Estimating the size of people who inject drugs: Utility, means and challenges

The HIV epidemic was first identified in several countries among key populations, including female sex workers, men who have sex with other men, transgenders and people who inject drugs (PWID). These key populations are mostly ‘hard-to-reach’ or ‘hidden’ populations as their behaviours are considered unaceptable in some settings; hence, the size of these populations is generally unknown1. This editorial article discusses the importance of knowing the size of PWID, the different methods used to estimate their size and the challenges faced in the field while conducting the size estimation activities.

There are several reasons behind estimating the size of key population e.g. PWID but we broadly can categorize them into two major areas: (i) policy related reasons; and (ii) programme related reasons. The policy related reasons include advocacy, response planning and resource allocation, plus estimations of numbers infected with HIV and projections of the burden of disease. The area of programme related reasons encompass intervention planning, measurement of coverage and monitoring and evaluation of interventions2. The estimated size of key population groups including PWID have also been used to garner political support and commitment for the national response for HIV; direct funding by characterizing the extent and pattern of the epidemic; plan HIV programmes for key populations at the national, provincial/state as well as district/city levels; and monitor and evaluate programmes in terms of coverage, quality and effectiveness. The results of a size estimation exercise study can mobilize necessary media attention to demystify so-called age-old traditional notions that the actual number of some key population groups, e.g. PWID, is considerably insignificant and hence not worthy of public health action. Most importantly, size estimates of PWID can be used as denominators for reporting on several international monitoring indicators and for grant applications12.

Today, we have several methods in our knowledge to estimate the size of PWID. However, it is a real fact that each of these available methods has its own strengths. Therefore, it is strongly recommended in various methodological guidelines that we must use multiple size estimation methods, wherever possible, to arrive at a consensus estimate345. All the available size estimation methods can broadly be categorized into two broad categories: (a) direct methods (census and enumeration) and (b) indirect methods (capture–recapture, multiplier, reverse tracking, network scale-up and successive sampling). Direct methods count members of the population directly, whereas indirect methods use data from different sources to estimate their size. In addition, these size estimation methods can also be classified according to their data source: (a) data collected from key populations and (b) data collected from the general population678. For a detailed description of each of these methods, please refer to the recent WHO guidelines document on size estimation and other relevant references.

In India, so far only three different methods have been used to estimate the size of PWID, including capture–recapture, multiplier and programmatic mapping and population size estimation (p-MPSE), which is a variant of the census method. The use of capture–recapture as well as multiplier method to estimate the size of PWID was mostly limited to small geographical areas such as a few towns or cities or districts only, whereas p-MPSE was used widely at state or regional or national level on several occasions over the past two decades.

As early as in the 1990s, some scientists from the Indian Council of Medical Research (ICMR) made efforts to estimate the size of PWID in a few cities of the northeastern part of India using capture–recapture method. Later, in the 2000s (in 2005–2006 and 2009), the Bill and Melinda Gates Foundation (BMGF) supported five ICMR institutes and FHI 360 to undertake size estimation of PWID across some selected districts of the country (intervention districts of the Avahan programme) using multiple methods of size estimation, including multiplier method9.

National-level PWID size estimates were published by the National AIDS Control Organization (NACO) in 2006 and 200910; however, there has not been another major update by them since then. One of the early attempts to develop national size estimates for key populations, including PWID, took place during the planning stages of the National AIDS Control Programme (NACP)-III in 200611. The estimates were developed by an expert group and were based on a review of available mapping data from targeted interventions (TIs) implemented between 2002 and 2005 under the NACP-II and supplemented with data from other projects being implemented in India at that time. Corrections were applied to account for missing districts and the lack of data from rural areas. The national estimates for PWID developed at that time were based on relatively limited data collected primarily in urban areas and included the most visible subset of the PWID group11. These involved several assumptions and compromises, such as the decision not to account for the less visible PWID, due to lack of information. Difficulties in covering all geographic locations, lack of depth in coverage, especially in places where there were no TIs, lack of clarity and non-standardized definition were all cited as limitations of these estimates, which were ‘crude estimates’.

In 2009, a major update to the size estimates was done. It was based on two data sources. The first was TI mapping data from 17 States, which were collected according to detailed NACP-III guidelines. The second data source came from a ‘reverification’ process, which was carried out in selected locations as part of the PSE update. It is important to note that under NACP-III guidance, the TIs were to address only the subset of key populations (KP) groups at the highest risk. Hence, for PWID, the focus was to be on those who share needles.

Another round of national PWID size estimates was undertaken by the All India Institute of Medical Sciences, New Delhi in 2019 in collaboration with the Ministry of Social Justice and Empowerment12. A multiplier method was used by them for estimating the size of PWID. Importantly, their estimate of PWID turned out to be nearly six times higher than the estimate available with the NACP. An expert group was formed to critically review the two separate estimates and resolve the differences. However, no consensus or conclusion was arrived at, and it was decided to wait for a future round of national size estimates planned by the national programme. Finally, in 2021–2022, the NACO implemented the latest round of size estimation exercises of key populations, including PWID, using the p-MPSE method. The estimates from this latest round of national-level size estimation exercise are still awaited (or not published yet).

Challenges faced while implementing the above size estimation exercises in the field were many in number. For example, carrying out the capture–recapture method was relatively straightforward, but fulfilling the necessary conditions was challenging. Similarly, implementation of the multiplier method (programme service multiplier or unique object multiplier) sounded easy and simple, but satisfying the underlying assumptions was tough. Most of the time, the two data sources were not fully independent. There were issues in defining the population (PWID) by two data sources in the same way. Furthermore, periods, age range and geographical areas of data sources were not fully aligned. More so, the most widely used size estimation method in India, i.e. p-MPSE, has several limitations. One of the major limitations of this method is that it is not a census but a variant of census that attempts to count all members of the population through key informant interviews. Selection, composition and knowledge of the key informants determine the quality of the estimates. Most importantly, this method misses all those who do not visit any physical location (as it covers only the physical venues). Since a major proportion of PWID may not visit any physical venue for buying, selling or injecting purposes, p-MPSE is likely to underestimate the size of PWID significantly. In addition, stigma and discrimination may prevent them in identifying themselves as PWID.

The most efficient and pragmatic PWID size estimation strategy for a big country like India would be, first, to integrate it with any national survey among PWID, e.g. national behavioural surveillance survey (BSS) or biobehavioural surveillance (BBS). Undertaking it as an independent exercise (like in the past except the ones supported by the BMGF) is seriously questionable and is a waste of limited resources for the programme. Second, a mix of multiple methods must be used instead of one single method. Third, the use of some new emerging methods like successive sampling should be explored. Fourth, a service multiplier method using RDS as a sampling method for the survey component could be the most appropriate multiplier method, particularly for the PWID group. Finally, regular (annual/biannual) size estimation exercises must be part of the programme implementation.