Of all countries in the Western Hemisphere, Brazil has the highest economic losses caused by dengue fever. We evaluated the cost-effectiveness of a novel system of vector surveillance and control, Monitoramento Inteligente da Dengue (Intelligent Dengue Monitoring System [MID]), which was implemented in 21 cities in Minas Gerais, Brazil. Traps for adult female mosquitoes were spaced at 300-m intervals throughout each city. In cities that used MID, vector control was conducted specifically at high-risk sites (indicated through daily updates by MID). In control cities, vector control proceeded according to guidelines of the Brazilian government. We estimated that MID prevented 27,191 cases of dengue fever and saved an average of $227 (median $58) per case prevented, which saved approximately $364,517 in direct costs (health care and vector control) and $7,138,940 in lost wages (societal effect) annually. MID was more effective in cities with stronger economies and more cost-effective in cities with higher levels of mosquito infestation.

Dengue viruses cause ≈50 million infections annually worldwide, and ≈1% of these infections require hospitalization because of dengue hemorrhagic fever (

The most accurate method of assessing dengue risk by vector surveillance is one that specifically counts dengue vectors that are actively in search of a blood meal: adult female

Fixed-position traps designed to capture gravid mosquitoes (e.g., MosquiTRAPs) (Ecovec SA, Belo Horizonte, Brazil) have been developed to reduce personnel costs and directly measure adult female mosquito abundance in Brazil (

We evaluated the cost-effectiveness of supplementing vector control methods with MID in 21 cities in Minas Gerais State, Brazil, after use during 2 dengue seasons. We also identified factors that affected efficacy and cost-effectiveness of MID. We reported direct savings for health care costs and vector control activities separately from indirect savings for lost wages so that results are relevant to public health budgets and societal concerns.

Monthly dengue cases during January 2007–June 2011 were obtained from each municipality in Minas Gerais, Brazil, by using Sinan Net (Information System for Notification of Grievances), a publicly available database of the Health Ministry of Brazil. Dengue cases were expressed as incidence per 100,000 inhabitants on the basis of the Brazilian Institute of Geography and Statistics (Rio de Janeiro, Brazil) 2010 population census.

MID was implemented in 21 cities in Minas Gerais during April 2009–June 2011. These cities are dispersed throughout the state in areas that included a range of population sizes and incidences (

Spatial distribution of 21 cities tested with Monitoramento Inteligente da Dengue (Intelligent Dengue Monitoring System [MID]), Minas Gerais, Brazil, 2009–2011. A). Size of city centroids (n = 218) (circles) is proportional to population size. B) Size of city centroids (n = 147) (circles) is proportional to total dengue fever incidence during 2007–2011. Gray circles indicate cities that never implemented MID, and black circles indicate cities that implemented MID during mid-2009–June 2011. Areas of higher and lower total incidence are positively clustered with each other (Moran’s I, p<0.0001). Cities that implemented MID and those that had not implemented MID are distributed throughout areas of high and low incidence. Only cities with populations >15,000 are shown. Incidence data were not available for all cities.

Dengue incidence was strongly seasonal, and outbreak probability varied substantially between cities (

Changes in incidence of dengue fever in 21 cities that implemented Monitoramento Inteligente da Dengue (Intelligent Dengue Monitoring System [MID]), Minas Gerais, Brazil, mid-January 2007–June 2011. A) Annual incidence in 21 cities that implemented MID (bars outlined in black) and 147 cities that had not implemented MID (bars outlined in gray). Horizontal lines in boxplots indicate medians of 1,000 medians. Whiskers indicate ± 2.7 SD. Circles indicate points that fall outside ± 2.7 SD. B) Distribution of population sizes in cities that implemented MID. C) Time that MID was implemented in each city. D) Median relative increase (RI) in incidence for cities that implemented MID versus cities that had not implemented MID. RI was calculated as the sum of monthly incidence after MID was implemented divided by the sum of monthly incidence before MID was implemented for the same number of months. For cities that implemented MID, the median is a single value for the 21 cities. For cities that had not implemented MID, 21 cities with the same distribution of population sizes as MID cities were selected at random 1,000 times and their median relative differences during the same set of time frames were calculated. Horizontal line in the boxplot indicates median of 1,000 medians. Whiskers indicate ± 2.7 SD. Circles indicate points that fall outside ± 2.7 SD.

To compare this large sample size in a case–control format to only 21 MID cities, we generated 1,000 random sets of 21 control cities with the same population distribution as the MID cities. Next, we calculated the relative difference in incidence (RI) for the same period before and after the start of surveillance for each treatment city (i.e., incidence for x time before MID/incidence for x time after MID). Likewise, for each set of the control cities, we calculated RI using the distribution of time frames in each group of treatment cities (_{control} – _{MID}). Under the null hypothesis that MID had no effect at decreasing RI, the median of the 1,000 differences would be 0. We tested this hypothesis using a sign test. The alternative hypothesis was that the median of the 1,000 differences would be significantly >0 if MID decreased the RI of treatment cities.

We identified factors that affected the effectiveness and cost-effectiveness of MID by using a generalized linear model (γ distribution, log link) with either RI or US dollars/prevented case as response variables. Factors considered were population size (PS); distance to 3 large populations (D3L); distance to 3 high-incidence populations; a ranking system for the effectiveness of using MID (PED); a measure of average mosquito infestation during the dengue season in 2011 (IMFA); population density; income per capita; and an index between 0 and 1 that included employment, income, education, and health, all with equal weight. Distances were the sum of Euclidian distance to 3 cities with population size (D3L) or density in the 90th percentile. We fit each variable individually and fit all possible linear combinations of the 8 variables.

We chose between competing models by using delta Akaike Information Criterion (ΔAIC = AIC of intercept only model – AIC of target model). A higher ΔAIC indicates a better model of the data. When comparing nested models that differed by only 1 factor, 2 AIC points is considered a significant difference (α = 0.05). Statistics for all single-variable models, model selection results, and fits of the best multivariable and full models are shown in of

We estimated the number of cases prevented by MID by predicting the number of cases that would have occurred in the absence of MID and taking the difference between those and the number of observed cases (i.e., cases prevented/year = predicted cases in the absence of MID [_{i}_{i}O_{i}_{i}_{i}_{control} – RI_{MID}

Effectiveness of Monitoramento Inteligente da Dengue (Intelligent Dengue Monitoring System [MID]), Minais Gerais, Brazil, mid-2009–mid 2011. Predicted number of dengue fever cases prevented per year during the time of MID are plotted against the annual incidence of dengue fever in each city during the same time. K is a percentage value of the population size in a city. Error bars indicate 2 SE. A) 29,533 cases were prevented when K = 50%. B) 24,263 cases were prevented when K = 20%. C) 16,578 cases were prevented when K = 10%. D) 9,219 cases were prevented when K = 5%. Shaded symbols distinguish population size classes as follows: black circles indicate 18,000–21,000; gray circles indicate 35,000–60,000; white circles indicate 70,000–90,000; triangles indicate 100,000–140,000; squares indicate 150,000–300,000.

All costs were in US dollars. Costs per dengue case were taken from the report of Sheppard et al. (

Total costs for MID in the 21 cities were measured directly by Ecovec SA (

Product or service | Cost in US dollars (%) |
---|---|

Royalties to UFMG and FAPEMIG | |

MID† | 29,918.96 (2.0) |

MI-Virus† | 38,894.65 (2.6) |

Consumables (licensing) | |

MosquiTRAP† | 77,533.50 (5.2) |

Sticky card | 112,776.00 (7.5) |

AtrAedes† | 131,572.00 (8.8) |

Web software | 21,000.00 (1.4) |

Mobile software | 96,041.00 (6.4) |

Services | |

MI-Vírus kit and analyses | 58,485.00 (3.9) |

MID | 61,303.96 (4.1) |

Shipping and freight | |

MI-Virus and traps | 11,056.45 (0.7) |

Stationary and materials | 668.40 (0.0) |

Technical and supervision (employees and taxes) | |

Technical support at Ecovec SA, 12 h/d | 115,499.00 (7.7) |

Technical support at cities visited | 80,208.00 (5.4) |

Full-time biologist | 372,520.54 (24.9) |

Technical visits on site | 19,200.00 (1.3) |

Taxes | 269,270.66 (18.0) |

Total | 1,495,948.13 (100.0) |

*MID, Monitoramento Inteligente da Dengue (Intelligent Dengue Monitoring System); UFMG, Universidade Federal de Minas Gerais; FAPEMIG, Fundação de Amparo a Pesquisa do Estado de Minas Gerais; MI-Virus, Intelligent Virus Monitoring System; MosquiTRAP, fixed-position trap designed to capture gravid mosquitoes; AtrAedes, synthetic ovipostion attractant. †Manufactured by Ecovec SA (Belo Horizonte, Brazil).

For each treatment city, we calculated the direct, indirect, and total costs of dengue. Direct costs comprised medical and nonmedical direct costs, as well as vector control and MID. Indirect costs comprised costs for lost wages and MID costs in treatment cities. Costs for MID cities were calculated from the number of observed cases (divided into ambulatory and hospitalized case-patients). Similarly, the estimated costs of dengue in the absence of MID were divided into ambulatory and hospitalized case-patients by multiplying the sum of the number of observed cases plus the number of prevented cases by the proportion of observed cases in persons who were ambulatory or hospitalized. The dollars saved annually were calculated by subtracting the cost of dengue in MID cities from the predicted cost of dengue if MID were not implemented. Underreporting was not accounted for because we had no city-specific data to inform estimates. Costs were not discounted because we considered all cases to be nonfatal and our study period was only 2.5 years.

The annual incidence of dengue in control cities varied more widely than in treatment cities, and the median annual incidence in treatment cities was generally higher (

Mean relative difference in incidence (RI) of dengue fever cases for treatment cities grouped by population size using Monitoramento Inteligente da Dengue (Intelligent Dengue Monitoring System), Minas Gerais, Brazil, mid-2009–mid 2011. Horizontal line indicates mean RI for the 1,000 median RI of control city sets. Error bars indicate 2 SD. Error bars for the largest population size group are too small to be shown. The black dot is an outlier that was excluded from the general linear model results.

The most parsimonious generalized linear model of RI in MID cities included PS and IDFM (

Under the assumption that dengue could affect 30% of a population, we estimated that the number of cases prevented by MID annually in the largest cities (>130,000 inhabitants) was 2,300–3,900 (

Effectiveness of Monitoramento Inteligente da Dengue (Intelligent Dengue Monitoring System [MID]), Minas Gerais, Brazil, mid-2009–mid-2011. Predicted number of dengue fever cases prevented per year during the time of MID are plotted against the annual incidence of dengue fever cases in each city during the same time. A total of 27,191 cases were prevented. Cases prevented/year = predicted cases in the absence of MID (_{i}_{i}O_{i}_{i}_{i}

Cost-effectiveness of and savings from Monitoramento Inteligente da Dengue (Intelligent Dengue Monitoring System [MID]), Minas Gerais, Brazil, mid-2009–mid-2011. A) For cost-effectiveness, the number of US dollars (USD$) spent per dengue fever case prevented is plotted against the annual incidence of dengue fever cases during MID for the city. Each point represents cost-effectiveness for a city. Points are coded by population size classes. Horizontal line indicates average cost-effectiveness ($227) per case prevented. B) Savings for each cost component from the benefits of MID. Direct savings include only health care, nonmedical direct savings, and vector control savings. Indirect savings include only savings in the work force. Total savings include direct and indirect savings. Negative values indicate dollars lost because of implementing MID. Vertical line indicates 0. Shaded symbols distinguish population size classes as follows: black circles indicate 18,000–21,000; gray circles indicate 35,000–60,000; white circles indicate 70,000–90,000; triangles indicate 100,000–140,000; squares indicate 150,000–300,000.

Accurate estimates of dengue incidence and its economic effects are more limited (

The trend of increased effectiveness in larger populations might not be significant in the multivariable model (which includes IDFM) because PS and IDFM showed a positive correlation (

PED, a measure of MID quality, was not correlated with effectiveness or cost-effectiveness of MID. This result suggests that variation in the force of infection between cities overwhelmed differences in PED, the current measure of PED is inaccurate, or both. The relationship between mosquito infestation and human incidence is highly variable in space and time (

One caveat to our method of estimating cost-effectiveness is that it was necessary to estimate the number of susceptible hosts in the absence of MID (we used 30%) to predict the number of cases that would be prevented. Although 30% is not unreasonable based on previous studies (

Although MID showed an average cost-effectiveness value of $227 (median $58) per case prevented in Minas Gerais, the average value increased to $616 in 6 moderately sized cities (population 73,000–117,000) that did not show any savings in direct costs (

Furthermore, cities in which MID was implemented had historically high dengue incidences relative to control cities (mean ± SD 2007 and 2008 were 549.1 ± 592 in MID cities and 240.4 ± 567.6 in control cities). Thus, average estimates of cost-effectiveness may be high in cities in which MID was implemented. However, factors determining incidence patterns in a given city, such as population immunity, infrastructure, or human behavior, may not be static over time because high population immunity is not protective against novel serotypes (or genotypes with high forces of infection) and human behavior and infrastructure are continually changing. A predictive model of serotype dynamics across cities formulated on the basis of serologic data would be useful for decisions on which cities should implement MID so that the most cost-effective strategy can be achieved statewide.

Our study showed that MID is generally effective for decreasing case loads and suggested that an MID strategy is theoretically better than other strategies. Although MID cost-effectiveness varied between cities, implementation of MID saved hundreds of thousands of dollars on health care and ≈7 million dollars in lost wages statewide, and half the cities had cost-effectiveness values

Monitoramento Inteligente da Dengue (Intelligent Dengue Monitoring System), Minas Gerais, Brazil.

Monitoramento Inteligente da Dengue (Intelligent Dengue Monitoring System), Minas Gerais, Brazil. Table 1, Factors that affected effectiveness and cost-effectiveness; Table 2, Top 5 multivariate models; Table 3, Fit of multivariable models; Table 4, Cost-effectiveness and savings for each city.

K.M.P. was supported by the Research and Policy for Infectious Disease Dynamics Program of the Science and Technology Directorate, US Department of Homeland Security; and the Fogarty International Center, National Institutes of Health. A.E.E. was supported by Conselho Nacional de Desenvolvimento Científico e Tecnológico (PRONEX-Dengue), Fundação de Amparo à Pesquisa do Estado de Minas Gerais, Departamento de Ciência e Tecnologia–Ministério da Saúde, and Instituto Nacional de Ciência e Tecnologia–Dengue.

Dr Pepin is an epidemiologist at Colorado State University in Fort Collins, Colorado. Her research interests are epidemiologic modeling and statistical analyses of infectious disease data to facilitate decision making for public health practices.