InvaCost Database. We selected all the entries in the regularly updated InvaCost database (version 4.0) referring to Aedes, a total of 10913, to compile the dataset presented here. This material, relevant to Ae. aegypti and Ae. albopictus, has been previously published11, 12, 13, 16.
Literature search. We updated our database with data from the literature published up to December 31, 2021, on the economic impacts of both Aedes species and the human disease-causing viruses they transmit. We used the same methodology as InvaCost13. To minimize the risk of omitting relevant materials, we carried out a bibliographic search in two online sources: the Web of Science (WoS) platform (https://webofknowledge.com/) and the PubMed repositories (https://www.ncbi.nlm.nih.gov/pubmed/). The search was done following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statements17 (see Supplementary Fig. 1 for details). We carefully composed appropriate search strings, and those considered as the most efficient from a set of potential candidates were consensually retained. Decisions were made after preliminary tests based on a handful of relevant articles provided by the authors13. Subsequently, we applied the following browse strings for Pubmed: (search string: cost-effectiveness[Title/Abstract] OR cost effectiveness[Title/Abstract] OR monetary[Title/Abstract] OR dollars[Title/Abstract] OR euros[Title/Abstract] OR sterling[Title/Abstract] OR DALY[Title/Abstract] OR expenditur*[Title/Abstract] OR economi*[Title/Abstract] OR cost of illness[Title/Abstract] OR cost-of-illness[Title/Abstract]) AND (zika[Title/Abstract] OR chikungunya[Title/Abstract] OR dengue[Title/Abstract] OR yellow fever[Title/Abstract] OR albopictus[Title/Abstract] OR aegypti[Title/Abstract]): and for WoS (search string: ("cost effectiveness" OR "cost-effectiveness" OR monetary OR dollars OR euros OR sterling OR DALY OR expenditur* OR economi* OR "cost of illness" OR "cost-of-illness")) AND (TS=(zika OR chikungunya OR dengue OR "yellow fever" OR albopictus OR aegypti)).
PRISMA methodology. The potentially relevant materials obtained from all the repositories were combined in a single file and screened for duplicates. The retrieved documents were individually assessed at three levels in accordance with PRISMA guidelines (titles, then abstracts, and finally full text; Supplementary Fig. 1). For the final investigation, we selected only relevant materials containing records of economic costs associated with Aedes aegypti and Aedes albopictus. The materials included in the review were required to meet the following inclusion criteria: (i) peer-reviewed article, book chapter or report by an official body; (ii) articles written in English, French, Italian, Portuguese or Spanish; (iii) mention of at least one cost record for a particular geographic area (municipality, region, country, continent) and a given period; (iv) costs exclusively associated with Aedes species or the diseases caused by the viruses they transmit (i.e., dengue, chikungunya, Zika or yellow fever); (v) costs that can be monetized, relating to medical services, management, and market losses. Exclusion criteria were: (i) records of only the cost per inhabitant (several papers estimated the medical costs per inhabitant, mostly for dengue) or the cost per patient without an adequate estimate for a particular geographic area and period; (ii) records of costs per disability-adjusted life years (DALY) without an estimate of the economic costs for a particular geographical area and period; (iii) future projections for vector control or vaccination methods (as they are potential costs); (iv) experimental field trials of vector control (as they cannot be transposed into operational actions for public health purposes at this stage). Two researchers double-checked each document to ensure transparency and validity.
Opportunistic search. Additional materials were also retrieved through an opportunistic search. Cost records were identified from other source materials when establishing the methodology for the project (e.g., when testing different search string combinations in the initial stages of the work), from bibliographic alerts set up by the review team or through directed grey literature searches of specific subjects we had identified as underrepresented (e.g., tourism costs)13 (Fig. S1).
Finally, we scrutinized all the relevant materials (including the “Aedes” references from InvaCost v 4.0, our update following the bibliographic search, and additional records identified from opportunistic searches of other sources) for data on economic costs (Supplementary Fig. 1). We also checked all entries in the database to ensure there were no duplicated records (i.e., multiple documents referring to identical cost records) or mistakes. The various steps in compiling the database are similar to those taken to generate the InvaCost database and were described in Diagne et al., 202013. Our procedures ensure that the database is, as far as possible, a comprehensive and up-to-date list of references.
Extraction, description and standardization of cost records. We scrutinized each document for any cost record that could be incorporated into the database, ensuring that the extracted data had the same structure and contained the same descriptors as the InvaCost database version 4.0 13. We inserted three additional columns: (i) a column listing the disease that the costs were associated with; (ii) a column giving details of further processing needed to estimate the costs where applicable; (iii) a column indicating who bore the burden of paying the cost. The final stage of the inclusion/exclusion took place during this data extraction phase. Full definitions and details of the description of each column in our dataset are given in Appendix 1. We screened the documents for reported cost records that were not retrieved by our initial literature search. Also, we assessed the original sources of the data, where available, to better characterize the reported cost (the ‘Previous materials’ column in Appendix 1). We extracted raw cost data as reported in the original sources. If several cost values were provided for a single situation (e.g., different cost records according to several management scenarios for the same invasive population) we estimated median values following the previously established criteria11,13. Any further processing to obtain cost records is reported in the ‘Additional processing’ column. When costs were recorded at different time scales in the same document, we chose those records that summarized most effectively the figure(s) presented in the study. If such a record was not obvious in the body of the text, we extracted every relevant cost record, and where several cost records were reported in a study, we also extracted the minimum and maximum records.
After extracting the costs, we standardized the raw cost data as cost records per year (‘Annualized cost estimate’ column) following the same methodology as used for the InvaCost database13. Each cost record was characterized by a number of descriptors (Appendix 1), with those specifically related to this study mentioned up above. Each of the Aedes costs in a particular country in a particular year was related to one or both Aedes species, based on historical data on Aedes colonization in each country or region when unspecified in the document (Appendix 2). In some instances, costs in the original sources were estimated for two or more diseases together (e.g., dengue and chikungunya; dengue and Zika; dengue, Zika and chikungunya). These were entered into the database and are referred to throughout the paper as DEN-CHIK-ZIKA. Once all the data had been collated, the final dataset was checked for errors. Each area was related to the country in which it is located, leading to some mismatches between the columns ‘Geographic region’ and ‘Official country’ as some countries have non-contiguous overseas territories. For instance, costs associated with invasive species in La Réunion (a French overseas department) were attributed to Africa as ‘Geographic region’ and France as ‘Country’ although France is on the European continent.
Characterizing the typology of each disease cost was mainly based on the InvaCost database 13 , with further sub-typologies (Table 1), as explained in the next section (Nature of the costs). We evaluated the reliability of the authors’ methodologies for obtaining cost records 11,12, 13 . The nature of the information retrieved and the choices made to characterize each cost are summarized in Appendix 1. The cost data were standardized to 2017 US$ using the same methodology as in Diagne et al., 2020a13 . We further transformed these costs into 2022 US$ values by multiplying them by an inflation factor of 1.193, based on Consumer Price Index from the World Bank ( https://data.worldbank.org ).
Table 1
Classification of the types of economic costs due to Aedes and Aedes-borne diseases.
Type of stakeholder
|
Type of cost
|
Damage / Losses
|
Management
|
Health providers (public or private bodies)
|
• Direct medical costs
• Indirect costs
|
• Surveillance (entomological and/or epidemiological)
• Vector control
• Prevention (entomological and/or epidemiological, includes communication campaigns)
|
Individuals
|
• Indirect costs (wage losses, debt)
• Direct medical and non-medical expenses (out-of-pocket costs)
• Losses of non-market values (quality of life)
|
• Personal protection
|
Community
|
• Indirect costs (lost productivity, disability benefits)
• Losses (in tourism, trade, economic growth)
|
• Research and funding
• Community based vector control
|
Nature of the costs
We developed a specific typology for the economic costs considered here, for which we considered the classification from previous works11, 18 as a basis, as well as the particular characteristics of invasive Aedes species and the diseases they transmit. This constitutes an original framework provided in this article (Table 1). We focused on monetary costs and made a major distinction between damage/losses and management costs.
We defined damage/loss costs as those associated with bearing the related losses or with repairing the damage, caused by Aedes and the diseases they transmit. Damage costs include several cost subcategories, widely described within the health economics literature. Direct medical costs are defined as the incurred expenses relating to diagnosis, hospital admission, hospitalization, ambulatory cases, patient care, and treatment of the illness. Non-direct medical costs are related to other medical expenses besides treatment, such as transportation, food and accommodation for patients and caregivers that are typically paid for by the patients or caregivers themselves. Indirect costs are those associated with lost productivity due to illness, morbidity or premature death. Losses are defined as the value lost from products/services traded on the market, which in this case covers tourism, trade, and economic growth, and are borne by communities and, in some cases, individuals. Some non-market values, defined as those that are not traded on markets, such as loss of quality of life, can be captured by ad hoc economic evaluation methods18. However, our analysis was limited to monetary costs derived from market values.
We defined management costs as those assigned to managing invasive Aedes vectors and Aedes-borne diseases. These include the costs of (entomological and epidemiological) surveillance, vector control and other (entomological and epidemiological) preventive actions, including vaccination, screening the blood-supply system, implementing communication campaigns and personal protection measures, and research and innovation.
Within these categories, we looked at how the costs are distributed among the various stakeholders (as proposed by Castro et al. 2017)19: health providers, the individual (or household), and the community. The “health providers” category includes public and private health providers, as well as other administrative/private bodies such as ministries of agriculture, education and tourism, NGOs or UN agencies, according to the type of intersectoral collaboration for Aedes-borne disease control20. Community refers to the administrative unit of analysis (e.g., neighborhood, village, municipality, province, state, country) bearing large-scale costs associated with Aedes invasion and Aedes-borne disease infections in the area19, but also refer to costs that represent a transfer in purchasing power from the general taxpayers (e.g., disability and welfare payments). At the individual or the household level, out-of-pocket costs represent direct payments made by those affected by the illness or by their family members/caregivers.
Data mining and analysis. After extracting the cost data, we used the InvaCost package to carry out the analysis21, updating it for our particular needs. We provide the R code and data in the supplementary material. After selecting for Aedes genus (genus = ‘Aedes’), we filtered out unreliable costs by selecting only for high reliability data (Method_reliability = ‘High'), observed costs (Implementation = ‘observed’)and costs based on direct observations or estimations (Acquisition_method = ‘Report/Estimation' and ‘Extrapolation’). Potential costs (i.e., estimated under hypothetical scenarios) were also excluded. For example, the costs if a disease became endemic in a place where there are currently only sporadic outbreaks, or the costs related to lost productivity due to premature mortality. Potential future costs, i.e., those not incurred but expected- for example, costs due to disease sequelae - were treated as a separate category. Besides these criteria, and to avoid data duplication in extracting the cost records, we took several steps to address the potential issue of double-counting. The methodology is standardized using a decision tree (Figure S2) to determine which cost would be retained where the same disease, species, cost type, geographical area or/and time frame were involved. Estimates of global-level economic burdens were excluded at this stage and were instead used to make comparisons with our results. Country-level cost records were selected over site-level costs when they overlapped in space and time. For example, if two studies estimated the costs of disease in Colombia in 2010, but one concerned a hospital in the city of Medellin, Colombia, and the other the entire country, the latter was retained. If two country-level records overlapped in time, we used complementary criteria to evaluate the reliability and completeness of the cost records13. If reliability was the same, the study that provided more detailed information on costs (clearly defined cost typologies, a longer time span) was retained. Although we made great efforts to avoid double counting, we are aware that “hidden” sources of double-counting could have been missed, particularly at the cost estimation level in the original sources. For example, if authors counted the cost of a doctor’s salary separately but also in the total medical care fee as two different values. To carry out temporal analyses of the costs, we standardized the temporal variable ‘Impact_year’ which refers to the year in which the costs were incurred. For one-year costs, the year of the cost was the same as the year under study. When the cost referred to a period of several years, the mean yearly cost was extended to cover all the years of the study period by applying the ‘expandYearlyCosts’ function of the InvaCost package21. There is, however, a temporal bias that we can at least acknowledge. We can expect a delay between the economic impact of an invasive species and the time when people start estimating the value of the impact, then publish their findings in a report or a journal. Hence, any analysis covering recent years will be based on incomplete data and is therefore highly likely to underestimate the actual costs. We also calculated the observed cumulative and average costs over a 5-year period using the summarizeCosts function of the InvaCost package21. This has been done for total costs, for damage and management costs, Aedes species costs and for specific disease costs (either for dengue, chikungunya or Zika, but not for DEN-CHIK-ZIKA, as we were unable to separate the contribution of each disease).
To adequately assess the behavior of temporal trends in damage and management costs, we used the modelCosts function of the ‘invacost’ package21, which fits multiple models to cost data as a function of time. We identified the best model using quantitative criteria (see Supplementary Note 1 and Supplementary Table 1) and used it to describe the temporal relationship for the accumulation of damage and management costs between 1990 and 2017. We deemed this specific time span as the most appropriate for assessing the temporal relationship for the accumulation of damage and management costs over time because of the sensitivity of the models to the time lag in cost reporting and the discontinuity of data for damage costs before 1990. The rational for this choice is further detailed in the Supplementary Note 1.