Epidemiological yield and cost-effectiveness of non-targeted community HIV screening in a low-prevalence district: a prospective quasi-experimental study.

OBJECTIVE

To evaluate the epidemiological yield, cost-effectiveness, and mechanisms of diagnostic cascade attrition of active, expanded community-based HIV screening in a low-prevalence region, and to contextualize the regional management burden.

METHODS

A prospective, non-randomized, quasi-experimental study was conducted (August-December 2025) across six demographically matched sub-districts. Three implemented active community screening (intervention); three maintained routine facility-based testing (control). A deduplicated dual-track database mapped the diagnostic cascade, supplemented by a 7-month longitudinal cohort to assess regional disease burden. Micro-costing determined fixed and variable costs from a health-system perspective. The Number Needed to Screen (NNS), Cost Per New Diagnosis (CPND), and Incremental Cost-Effectiveness Ratio (ICER) were calculated, supported by Propensity Score Matching (PSM), permutation tests, and deterministic sensitivity analyses.

RESULTS

Among 17,305 unique participants (intervention: 14,434; control: 2,871), the intervention expanded coverage in the 35-49 age cohort (35.0% vs. 25.5%, p < 0.001). However, only one true new HIV diagnosis occurred in the intervention arm (yield: 0.0069%), showing no statistical difference from the control (0%, Fisher's exact test p = 1.000), a finding robustly confirmed by PSM and permutation tests. The overall NNS was 14,434, remaining undefined (>1,250 and >5,055) for the highly prioritized 18-49 age cohorts. Longitudinal cohort data revealed a stark demographic mismatch: 76.0% of true local incident cases were aged ≧50 years, and the regional management burden was increasingly driven by migrant populations rather than local transmission. Notably, 80.0% (8/10) of reactive initial screens in the intervention group were previously diagnosed individuals, predominantly older males (87.5% male; 75.0% aged 50-64). While temporal correlation was not statistically significant (Spearman's ρ = 0.564, p = 0.322), the absolute metrics highlight substantial known-positive interference. Consequently, the intervention yielded an ICER of 504,886 CNY per additional diagnosis, remaining above the willingness-to-pay threshold even under extreme sensitivity scenarios.

CONCLUSION

In epidemiologically saturated settings, non-targeted community screening yields profound diminishing marginal returns. Material incentives inadvertently capture previously diagnosed older males who undergo repeated testing, resulting in substantial resource redundancy rather than identifying occult infections. To optimize cost-effectiveness, public health strategies must implement digital pre-screening interception mechanisms and strategically reallocate funds toward key-population friendly services and peer-driven interventions.