Network Scale-Up Correction Factors for Population Size Estimation of People Who Inject Drugs and Female Sex Workers in Iran

Introduction

The results of the network scale-up (NSU) method in estimating the size of key populations for HIV might be biased if the recruited subjects are not fully informed of the risky behaviors of people in their networks (low visibility), or key populations have a smaller social network (low popularity). We aimed to measure such biases in the size estimation of people who inject drugs (PWIDs), and female sex workers (FSWs) in Iran.

Methods

We interviewed 163 male PWIDs, 76 FSWs (known as egos) and 600 subjects from the general population. We selected twenty first-names (ten males and ten females) and asked the study subjects separately how many people they knew with one of these names (known as alters). Visibility Factor (VF) was defined as the percentage of FSW or PWID alters that were aware of their behavior. In addition, the popularity factor (PF) was calculated by dividing the number of alters reported by FSWs and PWIDs into that of the general population. The 95% uncertainty intervals (UI) were calculated using bootstrap technique.

Results

The VF was estimated at 54% (95% UI: 52%–56%) for PWID and 45% (95% UI: 42%– 48%) for FSW. The VF among the peer alters was significantly higher than non-peer ones. The PF for PWID and FSW was 69% (95% UI: 66%–73%) and 77% (95% UI: 72%–83%), respectively. The cross-validation and name splitting analysis showed that our estimates were not influenced by any single name.

Conclusions

Both correction factors, particularly VF were far from one, and NSU results without correction, could lead to up to 4 times underestimation of the sizes. Therefore, applying these coefficients is necessary in NSU projects.