BACKGROUND: To detect and minimize the risk of local Zika virus (ZIKV) transmission throughout the 2016 mosquito season, the New York City (NYC) Department of Health and Mental Hygiene needed to identify areas with potentially viremic travel-associated ZIKV human cases. Imported ZIKV cases identified through passive surveillance might be spatially non-representative, given incomplete case ascertainment of individuals who are asymptomatic or symptomatic but did not seek care or testing. We used statistical modeling to identify areas at greatest risk for ZIKV importation.
METHODS: For each of 14 weeks during June–September 2016, we constructed zero-inflated Poisson regression models to output the predicted number of weighted ZIKV cases per census tract; case counts were weighted by the time since illness onset (i.e., cases with onset in the 28 days before a data pull were counted fully, cases in the past 29–56 days were counted as 1/2 cases, etc.). An automated forward selection process was applied to eleven census tract-level covariates, from static sociodemographic census data and the latest surveillance data. To account for testing bias, we plugged the maximum cumulative observed ZIKV testing rate in any census tract into the final regression equation. Covariates selected for inclusion in the model and with P<0.05 for at least 9 of 14 weeks were summarized. Spearman rank correlation coefficients were used to characterize fluctuation in modeled risk areas over time. Census tracts were grouped by natural breaks into risk categories.
RESULTS: Of 2,140 populated census tracts in NYC, 1,659 (78%) had zero observed ZIKV cases during June–September 2016. Covariates consistently selected for modeling included: the proportion of the population living below the federal poverty level, borough of residence, number of persons with ZIKV testing in the prior 30 days, and the number of dengue and chikungunya cases since 2013 with travel to countries with active ZIKV transmission. Although modeled risk areas fluctuated (e.g., last weeks of July and August 2016: Spearman correlation=0.63, P<0.0001), areas in the Bronx and upper Manhattan were consistently identified as high-risk.
CONCLUSIONS: We used observable characteristics of areas with recent, known travel-associated ZIKV cases to identify similar areas with no observed cases that might also be at-risk in any given week. The lack of data on the spatial distribution within NYC of persons arriving from ZIKV-affected countries is a limitation. Findings were used to target activities including public outreach, sentinel surveillance, and trapping and controlling Aedes spp. mosquitoes.