Mosquito pool surveillance is a cornerstone of arbovirus risk assessment, yet traditional metrics such as the minimum infection rate (MIR) can understate infection probability and, as conventionally reported, do not provide uncertainty quantification that properly accounts for pooling. We analyzed 7 yr (2018-2024) of mosquito pool surveillance data using pooled maximum likelihood estimation (MLE) with transformed Clopper-Pearson confidence intervals to estimate per-mosquito infection probabilities. The dataset comprised 4,930 pools (246,500 mosquitoes) collected across 80 surveillance sites and tested for 6 arbovirus assay targets covering eastern equine encephalitis, St. Louis encephalitis, West Nile, chikungunya, dengue, and Zika viruses. Viral circulation was sparse (5 positive pools, 0.10% overall positivity), but the pooled-likelihood framework still produced interpretable infection-probability estimates together with conservative upper confidence bounds in zero-positive and sparse-positive weeks. Peak infection probabilities ranged from 0.678 per 1,000 mosquitoes for West Nile virus in August 2022 to 2.667 per 1,000 for chikungunya and Zika virus detections in October 2018. In this low-positivity setting, MIR, pooled MLE, and the Firth/Burrows bias-reduced estimator were numerically close in the observed positive weeks, whereas the main practical gain came from interval estimates that made sampling uncertainty explicit. These results support moving from MIR-only summaries to a broader pooling-aware reporting framework in mosquito surveillance. Reporting both per-mosquito infection estimates and interval bounds places sparse detections and zero-positive weeks on a common operational scale and supports interpretation alongside other surveillance indicators.
Keywords: Arbovirus surveillance; infection probability; maximum likelihood estimation; mosquito control; pooled testing.