1. Estimating size of invisible
risk populations
Dr K.A.M. Ariyaratne
Consultant Venereologist | Coordinator Strategic Information
National STD/AIDS Control Programme, Sri Lanka
3. Why do we need to know the
sizes of key populations?
• To know the HIV epidemic in a country
• For programme planning
• For resource allocation & advocacy
4. Why key population size
estimations are so difficult?
• Hard-to-reach, less visible
• High mobility
• Geographically
• In and out of KPs
• Variable definitions
• Identity, behaviour, culture
• Time frame: ever, last year, last month...
5. PSE methods
A. Collect data directly
from the key populations
B. Collect data from the
general population
1. Census and enumeration
2. Multiplier
3. Capture-recapture
1. Population survey
2. Network scale-up
7. Census and Enumeration
• Develop a complete list of places that the
population may congregate (Mapping)
• Census: Count Key pop. of all places/hotspots in the
list
• Enumeration: Count key poos in a sample of
places/hotspots
8. Census and Enumeration; primary
prerequisites
• The population can be counted in hotspots
• (Visible at least to community members and
concentrated in distinct areas)
• Able to make a map (Mapping)
9. Mapping
Mapping is a process or tool,
not a method of estimation
in itself.
Done in 2 levels
• Level 1 - Data from 2ry
and 3ry key informants
• Level 2 - Interview Key
pops. at hotspots
12. PSE methods
Collect data directly from
the key populations
Collect data from the
general population
1. Census and enumeration
2. Multiplier
3. Capture-recapture
1. Population survey
2. Network scale-up
14. Multiplier method
Compares two independent sources of data
• Data Source 1-
• Service multiplier/Unique object multiplier
• Data Source 2: a representative survey that
measures the proportion of key population belong
to data source 1 as well.
15. M= Multiplier
(MSM who are given a Key tag)
M=5
P= Proportion
(Survey finding: MSM who are
given a Key tag)
P=3/9 (0.33)
Data source 1 Data source 2
Same
period
of time
17. PSE methods
Collect data directly from
the key populations
Collect data from the
general population
1. Census and enumeration
2. Multiplier
3. Capture-recapture
1. Population survey
2. Network scale-up
19. Capture-recapture Method
Based on two independent captures (samples)
• Capture 1: ‘tag’ and count number tagged;
• Capture 2: ‘tag’ and count who is ‘retagged’ and
who is ‘first time tagged’
24. Capture-recapture assumptions
• Two captures/samples must be independent and
not correlated
• Each population member has an equal, or
known, chance of selection
• No major in/out migration occurs
• The sample sizes of each capture must be large
enough to be meaningful
25. PSE methods
Collect data directly from
the key populations
Collect data from the
general population
1. Census and enumeration
2. Multiplier
3. Capture-recapture
1. Population survey
2. Network scale-up
27. General population surveys
Advantages
• Routinely administered in most countries
• Straightforward to analyse
Disadvantages
• Difficult to use when behaviours are stigmatized
or illegal
• Large sample size required for rare behaviours
28. PSE methods
Collect data directly from
the key populations
Collect data from the
general population
1. Census and enumeration
2. Multiplier
3. Capture-recapture
1. Population survey
2. Network scale-up
30. Network scale-up
• Step 1 – Determine personal network size
• Step 2 – Ask the general population how many
people they know in the hidden population
• Step 3 – Calculate the estimated population size
and adjust for known biases
32. Summary
• Many methods - there is no gold standard
• Choosing a method depends on many
factors…
• The most common method - Multiplier
• Better to use different methods and use data
for programmatic decisions
Editor's Notes
Thank you. Good morning to every one. As you know some members of our society have higher risk of getting HIV. They are know as high risk populations, most-at risk populations, key affected population or simply as Key populations. These populations are often hidden due to legal issues and due to stigma and discrimination.
These are some of the high risk populations for HIV infection in Sri Lanka. They represent diverse populations; but one thing is common they are often invisible and hard to count.
Size estimates are necessary know the HIV epidemic in a country. These need to estimate the number of people living with HIV. In addition, the size estimates are needed to predict where future infections will take place.
Also we need sizes of key populations when we are planning programmes such as condom distribution, planning needle exchange or drug treatment programmes
In addition we need these sizes for resource allocation and raising funds from donors
Why is key population size estimations are difficult?
They are often hidden and not visible
These populations move from location to location. For example sex workers in Colombo migrate to Kataragama during Perehara season. Also they can move in and our of risk behaviour. A woman could be a sex worker this year. But next year she can be a migrant worker who may decide to be a FSW few years later.
Size of the population is also depend on the definition. For example, MSM can be defined as a man who had oral or anal sex with a man in the past 12 months. But if we include only men who have anal sex during past 6 months, the number obviously changes.
There are two main categories of methods for size estimations. Category one, data has to be collected directly from Key populations. In category two, data is collected from the general population.
In category 1 there are three methods
In category 2 there are two methods
I am going to briefly explain these 5 methods during the time allocated to me.
Lets talk about census and enumeration first.
In both census and enumeration, we need to prepare a complete list of places where key pop. members gather, these are called hotspots. These places are marked on a map. Therefore this process is named Mapping.
In census all hotspots are reached and number of key pops are countered.
In enumeration, a sample of hotspots are visited and count the number of key populations. Then you apply the average number of key population to all the hotspots to make the size estimation.
To conduct census or enumeration there are some requirements,
Mapping the process of preparing the list of locations of hotspots.
Usually done in two levels. In L1 we interview secondary and tertiary key informants. For example 2ry informants in case of sex workers would be pimps, three wheel drivers. 3ry key informants would be STD staff or police officers
During level 2, key populations themselves and get the count.
Census method start with making a list of all hotspots and marking on a map. Then you visit each and every hotspot and count the number of key populations. E.g. sex workers in brothels. This count has to take place in a very short period of time. Otherwise, key populations can move between sites may lead to double counting.
Enumeration method also start with making a list of all hotspots. But instead of counting key pops. in all hotspots, you count key populations in a sample of hotspots. Calculate an average number and apply it to all locations to get the total population size.
Second is multiplier method
Multiplier method ccompares two independent sources of data for populations to estimate the total number in the population.
Depend on the data source one, there can be two multipliers
Service multiplier (get data from a service center like STD clinic or VCT center)
Unique object multiplier (get data by distributing a unique object)
Data source two is a representative survey to find out what proportion of key population belong to data source one as well.
Lets try to illustrate this.
In the data source one. Lets say researchers have given a unique object such as a key tag to 5 MSM.
They you do a representative survey among MSM and count the number of MSM who have received this unique object. Lets say 3 out of 9 surveyed got a key tag.
So the proportion of the MSM who have received a key tag is 0.33.
“S” is the estimated MSM population. We can calculated this by dividing the multiplier by the proportion using this simple formulae. According to these small numbers, estimated MSM population is about 15.
Now let’s talk about the 3rd method in the first category of size estimation methods.
Capture–recapture method is based on two independent captures of the key pop. members.
Interestingly, this method was first used in 1662 to estimate the population of London.
Since then this is being widely used in wildlife research to estimate number of animals or fish.
During the last few decades, the method has been applied to estimate the sizes of those hidden populations such as sex workers (SWs), MSM and intravenous drug users.
Let’s illustrate this and see. Will say there are number of Key pop in an area. Please don’t count the number as I am going to calculate using capture recapture method.
Now we visit these are and do the 1st capture and count the number there. In practice we can distribute a memorable object. In one study they have give an umbrella to MSM. Idea is to mark or tag them.
Now you should do a second capture and count the number. For example we can visit on a different date and give a different object. In addition to counting the number in the fist capture, we also have count the number also captured both times. In this illustration, there are two persons recaptured when we do the 2nd capture.
This is the mathematical logic to calculate total number of population.
We can calculate two fractions or proportions. The first one proportion of population in the first capture out of total population.
The second one is proportion recaptured out of 2nd capture.
Assuming the capture is random, both fractions should be equal. i.e. population proportion is same as sample proportion
Now we can calculate the unknown population size or “S” as we know other numbers. In other words, you multiply number in the 1st capture by number in the 2nd capture and divide by number recaptured in the second capture.
In capture recapture we should try to meet following assumptions.
Let’s now talk about second category of pop. Size estimation methods where data is collected from the general population
General population surveys like DHS are used to estimate the key populations in some countries. These are not always easy questions to include in a survey given the stigma, discrimination and illegality of these behaviours.
Network scale-up is the other method based on general population data.
The method uses information collected in general population household surveys. However, instead of asking about the respondent’s own behaviour, it asks about the behaviour of the respondent’s acquaintances.
There are three steps in Network scale-up.
This graphically represent Network scale-up method. Red dots represent key pop. members Green circles represent all other people. Yellow circle represent our respondents’ network size. Of the 9 people known to respondent, 2 are key pop members. This proportion is applied to total population to calculate total number of key population members after adjusting for biases.
Choosing a method depend on..
What resources (i.e. budget) are available
Which data is available ? Lists, maps etc
How are we going to use the data?
What are the characteristics of the population ? Are they visible?