Dengue is the most significant viral disease transmitted by arthropods around the world. It is distributed in tropical and subtropical zones and affects approximately 50–100 million people annually in over 100 endemic countries, where about half of the global population lives [1
]. In the Andean region of Latin America, Colombia is the country most affected by dengue [2
] and represents an economic cost of around $79.17 per ambulatory case and US $733.32 per hospitalized case [3
]. In this country, all places below 1800 m can support Aedes aegypti
, which is the main vector of the Dengue virus (DENV) in America [4
Colombia has been considered free of the disease since the 1970s due to vector eradication, but the epidemiological landscape has changed slowly over time, with a subsequent reinfestation by Ae. aegypti
. Since the 1980s, dengue outbreaks have begun to be common in some Colombian municipalities, and severe dengue epidemics spread rapidly across the country. Dengue epidemics occur every two or three years and have had a strong socio-economic impact on Colombia [5
Due to the absence of a specific treatment or vaccine for the control and prevention of the disease, the WHO recommends maintaining mosquito populations at low proportions, with the aim of reducing vector-human contact. As risk indicators, endemic places use entomological parameters as aedic indices, which measure the number of recipients positive for immature stages and the number of sampled houses. While these indices are useful for measuring risk in some places [6
], in other places, it is not possible to establish a positive relation between dengue case occurrence and aedic index values [10
]. To address this, several authors have proposed indices based on pupal stage detection [13
] as the best way of quantifying the population of adult mosquitoes, which are responsible for transmission. However, it is very difficult to quantify the number of pupae because this stage lasts only for a short time and is not likely found in the breeding populations.
These facts have led other authors to exclude entomological information and to develop indices based only on dengue cases, which, indirectly, provide evidence of the presence of infected mosquitoes. In [14
] three indices were developed that measured dengue risk in Taiwan and characterized the epidemic that occurred in 2002. Then, [15
] used a different classification methodology based on the standard deviation of the indices to develop a threshold, and they argued that the new classification was more accurate at establishing a risk scale than that proposed by [14
]. However, even though these approximations describe the epidemic behavior and how to classify the risk of dengue occurrence, any of them can be used for early detection of dengue outbreaks in an endemic period. Taking into account the difficulties of establishing the relation between entomological and epidemiological data in some places, the aim of this study was to evaluate the indices developed by [14
], which use only epidemiological information, in epidemic and endemic years. To do so, we selected as our study area a Colombian municipality that was endemic for dengue, for which correlations between entomological and epidemiological data were not observed [11
Dengue is considered an important vector-borne disease and a serious public health problem in tropical and subtropical countries [21
]. From 2008 to 2012, 1657 dengue cases were reported in Bello municipality (Colombia), which, based on the regular behavior of the disease in this location, exceeded the response capacity to control the disease. Bello has 10 main divisions (zones) in which a variable number of neighborhoods are grouped. Indices proposed by [14
] were applied to these 10 zones over five years (2008–2012) to identify dengue risk zones across space and time. We find that these indices are not accurate in detecting risk in endemic or epidemic years in Bello (Table 3
). Although [15
] implemented indices developed by [14
] by modifying only the system of classification, it was not enough to detect differential risk across space. Therefore, we concluded that the classification system was not the only reason for the issues related to risk detection; we mathematically modified these indices to improve their capability to predict dengue in different scenarios (endemic and epidemic).
Of these results, we deduced that some of the parameters evaluated were not appropriate for measuring risk in endemic periods and, therefore, we considered that [14
] states that dengue risk is inversely related to the number of epidemiological waves; however, this is not applicable to endemic periods because a large wave longitude (temporal duration) is not necessarily responsible for outbreaks. It is clear that short epidemiological waves can also produce dengue outbreaks. In this sense, we took into account the largest wave amplitude in a year plus the number of weeks involved in these waves. The results obtained showed that the modified indices had a greater capacity for classifying dengue risk both in endemic and epidemic years, and our results correspond to the epidemiological data from Bello (Table 3
). Therefore, we modified the β and γ indices because they are calculated using the epidemiological wave value.
After evaluating the performance of the modified indices, we merged them to build a scale of risk. We first used the methodology implemented by [15
], which uses standard deviation to classify the risk; however, this methodology is not sensitive enough to detect variation in risk across space and time (Figure 3
). We then calculated the risk scale for the modified indices merged using the LISA methodology, which takes into account the values of the surrounding areas. With these results, we checked that the calculated scale was comparable with the epidemiological landscape for dengue in Bello (Table 3
). Additionally, it is necessary to have a very fine classification because the idea behind this type of methodology is to reduce the cost of interventions by focusing on risk without neglecting zones that may be epidemiologically important. With our results, the indices can be classified into four categories that indicate low, moderate, high, and very high risk, which is a didactical form that can be used and understood by health authorities in each endemic place. Modified indices can also be used in other endemic places because they are insensitive to differences in dengue incidence, as was shown in the different scenarios proposed (Table 2
). Additionally, classification made with the modified indices, show high correlation with dengue case occurrence, ranging between 70% and 85%, indicating its goodness as a predicting tool.
In this study, we could effectively classify the risk zones with only epidemiological data because we did not include entomological, climatic, socioeconomic variables, or environmental variables, which have been relevant in other studies to predict dengue occurrence [14
]. The simplicity of this procedure allows implementation by health entities so that intervention decisions can be made quickly. Additionally, the data used are available for all places, and it is not necessary to procure other information that is difficult to access.
Finally, the modified indices and methodology used in this research could be set to the finest temporal scales to build a permanent system of surveillance to determine, in real-time, zones for which it is mandatory to apply control strategies and monitor the effectiveness of these interventions.
This study showed the necessity of modifying the risk indices proposed by [14
] so that they could be used in areas during epidemic and endemic periods. The mathematical modification was achieved such that the risk measure scale is not affected by the dengue incidence of the area, due to the definition of epidemic being relative in the area. Therefore, the importance of the maximum value of the epidemiological wave for endemic scenarios was shown and indicated that a few successive weeks with dengue cases also represent a risk for a dengue epidemic.
If the dispersion of the disease from zone to zone is taken into account, LISA classification is more adequate for measuring risk than SD because it calculates risk based on the surrounding zones, whereas SD only allows for evaluating the risk of the zone for which it is being calculated.
Finally, we observed that using only epidemiological data, it is possible to determine the risk without introducing external variables such as climatic, entomological or environmental variables. The results of this risk determination could be used to generate early warning systems and to apply control measures to reduce the economic costs generated by dengue.