines ___, chevrons <<2>>, pound signs #3,
will all be read as fields.  In the following examples you would
remove all of these symbols except the one # sign at the end,
which is the field you want.

Are you using a method now? <T>Yes___(1), No___(0),
<T>DK/NR___(9) #
Are you using a method now? <T>Yes <<1>>,    No <<0>>,  <T>DK/NR
<<9>>#
Are you using a method now? <T>Yes (#1),    No (#0), <T>DK/NR
(#9) #

Rapid survey questionnaire: Family planning
Complete for all married women aged 15-44 years who are currently
living in the household.
CASE NO:  IDENTIFICATION  , +, +, +, , , Office Use<<ID number>>
     
1. , {Study} No        , +, , , ,  1.##
2. , {Province} No       , +, , , , 2.##
3. , {Interviewer}      , +, , , , 3.##
4. , {Date} of Interview        , +, +, , , 4. 

, , +, +, , , <<dd/mm/yy>>

5. , ID Number (4 digits) : {Cluster No}. {Woman No}. in Cluster,
+, +, , , 5.##

, {NAME} OF RESPONDENT:, +, +, , , 
6. , How {old} are you? (Probe) years (if DK/NR,, enter 99), +,
+, , , 6.##
7. , How many living {children} do you have? (if DK/NR,, enter
99) , +, +, , , 7.##
8. , Are you or your husband currently using any family planning
{method}? , +, +, , , 8.#  
    Yes (1)     No (0)   Go to Q 13    DK/NR (9)  Go to Q13, +,
+, +, , , 
9. , Which method are you/your husband using now (select
principal {method only})?, +, +, , , 9.##
, Tubectomy, (1), NORPLANT, (6), , 
, Vasectomy, (2), Condom, (7), ,      <%-5><<dd/mm/yy>><%0>
, IUD, (3), Foam,, emco,, jelly,, cream,, diaphragm, (8), , 
, Oral pill, (4),  <%-2>Safe period,, withdrawal,, abstain<%0>,
(9),  , 
, Injection, (5), Other:       , (10), , 
,  DK/NR, (99), , , , 
10., {How long} have you been continually using this method?, +,
+, , , 10.#  
, 0-3 months, (1), 1-2 years, (4), , 
, 4-6 months, (2),  3-4 years, (5), , 
,  7-12 months, (3),  5 years or more, (6), , 
, DK/NR, (9), , , , 
11., For how long have you been practising family planning,,
i.e.,, continuously using one method or another without
{interruption}?, +, +, , , 11.#  
, 0-3 months, (1), 1-2 years, (4), , 
, 4-6 months, (2), 3-4 years, (5), , 
, 7-12 months, (3), 5 years or more, (6), , 
, DK/NR, (9), , , , 
, , +, +, , , 
, , +, +, , , 
12., What is the main source of your family planning service or
{supplies}?, +, +, +, +, 12.##
,  Govt. hospital/clinic, (1) , Private hospital/clinic, (6),  , 
, Govt. field worker, (2), NGO clinic, (7), , 
, <%-2>Social marketing prog.<%0>, (3), NGO field worker, (8), , 
, Private physician, (4), Other: _______, (9),  , 
,  Pharmacy, (5),  DK/NR, (<%-6>99)<%0>, , 
, Go to Q17, , , , , 
13., If you are not using family planning now,, have you or your
{husband ever} used any method in the past?, +, +, , , 13.#
, Yes (1), (1), No, (0), Go to Q15, 
, DK/NR , (9) , , , , 
14, Which method did you/your husband use most recently (select
{latest  method} only)?, +, +, , , <R>14.##
, Tubectomy, (1), NORPLANT, (6), , 
, Vasectomy, (2), Condom<T>, (7), , 
, IUD   , (3), Foam,, emco,, jelly,, cream,, diaphragm, (8), , 
, Oral pill, (4), Safe period,, withdrawal,, abstain, (9), , 
, Injection, (5), Other, (10), , 
, DK/NR, (99), , , , 
15., Do you/your husband intend to practice family planning in
the {future}?, +, +, +, +, 15.#
, Yes, (1) , No, (0), , 
, DK/NR , (9), , , , 
16., What is the most important reason you are not using family
{planning now}?, +, +, , , 16.##
, Want more children, (1), <%-6>Method/service unavailable<%0>,
(6), , 
, Husband objects, (2), Sterilily, (7), , 
, Health reasons, (3), Breast feeding, (8), , 
, Religious reasons, (4), Pregnant, (9), , 
, Fear side effects, (5), Other, <%-2>(10)<%0>, , 
, DK/NR, <%-2>(99)<%0>, , , , 
17., What is the name of the local {Community} Health Worker?<T>,
+, +, , , 17.#
,  Knows (said name) , (1), Does not know, (0) ,   , 
, No response, (9), , , , 
18., Has the Community Health Worker visited or contacted you
during the {last three months}?, +, +, , , 18.#
, Yes , (1), No, (0),   , 
, DK/NR, (9), , , , 
This concludes the interview.  Thank you for taking the time to
participate in this survey., +, +, +, +, +, +


In this version the data are entered in a column on the right;
many researchers prefer this format.  However, it is important
for only the data entry version of the questionnaire. The
interviewers can enter responses anywhere on the paper. That is
why you should save one version of the questionnaire for your
interviewers and this second one for the data entry program.

3.Enter names for each field by placing { } around key words in
each question.
You should still be working in your word processor on the *.QES
version of your questionnaire.
Next you must give each field a name so that you can analyse the
data later. Examples:
How {old} are you?  Field name = OLD
Which {method} are you using {now}? Field name = METHOD NOW
{A.1}  Which method are you using now? Field name = A.1
If your question is longer than one line, the field name must be
in the last line.  For example:
8. Are you or your husband currently using any family planning
{methods}?
 
Actually, you do not need to enter field names. This is an
optional step, because Epi Info will create field names
automatically if you do not specify a name.  It selects the first
10 non-punctuation characters before each field.  See page 53 of
the Epi Info User's guide for more detail.  The major advantage
to entering your own field names is that they will be immediately
recognisable to you.
When finished, save the file in ASCII format with a QES
extension.
Example: save the file as RAPIDFP.QES

How to prepare a questionnaire within EPI info.
1. Load EPI5.  At the C prompt type <<EPI>> and press <<Enter>>.
2. Open EPED.  In the main menu move the cursor to EPED and press
<<Enter>>.
3. Type the questionnaire onto the screen.
Type the questions, instructions, codes, etc., directly onto the
screen, using the program's wordprocessing commands to tab,
backspace, delete, 
If you have a questionnaire on your disk that you want to use or
edit, load the file by pressing <<F2>>.  Enter the file location
and name, e.g., b:\RAPIDFP.QES. Press <<Enter>>. Then make your
corrections, additions, etc.
4. Enter the fields and field names, and remove unwanted symbols.
You can enter the field symbols directly, as described above. 
You can also call up a menu of field symbols by pressing <<CTRL +
Q>>, then <<Q>> again.  To insert one of these in your
questionnaire, first make sure the cursor is where you want to
make the insertion.  Then press <<CTRL + Q>>, Q, then highlight
the symbol you want and press <<Enter>>.  If you know what the
symbols are, it is easier to type them in directly.

You may skip this step if you don't want to insert any data entry
instructions

5. When finished, save the file. Press <<F9>>, then <<F10>> to
exit.
 Inserting data entry instructions in Epi Info
(See Chapter 10 of the Epi Info User's guide)
You can insert codes into your basic questionnaire to check for
errors, to do automatic coding of some entries, and to skip over
inappropriate questions. The instructions describe in this
section how to insert the following into your questionnaire:
Range Checks:  Specify the range of values that can be entered;
e.g., 1-5. The programme will reject all numbers outside the
range, e.g., Q, 6, 8 etc.
Legal Values:  Specify individual values that can be entered;
e.g., 2, 8, 9. The program will reject all other numbers (e.g.,
1,3,4, N, etc.).
Must Enter:  Specify that an entry must be made in the field, it
cannot be left blank or skipped.  The program will not move to
the next field until an entry is made. However, you can override
this by pressing the <<Down Arrow>> or <<page Down. >>
Repeat:  Specify that the same value entered into a field will be
repeated in all subsequent records until it is changed.  For
example, enter Q 3 for province, which will be entered
automatically on all subsequent questionnaires until you change
it.
Skips: Specify which questions will be skipped, depending on the
previous answer.  For example, "If No, go to Q16."

1. Create a data file.
Before you can enter these instructions, you must have a data
file in which to insert them. This is a file with a *.REC
extension. You can create one as part of this step or as part of
the data entry step.
Load EPI5 by typing EPI and pressing <<Enter>>. At the main menu,
move the cursor to ENTER and press <<Enter>>.
At the prompt type in the name of the questionnaire file you just
prepared, e.g., RAPIDFP, but leave off the extension.  Don't
forget to include the path name.  Example: b:\RAPID and press
<<Enter>>.
Follow the commands. Press <<ESC>>. Enter the name of your
questionnaire file, e.g., b:\RAPIDFP.QES and <<Enter>>.
Inspect the file, especially to make sure that all of the fields
are included and are in the correct places. Then press <<F10>> to
exit. If you need to make corrections, go back to the EPED
programme to edit your file.
 
2. Load the CHECK programme from the main Epi Info menu.
Load EPI5 by typing EPI and pressing <<Enter>>. At the main menu,
move the cursor to CHECK and press <<Enter>>.
@SPACI KECIL = 

3. Load your questionnaire.
You should load the data entry version of your questionnaire.
That is the one with the *.QES extension.  At the prompt, enter
the path and name of your questionnaire (e.g., b:\RAPIDFP.QES)
and answer <<Y>> to the question "Are you ready?" Press
<<Enter>>.
 
4. Enter the appropriate check codes in the appropriate fields.
Place the cursor in the first field to be changed and make the
appropriate entry; see the menu at the bottom of the screen. 
Move to each field to be changed until finished.
Range check:  Enter the minimum number that will be accepted,
press <<F1>>, then enter the maximum number that will be
accepted, press <<F2>>.
Example: Age 15-44.  Put the cursor on the field symbol in the
age question, type: <<15>> <<F1>> <<45>> <<F2>> and <<Enter>>. To
remove a range, see "6. Editing the commands," below.
Legal Values:  Enter the letter(s) and/or number(s) that will be
accepted, press <<F6>>.
Example:  Male=M, Female=F.  Put the cursor on the field symbol, 
type: <<M>> <<F6>> <<F>> <<F6>>, and <<Enter>>.
Example:  Male=1, Female=2, Unknown=9; Type:<R> <<1>> <<F6>>
<<2>> <<F6>> <<9>> <<F6>> and <<Enter>>.
Press <<SHIFT + F6>> to display all legal values for the field.
To remove a legal value, enter it into the field and press <<CTRL
+ F6>>.
Must enter data in this field:  Place the cursor in the field and
press <<F4>>. Press <<F4>> again to remove the command.
Repeat entry made in this field in subsequent records:  Move the
cursor to the field symbol, enter the number or characters you
want repeated, and press <<F3>>. Example:  Study No <<6>> <<F3>>. 
Press <<F3>> again to remove the command.
Skip:  Place the cursor in the field.  Enter the value that
triggers the skip, press <<F7>>, move to the question to be
skipped to, press <<F7>>.
Example:  Q 8 says "If 'No', go to Q13."  Place the cursor on the
field symbol in Q8. Type:  <<N>>, press <<F7>>, (move the cursor
to the field symbol in Q13, press <<F7>>.
You can have several different skips for the same question.  For
example, in addition to the skip above, you might have: "If Yes,
go to Q15."  Place the cursor on the field symbol in Q8 again,
type <<Y>>, press <<F7>>, move the cursor to the field symbol in
Q15, type <<F7>>. The program will now skip to Q13 if the
response is "no" and to Q15 if it is "yes."
If you want to skip to another question regardless of the
response, place the cursor on the field symbol and press <<F7>>
when the field is blank.  Move to the field symbol in Q15 and
press <<F7>> again. To display all current skips for a field,
place the cursor on the field symbol and press <<SHIFT + F7>>.
To remove a skip use <<CTRL + F7>>.
 
5. Save your changes.
Press <<F10>> when finished to save. "Write Data to Disk [Y/N]?"
appears.  Press <<Y.>>

6. Editing the commands.
Place the cursor in a field you wish to change and press <<F9>>. 
The commands for that field will be displayed in an indented
hierarchy.  You can edit, add, delete commands directly. 
Example: To change the Male and Female codes from M and F to 1
and 2, simply press <<F9>>, move the cursor to M and replace with
1, move to F and replace with 2.

How to enter data into EPI info 
(Chapters 8 and 17 of the Epi Info User's guide)
There are two ways to enter data into Epi Info. The first is to
enter the data from one questionnaire at a time directly into the
Epi Info data file.  The second, is to import all of the survey
data from a spreadsheet, dBase or ASCII file.
 
How to enter data from questionnaires (see Chapter 8)
1.  Load ENTER from the main menu.
Load EPI5 by typing EPI at the C prompt and pressing <<Enter>>. 
At the main menu move the cursor to highlight ENTER and press
<<Enter>>.  If you are creating a new data file, type in the path
and name of the questionnaire file that you want to use, but
leave off the extension, e.g., a:\RAPIDFP and press <<Enter>>. 
The program will create a data file and give it a *.REC
extension, such as RAPIDFP. REC.   If you are loading a data file
that has already been created, follow the same steps.  When you
type the name of the data file (RAPIDFP), the program will find
RAPID.REC and load it.  Or you can type in the path, e.g., b:\
and press <<F1>> to display a list of REC files.  Move the cursor
to the one you want to load and press <<Enter>>.

2. Enter the data from each completed questionnaire.
The case identification number <<idnum>> is entered
automatically, and the number increases by 1 for each record. 
The cursor will move to the first field that you specified to
receive data.  In the sample Family Planning questionnaire, this
field is "Study No"  Type in the number (e.g., 6) and then press
<<F3>>.  Since the study number is the same for all
questionnaires, you need to enter that only once.  The <<F3>> key
turns the "Repeat" function on and off.  When you turn it on, you
will not have to enter the same number for each questionnaire.
The Province number is also the same.  The interviewer number
will be the same for 7 or more cases, as will the date of the
interview and cluster number.  When the number changes, e.g.,
from 6 to 7, just type in 7 and it will repeat until you type in
a new number.
In most cases, after you enter data in a field, the cursor will
automatically move to the next field. If it does not, press
<<Enter>> to move to the next field.  This happens when the
spaces in the field are larger than the number entered.  For
example:  "How many living children do you have?"  Allow 2 spaces
for a two-digit number (12, 15, 10).  When you enter two digits,
the cursor goes to the next variable.  If you entered one digit
(2, 3, 1), you would also need to press <<Enter>> to move to the
next question.
If you have installed range checks and legal numbers, the program
will only accept the numbers you indicated.  "How old are you?"
will accept any number between 15 and 45. If you enter 13 or 54,
the program will not accept the entry.  For a "yes"/"no"
question, the program will accept Y, N, or blank.  If you enter Q
by mistake, it will not be accepted.  Dates must be entered in
the correct order: dd/mm/yy. The first entry (day) cannot exceed
31, the second (month) cannot exceed 12.
The program will skip questions that do not apply (identified on
the questionnaire as "Go to QXX").  You cannot enter data in
fields that are to be skipped.
Some fields are designated as "Must enter."  The program will
beep if you try to skip by pressing <<Enter>>. If the data item
is missing on the questionnaire, you can leave the field blank by
pressing the down arrow or <<Page Down>>.
 
3. Make corrections, as needed.
If you make a mistake, use the up or down arrows, the <<Home>>,
<<End>>, <<PgDn>>, and <<Pg Up>> keys to move the cursor to the
error and re-enter the correct data.
When you have finished a form, the message, "Write dat