Mosquito Control Activities During Local Transmission of Zika Virus, Miami-Dade County, Florida, USA, 2016

Janet C. McAllister; Mario Porcelli; Johana M. Medina; Mark J. Delorey; C. Roxanne Connelly; Marvin S. Godsey; Nicholas A. Panella; Nicole Dzuris; Karen A. Boegler; Joan L. Kenney; Linda Kothera; Lucrecia Vizcaino; Audrey E. Lenhart; John-Paul Mutebi; Chalmers Vasquez


Emerging Infectious Diseases. 2020;26(5):881-890. 

In This Article

Effect on Mosquito Abundance and Zika Infections

Because the response to the Zika outbreak in southern Florida was an emergency public health intervention, there was no time to set up proper controls. Therefore, we cannot evaluate properly using common comparison techniques the effect of the interventions. Instead, we used a changepoint analysis. A changepoint occurs if a time at which the statistical properties of the ordered sequence of observed case counts change. Case counts evaluated here are adult Ae. aegypti counts from BG Sentinel traps. A sequence can have >1 changepoint. In this analysis, the characteristic we assessed is the mean Ae. aegypti count change during the time observed. We consider 2 hypotheses: 1) Ae. aegypti counts during the entire period derive from a Poisson distribution with a constant mean, and 2) ≥2 time intervals exist, in each of which the Ae. aegypti counts derive from Poisson distributions with different means. We used a likelihood approach using binary segmentation, as described previously,[9] and implemented in the R package changepoint[10] to identify whether the data were consistent with hypothesis 1 or hypothesis 2. With each binary segmentation of the sequence, the Akaike information criterion with a correction for small sample size (AICcs) of the models with and without a changepoint, were computed. A model with the changepoint was considered a better fit if its AICc was smaller by ≥6 than the AICc of the model without the changepoint, which corresponds roughly with a type I error of 0.05.

We obtained dates of human cases from the Florida Deptartment of Health website ( Because the Little River red zone was declared when mosquito numbers were naturally dropping due to seasonality and mosquitoborne transmission had ceased by the time that red zone was identified, we did not evaluate changepoint analysis and transmission after spraying.

Although we cannot make a statistical association with the location of the changepoints in the other red zones, it is interesting that the first changepoints occurred after adulticide treatments began. In the Wynwood red zone that received both aerial adulticide and larvicide (Figure 3, panel A), this changepoint represents a large drop in mean Ae. aegypti counts. With further adulticiding and larviciding, the counts remained low and, in fact, dropped further on August 24. On approximately this date, insecticide applications stopped, and this date is followed by a third changepoint at which mean Ae. aegypti counts increased again.

Figure 3.

Changepoint in mean counts of Aedes aegypti mosquitoes from areas receiving adulticides and larvicides, Miami-Dade County, Florida, USA, 2016. Vertical lines indicate dates of changepoints for mean Ae. aegypti counts. A) Wynwood neighborhood; B) 10-mile region around the Wynwood neighborhood; C) combined Wynwood neighborhood (solid line) and 10-mile region around the Wynwood neighborhood (dotted line); D) southern Miami Beach; E) northern Miami Beach; F) Wynwood and Miami Beach combined. Points on the horizontal axis represent the first day of insecticide spraying; vertical lines show the first changepoint.

The only changepoint in the 10-mile area around Wynwood that received adulticide only occurred after the adulticiding began. We do not know what the mean Ae. aegypti counts were before August 9 (Figure 3, panel B). However, superimposing the counts for 10-mile region around the Wynwood neighborhood over those for the Wynwood neighborhood (as shown in Figure 3, panel C) showed that the mean Ae. aegypti counts for August 9 were comparable. Counts before August 9 might also have been comparable, but we have no way to verify that possibility. Ae. aegypti counts then increased in the region around Wynwood, whereas mean counts within Wynwood remained low. One possible explanation for this increase is that larviciding was not done in the 10-mile region around Wynwood. This observation reinforces the concept that both adulticiding and larviciding are needed to quickly reduce mosquito populations and maintain suppression. As reported previously,[7] detection of new Zika virus infections in Wynwood stopped after adulticide treatments began (Figure 4).

Figure 4.

Average number of Aedes aegypti mosquitoes and locally acquired Zika virus cases by epidemiologic week during the period of insecticide application, Miami-Dade County, Florida, USA, August–November 2016. A) Wynwood; B) southern Miami Beach; C) northern Miami Beach. Gray bars indicate mosquito counts; red line indicates Zika cases. Star indicates week cluster of locally acquired cases identified; square indicates first aerial adulticide application; triangle indicates first truck adulticide application; circle indicates first areawide (truck or aircraft) larvicide application. Cases are reported by date of symptom onset or date of specimen collection if no symptoms were present. Actual infection occurred before reporting date and is typically ≥1 week before the reporting date.

In the southern Miami Beach red zone, we again saw that the first changepoint occurred after the first adulticide treatments (Figure 3, panel D). There was a slight increase in mosquito count after the initial decrease (although this is not statistically a changepoint). Once the larvicide treatments began on September 6, mean Ae. aegypti counts decreased again, and 2 changepoints in mean counts followed the start of the larviciding. New cases of Zika virus ceased immediately after the first aerial adulticide treatments. However, for 4 weeks, single cases occurred roughly weekly after the last aerial adulticide treatment.

In the northern Miami Beach red zone, the first changepoint again occurred after the first adulticide treatments (Figure 3, panel E). After larviciding, the mean counts remained low and were followed by 2 more changepoints in mean Ae. aegypti counts. Again, for a third time no new Zika virus infections occurred after the first adulticide treatments.

We do not know what the Ae. aegypti counts were ahead of treatments or what would have occurred if treatments had not been initiated. However, graphing the Ae. aegypti counts from Wynwood and Miami Beach together (Figure 3, panel F) suggests that, before treatments began, the mosquito counts remained consistently high throughout the season (≈30–50 mosquitoes per trap). Only after the first adulticiding in each area did the mean mosquito counts drop statistically and vector-transmitted Zika virus infections cease.