P1-S6.07 Ecological analysis of the factors influencing changes in HIV prevalence over time among FSW following a targeted intervention

Background Avahan is a large scale intervention that targets high-risk groups, including female sex workers (FSW), in many epidemiologically heterogeneous districts in southern India. Changes in HIV prevalence post intervention may depend on setting and intervention characteristics. We conducted an ecological analysis to identify which factors were associated with greater changes in FSW HIV prevalence after Avahan start in 2004. Methods All variables were derived from two serial rounds (R1, R2) of cross-sectional FSW surveys, conducted ∼3–4 years apart, from 24 districts of 4 Southern Indian states. We examined the association between the difference in FSW HIV prevalence between rounds (R2−R1)(D.FSW HIV) and different classes of factors (Abstract P1-S6.07 table 1). Intervention factors included differences between rounds in consistent condom use (CCU) with occasional clients, difference in STI prevalence, or fraction of FSW in contact with the intervention at R1, and others (see Abstract P1-S6.07 table 1). Baseline contextual factors included FSW HIV or STI prevalence, fraction of FSW ever asked for anal intercourse (AI), weekly client number per FSW etc at R1, estimates of CCU in 1998, and increase in CCU before R1. Design factors (date of R1, time between R2−R1, differences in response rate between R2 and R1), and differences in contextual factors between rounds as listed Abstract P1-S6.07 table 1 were also explored. Pearson correlations, univariate and multiple linear regression analysis were performed. Abstract P1-S6.07 Table 1 Results of univariate analysis between the difference in FSW HIV prevalence (R2-R1) and different independent variables for each class of factors Types of independent variables N Coefficient of correlation (r) (if pv <0.1) p value Intervention factors Difference* in consistent condom use (CCU) by FSW with occasional clients 27 – ns Difference in Syphilis (Tp) prevalence 23 0.36 0.06 Difference in HSV-2 prevalence (D.FSW HSV2) 27 0.45 0.03 Difference in gonorrhoea or Chlamydia prevalence 27 – ns Difference in the fraction tested for HIV 27 – ns %FSW contacted by NGO at R1 27 – ns %FSW who visited NGO clinic at R1 27 – ns %FSW who received condom from NGO at R1 27 – ns Baseline contextual factors (mainly at R1) R1 FSW HIV prevalence 27 −0.53 <0.01 R1 Syphilis (Tp) prevalence 27 −0.41 0.03 R1 HSV-2 prevalence 27 −0.55 <0.01 R1 Gonorrhoea or Chlamydia prevalence 27 – ns R1 % of FSW ever been asked for anal intercourse (R1 AI) 27 −0.34 0.08 R1 weekly client number per FSW 27 – ns R1 % FSW who are brothel based 26 – ns R1 % FSW tested for HIV 27 – ns Estimated CCU by FSW with occasional clients in 1998† 20 – ns Estimated increase in CCU with occasional clients before R1† 21 – ns Design factors (related to conduct of surveys) Date of R1 27 – ns Time between R1 and R2 27 – ns Difference in response rate between survey rounds (R2−R1) 27 −0.39 0.06 Contextual changes (difference between survey round (R2−R1)) Differences in weekly client number per FSW 27 – ns Difference in the % FSW ever asked for AI 27 0.35 0.08 Different in the fraction of FSW who are literate 27 – ns Difference in the fraction of married FSW 27 – ns Difference in % FSW brothel based 26 – ns Difference in mean duration of sex work for FSW 27 – ns * Difference between rounds (R2−R1). † Adapted from Lowndes et al STI (2009). Results In univariate analyses, D.FSW HIV prevalence was negatively associated with R1 FSW HIV prevalence (r=−0.53), R1 HSV-2 and Tp prevalence, difference in response rate, % asked for AI at R1 (Abstract P1-S6.07 table 1). D.FSW HIV prevalence was positively associated with differences in syphilis (r=0.36) or in HSV-2 prevalence or in % asked for AI. In multivariate analysis, R1.FSW HIV prevalence (slope=−0.57) and estimated CCU in 1998 (slope=0.29)(R=0.73), or R1.FSW HIV (slope=0.19) and D. FSW HSV-2 (slope=−0.83) prevalence (R=0.66) were significantly associated with D.FSW HIV prevalence (p<0.01). Conclusion Contemporary time trends in HIV prevalence are influenced by epidemic stages and historical condom use for many years. HIV prevalence is more (less) likely to decline after effective interventions introduced in mature (early) epidemics. R2 was conducted too early after R1 to expect large decline in HIV. Without control group, mathematical modelling is required to simulate counterfactuals and estimate intervention impact.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.