Potential High-Risk Areas for Zika Virus Transmission in the Contiguous United States

Abstract

Objectives. To understand where transmission of Zika virus has the highest likelihood to occur in the contiguous United States with regard to its transmission both sexually and via Aedes aegypti mosquito bites.

Methods. We evaluated the 2 routes of transmission risk with predictors of sexually transmitted infections (percentage women of childbearing age, birthrate, gonorrhea and chlamydia rates, concentrated disadvantage) as a surrogate for unprotected sexual activity and the demographic distribution of the A. aegypti mosquito across 3108 counties in the contiguous United States.

Results. We found that 507 counties had the highest risk of virus exposure via mosquito vector or unprotected sexual activity; these were concentrated in southern states extending northward along the Atlantic coast and southern California, with the highest predicted risk in Mississippi counties.

Conclusions. Identifying areas with higher transmission risk can inform prevention strategies and vector control, and assist in planning for diagnosis and treatment.

The Zika virus has recently burst onto the scene in the Americas, particularly Brazil. There has been a dramatic rise in cases of both microcephaly and Guillain-Barré syndrome associated with the introduction of the Zika virus.^1,2 The evidence directly associating the Zika virus with these conditions has recently implicated Zika as a primary cause for both conditions.² This virus, once thought of as a mere nuisance, has taken on increased importance with the previous declaration of a Public Health Emergency of International Concern by the World Health Organization.³ On the basis of these latest actions, public health measures have focused on decreasing the risk of infection to pregnant women, as well as women of childbearing age.^4,5

The virus has at least 2 known methods for transmission—via a mosquito vector and by sexual transmission.^6–8 The mosquito vector, specifically the Aedes aegypti species, is the same mosquito that carries the dengue and chikungunya viruses, and is thought to be the primary method for infection.⁹ However, recent reports suggest that the Zika virus can survive within semen for significant amounts of time and, thus, the sexual transmission route of infection may be significantly underestimated.^5,6 Because of this, infection with the Zika virus may have significant long-term effects on sexual and reproductive health in addition to the neurological sequelae in newborns. The sexual transmission route is also concerning because sexually transmitted infections (STIs) tend to cluster geographically and occur disproportionately in areas with higher concentrated disadvantage^10,11 and limited access to sexual health resources (e.g., condoms, contraception, abortion).^12,13 It is currently unclear how efficient the sexual transmission of the Zika virus is, which has limited some of the prevention efforts.⁵ Namely, the relative transmission risk between mosquito-borne and sexual transmission is unclear.

The Zika virus has continued to spread northward from Brazil with cases now documented in Central America, Mexico, and the Caribbean. Notably, Puerto Rico has reported 93% of locally acquired Zika cases among the United States and US territories.⁵ There have now been locally acquired mosquito-borne and sexually transmitted Zika cases in the contiguous United States, specifically in Florida, Texas, and California.⁵ We sought to identify potential high-risk areas in the United States for the spread of the Zika virus by identifying significant routes of transmission, specifically via sexual transmission, in the context of exposure to the A. aegypti mosquito. Understanding where sexually transmitted Zika virus is likely to occur in parallel across the United States can inform public health efforts to prevent and control the spread of Zika virus.

METHODS

We used an ecological study design to identify potential areas within the United States considered higher risk for sustained Zika infection and transmission as well as to quantify the population at risk for exposure, with particular attention to women who are pregnant or of childbearing age because of the risk of microcephaly. We defined high-risk areas as localities with a higher probability of exposure to the 2 major methods of transmission, including the A. aegypti mosquito populations and unprotected sexual activity (as measured by STIs).

Population at Risk

We estimated the percentage of women who gave birth and percentage of women of childbearing age by using geographic boundary and demographic characteristics for US counties from the US Census Bureau’s 2014 Topologically Integrated Geographic Encoding and Referencing (TIGER) shape files and 2010 to 2014 (5-year) American Community Survey (ACS).¹⁴ We estimated the percentage of women of childbearing age by dividing the number of women aged 15 to 44 years by the total female population within each county, as provided by the 2010 to 2014 ACS 5-year estimates. In addition, we estimated percentage of women who gave birth in 2010 by dividing the number of women aged 15 to 50 years who had at least 1 birth in the past 12 months by the total female population within each county by using the 2006–2010 ACS. We used the 2006 to 2010 ACS to estimate percentage of women who gave birth because the more recent ACS data had large numbers of counties with missing or suppressed values because of small population counts. We downloaded these files that the Minnesota Population Center’s National Health Geographic Information System (NHGIS) had compiled.¹⁵ We ranked both variables and categorized them into quartiles for analysis to account for the large variance as continuous measures.

To estimate population exposure to the vector primarily responsible for the spread of the Zika virus, we downloaded data estimating the distribution of the A. aegypti mosquito for each of the 3108 US counties from the Dryad Digital Repository.¹⁶ These estimates for mosquito presence come from complex ecological niche models that have been widely used and accepted. These models utilized observed mosquito populations and then interpolated the geographic distribution of the mosquito species throughout the world. These data estimate the distribution of A. aegypti mosquitoes with methods that have been used previously and that are described in detail elsewhere.¹⁷ We used the estimated probability of A. aegypti mosquito presence within each US county, which is based on a boosted regression tree modeling approach that incorporated temperature, land cover, and environmental factors to predict the probability of the distribution of A. aegypti globally.¹⁷

We estimated the risk of unprotected sexual activity by using a surrogate marker of rate of STIs in the community. The STI rate is a well-established estimate for the rate of unprotected sexual activity within the population.^18–20 We obtained incident cases of gonorrhea and chlamydia infection that occurred during 2014 for each county in the United States from the National Center for HIV/AIDS, Viral Hepatitis, STD, and TB Prevention at the US Centers for Disease Control and Prevention (CDC).²¹ We calculated crude incidence rates of gonorrhea and chlamydia for each county as the ratio of the number of cases to the total population at risk. We calculated standardized incidence ratios (SIRs), representing the observed gonorrhea and chlamydia rate relative to the expected rate, for each county.²²

Transmissions of STIs tend to cluster in areas with higher concentrated disadvantage. To construct a measure of concentrated disadvantage, we collected and downloaded geographic boundary and demographic characteristic data for all US counties from the US Census Bureau’s 2014 TIGER shape files and 2010 to 2014 (5-year) ACS,¹⁴ which were compiled by the Minnesota Population Center’s NHGIS. This analysis included indicators of concentrated disadvantage including the proportions of the population that are African American, proportion of female-headed households, proportion of households receiving food stamps, proportion of individuals using public insurance programs (all ages and races/ethnicities), proportion of households with children younger than 18 years, proportion of households without employment during the past 12 months, proportion of households below the federal poverty line, proportion of the population with less than a high-school education (or high-school degree equivalent [i.e., general education diploma]), and the median household income (mean-centered).^23,24

We then used principal components factor analysis to create an index of concentrated disadvantage²⁵ by using the PSYCH package in R version 3.1.1 (R Foundation for Statistical Computing, Vienna, Austria). This yielded 2 factors with eigenvalues greater than 1 that cumulatively explained 71% of the total variance of the aforementioned 9 variables (Table A, available as a supplement to the online version of this article at http://www.ajph.org). We standardized these 2 factors (via z-score transformation) and extracted them as continuous variables for use in the regression models.

Statistical Analyses

We conducted the statistical analysis in a series of steps. First, we predicted county-level STI rates by using a spatial regression model with adjustment for concentrated disadvantage, percentage of women who gave birth, and the proportion of the county’s population that is of childbearing age. The purpose of this analysis was to identify counties with increased risk of transmission by sexual intercourse and significant at-risk populations (percentage of women of childbearing age in light of the increased risk of microcephaly). Second, we combined the STI model results with the probability of A. aegypti presence to estimate the total risk of Zika transmission (i.e., both sexual and vector transmission). We used the results of this analysis to identify counties with the highest potential risk based on both transmission means. Finally, we mapped these potential high-risk areas and tabulated the county population at risk.

We tested spatial autocorrelation in the residuals of the crude gonorrhea and chlamydia rates by using choropleth maps and Moran’s I.²⁶ Positive (negative) values of I indicate positive (negative) spatial correlation, meaning that nearby counties tended to exhibit similar (dissimilar) gonorrhea and chlamydia rates. We defined the spatial adjacency of the data by using rook contiguity and queen contiguity, and by using the 5 nearest neighbors to examine their effect on the results. Because results did not vary substantially across spatial adjacency definitions, we used the queen contiguity structure for all analyses.

Spatially dependent data violate the independence assumption required for generalized linear models, and can lead to an underestimation of standard errors, overly narrow confidence interval estimates, and, consequently, incorrect statistical inference if ignored.²⁷ To account for residual dependence, we augmented the linear predictor with a spatial random effect, as part of a Bayesian hierarchical model.²⁸ This random effect took the form of a conditional autoregression, which introduces spatial dependence through the adjacency structure of areal units.²⁸ The conditional autoregression model used Markov-chain Monte Carlo simulation to estimate model parameters.²⁹

To account for spatial dependence (as indicated by Moran’s I), we implemented a spatial Poisson regression model by using counties as the areal unit of analysis to predict gonorrhea and chlamydia rates. In particular, we fit the Besag–York–Mollié model with a conditional autoregression prior.²⁸ We used the Besag–York–Mollié framework to perform a spatial Poisson regression with offset, where the offset for the ith county was the county population. We assigned the intercept and regression coefficients a conservative normal prior with a mean of 0 and a standard deviation of 1 000 000.²⁹ We then drew 120 000 samples from the posterior distribution to estimate model parameters. We used a burn-in period of 20 000 samples and a thinning rate of 10, resulting in a total of 10 000 samples for analysis. We used the CARBayes package in version 3.1.1 of R for all analyses. We modeled the relationship between gonorrhea and chlamydia rates and percentage of women who gave birth, percentage of women of childbearing age, and the concentrated disadvantage factors. We then exponentiated the model coefficients (i.e., exp[Β]) to present relative risks with their corresponding 95% confidence intervals.

We also report choropleth maps of the crude distributions of all covariates, and the standardized incidence ratio (SIR) of gonorrhea and chlamydia adjusted for all predictors.³⁰ We then overlayed choropleth maps, with the adjusted SIRs from the spatial Poisson model with the distribution of the A. aegypti mosquito. We identified counties where the SIR was greater than 1.0 and the probability of A. aegypti presence greater than 0.3 as high risk. We chose counties with greater than 0.3 as high risk for A. aegypti presence because this is the mean probability among counties that A. aegypti is expected to inhabit.

To determine model sensitivity, we reduced the number of STI cases for each county by a factor of 15% and then again by 30%. We then fit separate spatial Poisson regression models for these scenarios. We will refer to the models as the 100%, 85%, and 70% transmission models, corresponding to the rate of STI transmission that is assumed for Zika sexual transmission. The justification for this modeling approach is that for Zika virus transmission is unknown and could be lower than that for gonorrhea and chlamydia. These more conservative models provide a gradient of the likelihood of sexual transmission given that current incidence of the sexual transmission of Zika is less than that of common STIs.

RESULTS

Rates of STI and concentrated disadvantage measures varied significantly across the 3108 counties (Table 1) in the United States (Figure A, available as a supplement to the online version of this article at http://www.ajph.org). The Moran’s I test for spatial autocorrelation in the residuals from crude Poisson models revealed significant spatial autocorrelation for all variables, indicating that the use of spatially dependent models was appropriate (results not shown). There was a statistically significant relationship between concentrated disadvantage, pregnancy measures, and STI rates (Table 2). Overall, counties with higher proportions of women of childbearing age, more disadvantage, and more births (not statistically significant) were associated with higher STI incidence rates. More specifically, the adjusted spatial Poisson model indicated that for a 1% increase in the percentage of women of childbearing age, the risk of STIs increased by 1.14 for counties in quartile 2, 1.31 times for counties in quartile 3, and 1.61 times for counties in quartile 4, whereas the relative risk for STIs for a 1% increase in concentrated disadvantage was 1.25 to 1.37 times greater after adjustment for the percentage of women who gave birth and spatial autocorrelation. Table 3 also shows the adjusted relative risks for the models that assume Zika virus transmission to be 85% and 70% of the STI model just described. Figure A displays the adjusted STI rate SIRs across US counties.

TABLE 1—

Zika-Related Characteristics of 3108 Counties: Contiguous 48 United States, 2006–2014

Characteristic	All Counties (n = 3108), Mean (SD)	High-Risk Counties (n = 507), Mean (SD)	Low-Risk Counties (n = 2601), Mean (SD)
Gonorrhea and chlamydia rate per 100 000 people in 2014	409.9 (318.0)	870.7 (286.0)	320.1 (235.5)
Women aged 15–44 y in 2010, %
Quartile 1 (11.0–32.1)	29.1 (2.7)	30.5 (1.4)	29.0 (2.7)
Quartile 2 (32.2–34.9)	33.6 (0.8)	33.8 (0.7)	33.6 (0.8)
Quartile 3 (35.0–37.8)	36.2 (0.8)	36.3 (0.8)	36.2 (0.8)
Quartile 4 (37.9–65.8)	41.6 (4.0)	42.3 (4.5)	41.3 (3.8)
Women aged 15–50 y who gave birth in 2006–2010, %
Quartile 1 (0.0–4.5)	3.5 (1.1)	3.5 (0.9)	3.5 (1.1)
Quartile 2 (4.6–5.5)	5.1 (0.3)	5.1 (0.3)	5.1 (0.3)
Quartile 3 (5.6–6.7)	6.1 (0.3)	6.1 (0.3)	6.1 (0.3)
Quartile 4 (6.8–21.2)	8.5 (2.1)	8.5 (2.5)	8.5 (2.0)
Measures included in the concentrated disadvantage index, 2010–2014
African Americans, %	9.1 (14.6)	32.3 (19.0)	4.6 (7.6)
Female-headed households, %	11.4 (4.4)	17.2 (4.2)	10.3 (3.4)
Households receiving food stamps, %	14.8 (6.6)	19.9 (6.6)	13.8 (6.2)
Public health insurance users, %	35.3 (7.6)	37.5 (7.5)	34.8 (7.6)
Households with children aged < 18 y, %	44.4 (6.4)	48.0 (4.8)	43.7 (6.5)
Households without employment past 12 mo, %	4.9 (2.0)	6.3 (1.9)	4.7 (1.9)
Households below federal poverty level, %	16.0 (6.0)	21.5 (6.1)	15.0 (5.4)
< high-school or GED education, %	28.2 (5.3)	31.5 (4.2)	27.6 (5.2)
Median household income, 2014 US$ (inflation-adjusted)	46 358 (11 944)	40 009 (10 044)	47 595 (11 894)

Note. GED = general education diploma.

Source. National Center for HIV/AIDS, Viral Hepatitis, STD, and TB Prevention (2014)²¹ and American Community Survey (2006–2010).¹⁴

TABLE 2—

Adjusted Relative Risk Estimates From the Spatial Poisson Regression Model Predicting County-Level Gonorrhea and Chlamydia Incidence Rates for Models Assuming 100%, 85%, and 70% Sexual Transmission: Contiguous 48 United States, 2006–2014

	100% Transmission Assumed		85% Transmission Assumed		70% Transmission Assumed
Variable	Crude IR (per 100 000)	aRR (95% CI)	Crude IR (per 100 000)	aRR (95% CI)	Crude IR (per 100 000)	aRR (95% CI)
Women aged 15–44 y, %
Quartile 1 (11.0–32.1)	289.1	1 (Ref)	243.4	1 (Ref)	200.2	1 (Ref)
Quartile 2 (32.2–34.9)	348.6	1.14 (1.03, 1.23)	295.0	1.18 (1.09, 1.29)	242.8	1.17 (1.09, 1.26)
Quartile 3 (35.0–37.8)	438.6	1.31 (1.18, 1.44)	372.2	1.36 (1.25, 1.55)	306.4	1.34 (1.24, 1.48)
Quartile 4 (37.9–65.8)	651.5	1.61 (1.42, 1.74)	553.6	1.67 (1.48, 1.94)	455.9	1.62 (1.51, 1.85)
Women aged 15–50 y who gave birth in 2010, %
Quartile 1 (0.0–4.5)	433.3	1 (Ref)	367.2	1 (Ref)	302.2	1 (Ref)
Quartile 2 (4.6–5.5)	521.7	1.04 (0.98, 1.14)	443.2	1.06 (0.99, 1.12)	364.9	1.04 (0.97, 1.17)
Quartile 3 (5.6–6.7)	627.7	1.04 (0.98, 1.10)	533.2	1.05 (0.97, 1.12)	439.1	1.04 (0.98, 1.13)
Quartile 4 (6.8–21.2)	593.0	1.02 (0.97, 1.10)	503.1	1.03 (0.97, 1.12)	414.2	1.02 (0.96, 1.14)
Concentrated disadvantage indices
Factor 1, Za	. . .	1.25 (1.21, 1.29)	. . .	1.26 (1.21, 1.29)	. . .	1.25 (1.22, 1.30)
Factor 2, Za	. . .	1.37 (1.30, 1.45)	. . .	1.37 (1.27, 1.46)	. . .	1.38 (1.32, 1.44)

Note. aRR = adjusted relative risk; CI = confidence interval; IR = incidence rate.

Source. National Center for HIV/AIDS, Viral Hepatitis, STD, and TB Prevention (2014)²¹ and American Community Survey (2006–2010).¹⁴

^aZ indicates a standardized factor that was created with principal component analysis.

TABLE 3—

Adjusted Relative Risks of Zika Virus Transmission: 100%, 85%, and 70% Risk: Continguous 48 United States, 2006–2014

Standardized Incidence Ratio	Probability of Aedes aegypti Habitation	No. of Counties Identified	No. of Women Who Gave Birtha in 2010	No. of Women of Childbearing Ageb	Total Population at Risk	Adjusted STI Rate per 100 000
100% transmission assumed
> 1.0	≥ 0.30	507	1 396 963	20 533 212	95 605 958	838.2
> 1.0	≥ 0.55	411	1 023 225	14 827 850	68 923 228	880.0
> 1.0	≥ 0.73	148	530 871	7 519 567	35 102 475	813.1
≥ 1.5	≥ 0.30	203	538 262	7 965 083	35 806 915	1 090.5
≥ 1.5	≥ 0.55	181	469 974	6 923 318	31 025 903	1 110.7
≥ 1.5	≥ 0.73	55	169 395	2 422 563	10 911 866	1 067.7
≥ 2.0	≥ 0.30	72	160 158	2 286 391	10 211 411	1 447.0
≥ 2.0	≥ 0.55	72	160 158	2 286 391	10 211 411	1 447.0
≥ 2.0	≥ 0.73	20	42 073	632 689	2 808 219	1 443.9
85% transmission assumed
> 1.0	≥ 0.30	358	1 102 517	16 081 094	73 883 264	770.1
> 1.0	≥ 0.55	299	817 786	11 748 698	53 673 009	813.8
> 1.0	≥ 0.73	100	412 457	5 794 517	26 397 120	751.2
≥ 1.5	≥ 0.30	115	280 404	4 097 889	18 236 981	1 086.0
≥ 1.5	≥ 0.55	112	269 116	3 869 530	17 250 458	1 094.1
≥ 1.5	≥ 0.73	32	81 502	1 048 002	4 728 616	1 099.2
≥ 2.0	≥ 0.30	44	126 058	1 810 780	8 072 763	1 279.0
≥ 2.0	≥ 0.55	44	126 058	1 810 780	8 072 763	1 279.0
≥ 2.0	≥ 0.73	14	31 499	464 684	2 114 192	1 283.2
70% transmission assumed
> 1.0	≥ 0.30	230	631 509	9 265 003	41 733 985	735.8
> 1.0	≥ 0.55	202	559 753	8 176 058	36 701 839	745.3
> 1.0	≥ 0.73	62	233 028	3 304 356	14 839 845	699.9
≥ 1.5	≥ 0.30	63	149 749	2 126 499	9 449 352	1 029.0
≥ 1.5	≥ 0.55	63	149 749	2 126 799	9 449 352	1 029.0
≥ 1.5	≥ 0.73	18	40 872	609 943	2 709 685	1 017.0
≥ 2.0	≥ 0.30	12	38 259	594 069	2 565 004	1 169.3
≥ 2.0	≥ 0.55	12	38 259	594 069	2 565 004	1 169.3
≥ 2.0	≥ 0.73	3	3 660	67 951	288 500	1 323.5

Note. STI = sexually transmitted infection.

Source. National Center for HIV/AIDS, Viral Hepatitis, STD, and TB Prevention (2014)²¹ and American Community Survey (2006–2010).¹⁴

^aWomen aged 15–50 years.

^bWomen aged 15–44 years (5-year estimate from 2010–2014).

The distribution of the A. aegypti mosquito is outlined in Figure B (available as a supplement to the online version of this article at http://www.ajph.org) and indicates that mosquito distribution is most heavily concentrated in the southeastern United States. Figure 1 displays counties (n = 507) with elevated SIRs (> 1.0) and higher than average A. aegypti mosquito presence (probability of A. aegypti > 0.3). These maps highlight a pattern of higher risk in the southern states extending northward along the Atlantic coast and in southern California, and a potential for high risk in the Mississippi delta region. Table 3 shows the number of counties and populations at risk for Zika virus infection according to varying criteria for the expected risk of STI and the predicted probability of A. aegypti mosquito populations. With 100% of the transmission risk assumed, one third of the contiguous United States population is living in an area that is considered potentially high-risk for Zika infection (Table 3).

FIGURE 1—

Counties With the Greatest Risk of Gonorrhea and Chlamydia, and High Aedes aegypti Mosquito Populations by (a) Overall Risk and (b) Adjusted Standardized Incidence Ratios (SIRs): United States, 2014

Note. The map in part a shows all counties with SIRs of gonorrhea and chlamydia that are greater than 1.0 and have greater-than-average probability of Aedes aegypti mosquito populations (predicted probability > 0.3). Darker shades of red correspond to counties that remain high risk in the models in which transmission rates were reduced to 70%. The map in part b shows the variability in SIRs among the high-risk counties shown in the upper map.

Source. National Center for HIV/AIDS, Viral Hepatitis, STD, and TB Prevention²¹ and Dryad Data Repository.¹⁶

DISCUSSION

The purpose of this study was to identify where mosquito-borne and sexually transmitted Zika virus infection are most likely to co-occur within the 48 contiguous United States when the virus has been introduced to the continental United States’ A. aegypti mosquito population. Although this study focused on regions in the United States that have yet to be impacted by the Zika virus, and assumes no countermeasures were deployed, we propose the need for significant planning and prevention in areas and populations most likely to experience the highest burdens from Zika infection. Our findings suggest that between 2.8 million and 95.6 million (0.3–41.7 million in the 70% transmission model) individuals could be at higher risk for Zika infection than individuals living in other counties in the United States. Of those at potential higher risk, 42 000 to 1.4 million (3700–632 000 in the 70% transmission model) may be pregnant women, the highest at-risk group (Table 3). Our results also show that complications from the Zika virus are likely to overlap with impoverished counties with large minority populations where resources are more likely to be scarce to combat a large-scale Zika virus outbreak.

These results align with other estimates of Zika infection risk in the United States based on air travel and mosquito populations.³¹ Air travel was used as a significant estimate for mosquito exposure. This study aimed to incorporate the mosquito exposure and sexual behavioral exposure to highlight prevention needs in specific high-risk areas. In this study we also assumed that the Zika virus will eventually become incorporated into the resident US mosquito population and focused on significant factors currently known to have an impact on disease transmission, particularly the presence of the A. aegypti mosquito, sociodemographic risk, and unprotected sexual intercourse.⁵ It is likely that the reports of sexually transmitted Zika virus are underreported as expressing symptomatology is rare, yet it continues to be unclear how long the virus can be transmitted sexually. This study highlights the endemic risk of Zika transmission, rather than the large-scale epidemic that may be exacerbated by international air travel.

Incorporating sexual transmission in modeling is important for various reasons. Although it is postulated that the majority of Zika infections occur via the mosquito vector, in areas with intense Zika activity, it would be challenging to quantify the impact of sexual transmission to an ongoing outbreak. What is more concerning is the fact that Zika infection remains asymptomatic in an estimated 80% of infected persons, regardless of the route of infection, meaning individuals may engage in sexual activity without any indications that they may have become infected or at risk for transmitting to others.⁵ Furthermore, Zika virus is usually cleared from the bloodstream; however, active viremia has been shown to persist in semen up to 188 days after infection.³² Zika virus transmission via sexual intercourse has been documented in more than 1000 cases. However, it is likely underestimated as a transmission vehicle because of the rare symptomatology and potential transmission without knowledge of infection, as well as infections occurring in endemic locations with varying routes of transmission.^7,33

Additional concern occurs with increased risk of microcephaly among pregnant women who become infected with Zika virus, especially during their first trimester.¹ This study was able to identify regions with higher risk of Zika infection to inform intervention opportunities to reduce risk among women early in their pregnancy. In addition, many women are unaware of their pregnancy status during the early stages of pregnancy, which may further complicate diagnosis and treatment. Although the magnitude of the relationship between Zika virus and newborns with microcephaly continues to be investigated,¹ our results suggest that developing and preparing intensive intervention approaches for the prevention of infection as well as treatment and care for families with affected newborns in these higher-risk areas should be prioritized. Enhanced vector control, Zika surveillance, and clinical management in these higher-risk areas will be critical for reducing the impact of a sustained Zika virus outbreak that may potentially occur particularly among economically challenged populations and communities that are least equipped to handle an outbreak.

Prevention planning for Zika virus needs to include basic communication strategies to inform higher-risk county residents of their risk of infection, how to prevent infection, and what treatment options exist if one becomes infected. Collaborative efforts need to respond to Zika virus outbreaks in communities. Partnerships need to consider public health infrastructure that provides resources for Zika virus testing, linking to treatment facilities, addressing the mosquito population, and developing appropriate prevention efforts to reduce the sexual transmission risk of the Zika virus.

Providing timely knowledge of risk factors, symptoms, ways to prevent mosquito biting, and access to tangible resources such as mosquito repellent and condoms should be considered as immediate interventions. Dynamic prevention messages need to be on the forefront of this effort to respond to the growing information of transmission risk. In addition, communication strategies should address diagnostic testing for pregnant women who suspect they have been infected, avoiding travel to high-risk areas, and answers to common concerns for women planning to conceive in the near future. This is particularly important information for US residents because of recommendations in highly affected countries that have warned aspiring mothers to delay pregnancy.

Health care systems need to be prepared for the testing, treatment, and management of cases that may develop more complicated health outcomes, including microcephaly and Guillain–Barré syndrome. Enhancing access to preventive resources is paramount, as many areas with higher concentrated disadvantage encounter limited resources.

Limitations

This study has limitations that should be noted. This study was limited to the contiguous 48 states in the United States. This analysis excludes US noncontiguous states and territories such as Hawaii, Puerto Rico, American Samoa, and the US Virgin Islands where locally acquired Zika infections have occurred and continue to occur.³⁴ However, our results remain robust regarding the potential higher-risk areas of the continental United States pending the exposure of the Zika virus throughout the A. aegypti mosquito population in the southern and southwestern states.

Second, pregnancy rates by county were not publicly available from the CDC’s Office of Vital Statistics and could not be obtained. This constrained us to use estimates from the US Census to approximate where births occur across the United States. These estimates, combined with percentage of women of childbearing age, provided a projection that we consider valid and useful.

Third, because of the limited amount of individual-level data regarding STI rates (e.g., age, race/ethnicity, income), this model may not fully account for such factors that facilitate STI transmission. Similarly, the model also assumes that STI rates are a valid surrogate marker for STI transmission rates. The use of STI rates as a surrogate for STIs has been widely used; however, it may underestimate the true STI transmission rate.^19,20

Fourth, these estimates do not account for travel-acquired infections, which may substantially increase transmission. However, many of the areas identified are impoverished and the likelihood of international travel is low because of economic constraints. In addition, the role of international travel may become less important once the Zika virus becomes introduced into the native mosquito population. Finally, these models assume that no aggressive Zika virus preventive measures were deployed, which would reduce the risk of infection.

Conclusions

Identifying areas in the United States that are at potential higher risk for Zika virus infection can help with planning primary disease prevention strategies and building infrastructure for diagnosis and treatment following infection. This study identified US counties at higher risk for a sustained Zika outbreak on the basis of past STI rates and likely A. aegypti presence. Timely strategies to communicate risk, control mosquito populations, and prevent disease transmission are imperative to preventing a large-scale Zika epidemic in the United States.