APMG’s WORK ON SEX WORK POPULATIONS SIZE ESTIMATION

IN BRIEF: AIDS Projects Management Group is working on developing population size estimation processes for sex workers at the national level. In this short brief, APMG Director Dave Burrows reports on this important development in HIV.

While many estimation methods are available for measuring the size of sex worker population as national level most are not suited to developing and transitional countries. Population size estimation is important for a number of reasons. Most significantly, coverage statistics only make sense if we know the population size. Coverage is expressed as the percentage of a total population which receives or participates in specific services of acceptable quality on a regular basis. While there is still a great deal of debate about which services - or combinations of services - are most important, and how to measure quality and regularity, none of these is useful if we don't first have at least a general agreement on how many people are in the population being covered.

Also, national government, Global Fund, WHO and UNAIDS processes - such as applying for GF grants, scaling up existing services, "Know Your Epidemic" and Universal Access to HIV Prevention and Treatment - all require knowledge of the total population of most-at-risk populations including sex workers.

APMG is carrying out population size estimation (PSE) for female sex workers and injecting drug users in Tajikistan at present. The report on these processes is due to be released before the end of 2009, but already there have been several lessons learned:

• Stand-alone PSE processes are likely to be much more effective than combining PSE with risk behavior and behavioral surveillance surveys: while it seems natural to combine these processes, the combination leads to great difficulties in PSE and less-than-ideal surveillance data
  
• PSE in countries with less developed health systems and health statistics requires a specific group of techniques to triangulate the most likely population size
  
• From the Tajikistan work, it appears that a combination of at least one (and, if possible, at least two) multiplier processes together with capture-recapture and local Delphi methods can provide the necessary data for the estimation exercise*
  
• The above processes are cheaper and tend to require less technical capacity among national institutions than more sophisticated studies using social network analysis
  
• Respondent-driven sampling is very useful for BSS and other survey work, but appear to be both expensive and less likely to lead to appropriate estimates than the above combination of methods
  
• Any estimation process among sex workers or drug users needs to be carried out collaboratively with local, state/ province and national authorities in both the health and law enforcement sectors. The final estimates should be agreed at a national consensus workshop.
  

APMG is now discussing the use of these PSE processes with countries in the Asia-Pacific region.
 
* Multiplier processes use known statistics such as detentions or arrests of sex workers together with survey data to estimate the population size.
Capture-recapture is a method that uses an object - in Tajikistan, a pocket calendar - distributed to a population at a defined time prior to survey implementation, combined with survey data.
Local Delphi methods assemble local authorities who have some experience of dealing with sex workers and ask them to agree on the number or the range of numbers of sex workers in the local area. These are combined at state and national level.