New HIV diagnoses continue in the Southern US despite widespread prevention efforts, underscoring the need for innovative deployment of prevention tools. Detection and response to genetically clustered infections is a pillar to the Ending the Epidemic initiative. We combined viral load (VL) and surveillance data to prioritize genetic clusters where re-engagement to care activities could be intensified.
We developed automated cluster analyses to prospectively monitor clusters in North Carolina; the system is routinely updated with pol sequences (from clinical and public testing sites), demographic, and clinical data. Clusters were constructed from pairwise genetic distances (TN-93), connecting edges <1.5% difference. Prioritization metrics were assessed for clusters with recent diagnoses (2017-2019) and based on the adjacent nodes to recent diagnoses (edges <1.5%), including members potentially disengaged from care (“Prompt” cases). Prompt cases were defined as members without VLs or persistent/rising viremia (VL>200 c/mL) in the prior 12 months. Connectivity of Prompt cases in clusters was estimated by number of edges to all adjacent nodes (i.e. node degree) per prompt case.
Of 15,558 persons with 25,509 sequences in the pipeline, 2195 had recent diagnoses; 59% (1294) of these were identified in 532 clusters. Clusters involved 2512 members: 1218 (48%) were past diagnoses (≤2016). Recent diagnoses in clusters were more likely to be MSM (65% vs. 46%), younger (33% vs. 15% 18-24 years), and have acute infection (9% vs. 5%) compared to non-clustered recent diagnoses (all p<0.01). Recent diagnoses tended to cluster with other recent diagnoses: 60% (775) clustered with ≥3 recent diagnoses (range 3-28). However, most clusters (65%) involved ≥1 Prompt case and the Prompt connectivity was associated with more recent diagnoses in clusters (Figure). A prioritization threshold of ≥5 recent diagnoses and connectivity ≥5 per cluster, yielded 39 priority clusters (698 members) with 187 Prompt cases (4.8 vs. 1.6 Prompt cases/cluster in non-priority clusters).
We detected a high rate of clustering among recent diagnoses with frequent involvement of past diagnoses. Harnessing longitudinal VL and sequence data allows for timely detection and monitoring of such clusters. Clusters with rapid growth and high network connectivity with past diagnoses without viral suppression can be prioritized for intensified care re-engagement and retention support.