This Lecture was delivered in the First Sri Lanka National Consultation Meeting on MSM, HIV and Sexual Health, 18th – 21st November 2009
Organised and conducted by Companions on a Journey and Naz Foundation International.
Ahmedabad Call Girls CG Road 🔝9907093804 Short 1500 💋 Night 6000
Size estimation of most at risk populations
1. Size estimation of most at-risk
populations
Dr. Don Ajith Karawita
MBBS, PGD VEN, MD Venereology
National STD/AIDS Control Programme, Sri Lanka
2. Overview of the presentation
1. Overview of population size
estimation methodologies
2. Survey-based methods
3. Mapping-based methods
4. Mapping of MARPs in Sri Lanka
1. Objective
2. Specific objective
3. Methodology
5. (A). Survey-Based methods
1. Study of individuals with high risk (Survey-Based methods)
1. Comparing two independent sources of data on high risk
groups (multiplier method)
1. Program data compare with a probability survey data
(Programmatic multiplier)
2. Distribution of unique object data compare with survey data
(Unique object multiplier)
2. Capture-Recapture method
3. Network scale up
6. Multiplier method: Programmatic multiplier
Source 1: Programme data
MSM population whose size is
going to be estimated Use counts from a pre-existing
programme listing during a specific
time period
e.g. No registered during last 3 m
No attended for STI screening
Source 2: probability survey data
Probability survey of MSM
Survey data should come from a
MSM
representative survey of the
Registere population whose size is being
d in NGO estimated
X
The survey area must encompass the
program listing area i.e. people in the
list have to be eligible for the survey
7. Example of how the programatic multiplier works
If we want an estimate of the size of
the MSM population
MSM population whose size is Source 1: Programme data
going to be estimated
NGO X reports that there were 80
Population size = ? MSM registered in their programme
as of May 2007
Source 2: probability survey data
Probability survey of MSM
In the survey, conducted in May 2007,
Proportion of MSM visited
40% of the respondents report
NGO X = 40%
receiving service from NGO X in the
MSM past year.
Registered
in NGO X
e.g. 80
Use of multiplier formula:
Popu=NGO X No/ proportion of MSM
found in the survey
Population=80/40/100 = 200
8. Problems with field implementation (Bias)
• Source 1: Programme data
– Failure to record beneficiaries
– Failure to remove inactive beneficiaries
– Duplication of data
– Failure to include appropriate beneficiaries (e.g. mixing up
DU and IDU)
• Source 2: probability survey data
– Recall Bias - Failure to recall the service delivery
point/service (Multiple NGOs providing services)
– Respondents who are in contact with interventions more
likely to be identified and sampled (survey is not
representative).
9. Multiplier method: Unique object method
Source 1: Unique object method
MSM population whose size is data
going to be estimated
A known number of “Unique objects”
(T-shirt, Key tag, Purse etc.) are
handed out to eligible individuals prior
to the probability survey usually 1-2
wks before the survey.
Probability survey of MSM
Source 2: probability survey data
No of Survey data should come from a
unique representative survey of the
objects population whose size is being
distributed
estimated
The survey area must encompass
distribution of unique object recipients
10. Example of how the unique object works
If we want an estimate of the size of the
MSM population
Source 1: Unique object data
MSM population whose size is
going to be estimated
The survey team distribute 150 special
key tags to MSW 2wks before the
Population size = ?
survey starts.
Source 2: probability survey data
Probability survey of MSM
Proportion of MSM received
In the survey, respondents are asked
the unique object = 10% whether they received a key tag & are
shown an example of the object.
No of 10% of the survey respondents reported
unique receiving the key tag
objects
distributed
150 Use of multiplier formula:
Popu=No of objects/ proportion of MSM
found received the object in the survey
Population=150/10/100 = 1500
11. Multiplier method: Unique object method
Source 1: Unique object method
MSM population whose size is data
going to be estimated
A known number of “Unique objects”
(T-shirt, Key tag, Purse etc.) are
handed out to eligible individuals prior
to the probability survey usually 1-2
wks before the survey.
Probability survey of MSM
Source 2: probability survey data
No of Survey data should come from a
unique representative survey of the
objects population whose size is being
distributed
estimated
The survey area must encompass
distribution of unique object recipients
12. Problems with field implementation (Bias)
• Source 1: Unique object data
– Object related – Object is not unique if commonly available in shops,
not a memorable one .
– Object distribution related - Giving to non-MSW, MSW outside the
geographic area, non distribution without reporting or keeping by
peer educators, giving more than one object, passing the object to
others, objects are given to people who are more likely to participate
in the survey or giving the object as they are recruited for the survey.
• Source 2: probability survey data
– Recall Bias - Failure to recall the object if recall period is long or object
is not memorable, not showing the object during the survey.
– Respondents who are known to peer educators are more likely to be
given the object and identified and sampled for the survey (survey is
not representative)
13. (B). Mapping-Based methods
1. Study of spots with high risk activity (Mapping-Based
methods)
1. Studying the whole list of spots (Census/Geographic
mapping)
2. Studying hidden networks (Network mapping)
3. Studying a sample of spots (Enumeration)
14. Mapping Vs Survey-based methods
Mapping Multiplier
Types of HRG Good for HRGs that are found in Listing or counts of HRG available. HRG
accessible or public area. Areas can can be sampled
be listed
Main data Site specific sizes and site profiles Overall population estimates for large
applications geographic area
Resources Field workers familiar with HRG Existing service records and / or
requirment areas (NGO, Intervention teams etc) probability sample survey
Key challange Obtaining a comprehensive list of Achieving independence of the two
sites sources of data