A quantitative framework for prioritizing intervention benefits public health officials needing to address many HIV transmission clusters under time and resource constraints. Using empirical and simulated data, we evaluate the hypothesis that HIV transmission cluster prioritization measures based on phylogenetically-derived lineage-level diversification rates will exceed commonly-used non-phylogenetic prioritization measures in highlighting clusters in urgent need of intervention, without subjectivity or need for supporting data.
35,752 HIV partial pol sequences collected from 9,822 participants between 1996-2019 were used to infer approximate maximum-likelihood phylogenetic trees rooted by root-to-tip regression. Transmission clusters were inferred using a patristic distance threshold of 0.02 substitutions/site, such that lineage-level diversification rates calculated for all tips could be summarized by cluster. The same methods were applied to simulated epidemic sequence data generated to mimic the BC epidemic using FAVITES.
Empirical data suggest diversification-rate based measures perform comparably to non-phylogenetic measures in recreating transmission cluster prioritization choices made by public heath teams (Figure 1). However, analysis of unbiased simulated data reveals a stronger relationship between diversification rate-based measures and future cluster growth, particularly long-term growth (median Spearman r = 0.39 vs 0.17). Diversification rate-based measures also highlighted groups with significantly more future transmissions than random groups of equal size (Mann-Whitney p<0.001). Furthermore, the relationship between diversification rate measures and future growth was notably more robust to decreased sampling proportion.
Phylogenetically-derived lineage-level diversification rate-based measures not only frequently outperform non-phylogenetic measures and show less sensitivity to lower sampling, but also offer several additional advantages beneficial to optimization of the public health prioritization process.