In order to address the inherent complexity of health decisions derived from factors such as uncertainty, conflicts of interest and subjectivity in the assessment of health technologies, amongst other aspects [1, 2], the multi-criteria decision analysis (MCDA) emerges as a structured methodology to support the decision-making processes, differing from other formal approaches due to its characteristic of mathematically modelling the subjectivity present in the f decision-makers’ judgement [3]. Hence, this methodological approach analytically guides decision support in a rational and transparent manner by aiming at the best choice in face of a set of alternatives, based on the decision maker’s preferences [4, 5].
Although the application potential of MCDA is evidenced in different areas such as health technology assessment, support to the selection of screening protocols and patient admission processes [4, 6, 7], there was a gap in the context of infectious diseases, like those transmitted by the Ae aegypti vector, given that the scientific productions that explore MCDA and tropical diseases focus on supporting the identification and risk areas spatial classification [8], mainly ignoring research opportunities to support the prioritization of interventions for the vector control, a process without consensus on the systematization of actions, despite the general guidelines recommended by the Pan American Health Organization (PAHO) [9].
In Brazil, the recurring mobilization for the control of arboviruses, especially those transmitted by Ae aegypti, is a testament to the great risks that the viruses transmitted by this vector have been causing to human health; with the proliferation of diseases such as Dengue, Zika and chikungunya; Ae aegypti has been the main cause for deaths by infectious diseases across the country [10, 11].
Mechanisms designed to mitigate the proliferation of these diseases are prioritized by public entities, the 16-year Bulletin (2003-2019) developed by the Health Surveillance Secretariat of the Ministry of Health presented the need for applied research to be considered as a strategic investment in a control intervention project [12].
The multi-criteria approach is flexible as well as capable of taking into account multiple factors considered important for public policy managers, for instance, the case in which the interventions applied are sustained for a pre-defined period of time without harming their overall performance [13]. The proper management of these and other factors are essential to the development of appropriate methods given that the action taken may not meet expectations due to inadequate planning [14], which justifies the need for a formal procedure that, at the same time, time that meets the general aspects recommended by PAHO, is adapted to the specificities of the analysed region.
Thus, the current work utilizes MCDA concepts to develop a multi-criteria model that supports the decision-making process for prioritizing Ae aegypti control interventions by a public health organization in the Northeast.
Four Operational Scenarios (OSs) are considered which can be distinguished by the risk severity of disease transmission. The expert evaluations, characteristics of operational scenarios and actions availability was observed to ensure that the means of action are adequate to the prompt response required by each OS.
Flexible and Interactive Tradeoff (FITradeoff) was the method utilized to build the model. The main contribution of the current work lies in the improvement of the use of the multi-criteria approach for health management and public policy, emphasizing disease vectors control, thus helping public policy managers not only in the environment in which the case study was conducted, but also in other geographic regions with similar climatic conditions, which could benefit from the support of a partial information method that requires less cognitive effort from the decision maker whilst preserving the robustness of the tradeoff elicitation procedure.
Multicriteria decision analysis
The multicriteria decision analysis, unlike other types of analysis, makes explicit a set of logical criteria that will serve as an accessible, accepted and exhaustive communication source, allowing for a thorough comprehension about preference shifts within the decision-making process [15]. Its potential to support the decision-making process is far superior than other traditional methods, as it enables complex decisions that include multiple criteria and simultaneously consider quantitative and qualitative data, in addition to involving multiple stakeholders [16].
Although all multi-criteria methods share the idea of systematic evaluation by decomposing the general alternative analysis into multiple criteria, there are several types of techniques, each associated with the structure of the decision problem [17].
[15] identifies four reference problems as follows: :
-
Choice or selection (problematic P.α) - amongst a set of actions, a subset is chosen to clarify the decision. This subset comprises “optimal” or “satisfactory” actions;
-
Sorting (problematic P.β) - each action is grouped into categories of specific characteristics;
-
Ranking (problematic P.γ) - realization of an arrangement with the regrouping of actions in an orderly manner;
-
Description (problematic P.δ) - through the description of actions and their consequences, the decision is clarified in term.
In this article, the problematic P.γ was utilised, which, according to [18], allows for the allocation of alternatives in order of increasing preference, based on the preference model. So that, in the spaces of actions under study, the construction of the ordered list of interventions that best applies to each of the operational scenarios analyzed will be carried out, allowing the comparison between all actions that make up the set of optimal alternatives.
FITradeoff method for ranking problematic
FITradeoff is composed by a group of segmented multicriteria methods in the utilization of the additive aggregation of preferences in Multi-Attribute Value Theory (MAVT). It was developed by the Centre for Development in Information and Decision Systems (CDSID) of the Federal University of Pernambuco (UFPE), under the coordination of Prof. Dr Adiel Teixeira de Almeida [19].
The method is distinguished in its category by overcoming frictions and inconsistencies commonly seen in the process of eliciting preferences by traditional Tradeoff, since it is motivated by flexible elicitation [20]. That is, it makes use of the traditional Tradeoff axiomatic structure and improves the execution of its procedures by directing cognitively easier questions to the DM, as it works only with strict preference statements; and assume partial or inaccurate information at the beginning of the elicitation process [18, 20].
For the ranking problem, FITradeoff seeks to arrange the alternatives in order of preference based on the understanding of the dominance relationships provided by Linear Programming Problems (LPP), according to the model described below [18].
$$max D(A_{i},A_{k})=\sum_{j=1}^{m}w_{j}v_{j}(A_{i}) - \sum_{j=1}^{m}w_{j}v_{j}(A_{k}) ~~~~~~(1)$$
s.t.
$$w_{1}> w_{2}> \cdots > w_{m}|\sum_{j=1}^{m}w_{j}=1 ~~~~~~~~~~~~~~~~~~~~~~~~~(2)$$
$$w_{j}v_{j}\left (x_{j}^{'} \right)> w_{j+1} ~~~ j=1 ~~~ to ~~ m-1 ~~~~~~~~~~~~~~~~~~(3)$$
$$w_{j}v_{j}\left (x_{j}^{\prime\prime} \right)< w_{j+1} ~~~ j=1 ~~~ to ~~ m-1 ~~~~~~~~~~~~~~~~~~(4)$$
$$w_{j}\geq 0 ~~~j=1,\cdots, m ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~(5)$$
The model seeks to maximize through the function MaxD(Ai,Ak) the global value (wjvj[A]) of the difference between each pair of alternatives (Ai,Ak) within a considered weight space. This weight space is built from the ranking of the scale constants (wi), according to (2) and superior limits definition (3) as well as inferior (4) presumed in each criterion. Furthermore, in (2) there is the normalization; and (5) the non-negativity of weights [18]. It is worth noting that, despite the term “weight” being used in the description of the variables w, these refer to scale constants and do not exclusively determine the degree of importance of each criterion.
Finally, after the operationalization of the LLPs, for each pair of alternatives, a matrix with the dominance relationships is provided, allowing the construction of the partial or complete alternatives ranking [18].
