Review of artificial intelligence platforms for early pandemic detection in Latin America

By: Diana Carolina Valladares Merejildo

Acknowledgement Note:
This project was carried out as part of the “Carreras con Impacto” program during the mentorship phase. You can find more information about the program in this entry.

Abstract

The research aims to analyze artificial intelligence (AI)-based epidemiological surveillance platforms for pandemic prevention and preparedness in Latin America (LATAM), using the COVID-19 pandemic as a case study. To this end, traditional COVID-19 detection methods, such as Polymerase Chain Reaction (PCR) and rapid antigen tests, will be compared with next-generation technologies like Next-Generation Sequencing (PacBio SMRT, Oxford Nanopore Technologies (ONT), and Ion Torrent) and AI platforms (HealthMap and BlueDot).

This analysis seeks to highlight the importance of AI in health systems to provide a timely response to health emergencies. Additionally, recommendations will be provided to public health decision-makers in LATAM, aiming to strengthen health systems' capacity to detect and manage infectious disease outbreaks, especially in resource-limited contexts.

Introduction

Early pandemic detection is crucial for effective public health emergency management, as it enables the implementation of preventive and control measures that can save lives and reduce economic costs. A study by the World Health Organization (2020) indicates that timely identification of infectious outbreaks is a key component in mitigating their impact. In this context, AI-based surveillance platforms have shown remarkable potential to improve early outbreak detection, which is particularly relevant in Latin American countries, where public health resources are often limited (Gonzalez et al., 2020).

Moreover, the COVID-19 pandemic underscored the urgent need to implement advanced tools to strengthen epidemiological surveillance and the capacity to respond to health emergencies. During this crisis, it became evident that traditional surveillance systems were often insufficient to address the rapid spread of infectious diseases (Camhaji, E., Galindo, J., & Oquendo, C., 2022; Ocaranza, C., 2020). In this context, innovative platforms like Bluedot and HealthMap have emerged as effective solutions, employing machine learning algorithms and data analysis to predict and monitor outbreaks in real time (Sullivan et al., 2019). These tools not only enable faster case identification but also facilitate informed decision-making by health authorities (Brownstein et al., 2008).

This critical scenario has motivated the choice of research topic on the effectiveness of AI-based surveillance platforms in early pandemic detection. The results obtained from this study can serve as a starting point to evaluate how new technologies can be integrated into LATAM's health systems, thereby improving responses to future emergencies. In this sense, the research aims to offer recommendations to public health decision-makers, adapted to the specific realities and challenges of LATAM, contributing to strengthening their response capacity to future infectious disease outbreaks.

Background

Global public health priorities have evolved significantly over the past few decades, especially in the face of the growing threat of pandemics. The World Health Organization (WHO) has identified biological risk as one of the main concerns, emphasizing the need for robust epidemiological surveillance systems to detect and quickly respond to infectious disease outbreaks (WHO, 2021).

This focus is crucial, as pandemics not only affect public health but also have profound social and economic repercussions. For example, the COVID-19 pandemic resulted in a global economic crisis, with millions of jobs lost and a recession in many countries. In Latin America, an estimated 30 million people fell into poverty due to lockdown measures and the disruption of economic activities, highlighting the interconnection between public health and economic well-being (ECLAC, 2021). Therefore, countries' ability to anticipate and manage these risks has become a priority on the global health agenda.

Biological risk assessment refers to the probability that a pathogen will cause harm to public health, and its management requires effective surveillance that allows for the early identification of outbreaks. According to the World Economic Forum (2020), the COVID-19 pandemic has exposed the flaws in existing surveillance systems, underscoring the need to integrate advanced technologies, such as artificial intelligence (AI), to improve emergency detection and response. AI-based surveillance platforms can analyze large volumes of data in real-time, facilitating the identification of patterns that could indicate an imminent outbreak (Gonzalez et al., 2020). This capability is essential to address global health priorities, which include not only pandemic prevention but also the mitigation of their impact on vulnerable communities.

Additionally, the experience gained during the COVID-19 pandemic has led to a reevaluation of epidemiological surveillance strategies. The WHO has urged countries to strengthen their public health systems by implementing innovative technologies and collaborative approaches (WHO, 2021). This includes creating surveillance networks that integrate data from various sources, allowing for a more agile and effective response to biological threats. In this context, research on the effectiveness of AI-based surveillance platforms becomes fundamental, as it can provide valuable insights into how these tools can be used to improve preparedness for future pandemics.

Epidemiological surveillance is defined as the systematic process of collecting, analyzing, and interpreting data on health problems in a population. Its main goal is to detect disease outbreaks, monitor health trends, evaluate the effectiveness of interventions, and provide useful information for research and public health policies (Thacker & Stroup, 1994). This approach is crucial for the prevention and control of diseases, especially in the context of pandemics.

Pandemics, such as COVID-19, have demonstrated that traditional surveillance systems can be inadequate for managing large-scale, real-time outbreaks. The need for a more advanced and efficient approach has become evident, particularly in regions with limited resources (Smith et al., 2021).

Traditionally, the detection of viruses like COVID-19 has relied on methods such as Polymerase Chain Reaction (PCR) and Rapid Antigen Tests. On the one hand, PCR is a widely used molecular technique that amplifies specific DNA or RNA sequences. This method is highly sensitive and specific, allowing for the detection of small amounts of pathogen genetic material in biological samples (Mackay et al., 2002). However, PCR requires specialized equipment and trained personnel, making it costly and slow compared to other methods. On the other hand, antigen tests identify specific viral proteins. These tests offer rapid results and are less expensive, but their lower sensitivity can lead to false negatives, especially in the early stages of infection (Dinnes et al., 2020).

In contrast to these traditional methods, Next-Generation Sequencing (NGS) has revolutionized the way pathogens are analyzed. Technologies such as Ion Torrent, which uses pH changes to detect nucleotides, enable rapid and efficient genomic analysis (Rothberg et al., 2011). Oxford Nanopore Technologies (ONT) is another innovative methodology that allows real-time sequencing by detecting changes in electrical current as DNA molecules pass through a pore. Its portability and ability to sequence long DNA strands are significant advantages (Loman et al., 2015). Similarly, PacBio SMRT (Single-Molecule, Real-Time) technology offers high fidelity in long reads, making it particularly useful for detecting genetic variants in complex analyses (Eid et al., 2009).

In addition to these methods, AI platforms such as Bluedot and HealthMap represent an innovative approach to epidemiological surveillance. Bluedot uses machine learning algorithms to analyze public health data and mobility patterns, enabling early identification of epidemic outbreaks (Sullivan et al., 2019). Meanwhile, HealthMap combines epidemiological surveillance data, social media, and media reports to track outbreaks in real-time, facilitating a rapid response from health authorities (Brownstein et al., 2008).

Thus, the impact of technology on pandemic detection becomes evident. The integration of AI into epidemiological surveillance not only improves early outbreak detection but also enables faster and more effective responses (Gonzalez et al., 2020). However, it is crucial to recognize that significant challenges must be overcome, including technical, ethical, and implementation barriers.

In this regard, the applicability of disease detection methods in Latin America reflects both opportunities and obstacles. Factors such as healthcare infrastructure, access to technology, and staff training play a fundamental role in the effectiveness of these methods. Therefore, it is essential to address these issues to maximize the potential of these technologies in epidemiological surveillance in the region.

As is evident, public health in Latin America faces significant challenges in implementing disease detection technologies. These challenges become even more complex when we consider the diversity of available methods and their application in clinical and community contexts. In this sense, it is crucial to analyze case studies that illustrate the effectiveness of different methods in detecting pathogens, especially in the context of the COVID-19 pandemic.

In the area of traditional methods, a study titled "Detection of SARS-CoV-2 in patients with COVID-19 in Brazil" (2020) is particularly relevant. This work used the PCR technique to detect the SARS-CoV-2 virus in nasopharyngeal samples. Through a thorough evaluation, the effectiveness and sensitivity of PCR were analyzed in various groups of patients, demonstrating its capacity to identify infections at different stages of the disease (Lima et al., 2020).

Additionally, another study titled "Performance of antigen-based rapid tests for the diagnosis of COVID-19 in Argentina" (2021) evaluated the accuracy of rapid antigen tests compared to PCR. The results showed that although antigen tests provided faster results, their sensitivity was lower than that of PCR. This raises important considerations regarding their use in clinical settings, especially in situations where precise detection is critical (Gonzalez et al., 2021).

Moving towards next-generation methods, sequencing has proven to be a fundamental resource. A study titled "Genomic epidemiology of SARS-CoV-2 in Chile" (2021) used Ion Torrent technology to sequence the virus's genome, providing valuable information about the evolution and transmission of SARS-CoV-2 in the country (Cáceres et al., 2021).

Another relevant study in this field is "Long-read sequencing of SARS-CoV-2 genomes from Latin America" (2022), which used PacBio SMRT technology to obtain long reads of the virus genome. This approach facilitates the analysis of virus variants and their relationship with transmission, contributing to a better understanding of the epidemiology in the region (Zapata et al., 2022).

Regarding AI-based methods, the study "Using AI to predict the spread of COVID-19 in Latin America" (2020) highlights AI's ability to model virus spread. Using Bluedot, this study analyzed mobility and epidemiological data to predict outbreaks, demonstrating how artificial intelligence can be an effective tool for public health surveillance (Sullivan et al., 2020).

Similarly, the use of platforms like HealthMap for real-time surveillance has proven beneficial. The study "Real-time surveillance of COVID-19 in Brazil using HealthMap" (2020) collected and visualized data on COVID-19 outbreaks, integrating information from social media, such as Twitter and Facebook, and official reports from health organizations like Brazil’s Ministry of Health and the World Health Organization (Sullivan et al., 2020). This data integration enabled a faster and more effective response to the spread of infectious diseases and provided an updated overview of the epidemiological situation, essential for informed decision-making (Brownstein et al., 2020).

Objectives

General Objective

Review of AI-based epidemiological surveillance platforms for the early detection of future pandemics in LATAM countries.

The purpose is to improve the response capacity to public health emergencies and contribute to strengthening health systems in the region.

Specific Objectives

• Analyze the effectiveness of different epidemiological surveillance methods (traditional, next-generation, and AI-based) in the early detection of pandemics.
• Identify the benefits and limitations of implementing AI technologies in LATAM.
• Develop recommendations based on the findings of the research.

This approach will not only contribute to strengthening surveillance but also improve the response to infectious disease outbreaks in resource-limited settings.

In this way, this research aims to provide a comprehensive framework that allows policymakers and public health professionals in LATAM to adopt the most appropriate tools for epidemiological surveillance. By addressing the opportunities and challenges presented by AI platforms, the study's results are expected to serve as an initial step toward optimizing the detection and response to health emergencies, thereby strengthening health systems in the region.

Methodology

The methodology of this research is based on a mixed approach that combines a systematic literature review with the analysis of empirical data obtained through analytical matrices evaluated using the weighted factors method (Charity Entrepreneurship, n.d.). This structure allows for a comprehensive assessment of disease detection methods in Latin America, covering both traditional and emerging methods (next-generation and AI-based). It is characterized as descriptive-evaluative and relies on the collection of information from secondary sources, including both qualitative and quantitative data.
Taking as a reference the work of Gabriela Paredes Villafuerte titled “Impact of Massive Sequencing Technologies on the Prevention and Detection of Pandemic Pathogens in Latin America and the Caribbean” (2024), matrices were developed with technical, economic, human talent, social, political, and management criteria, enabling a clear comparison and identification of patterns and trends in the results.
To establish the criteria, an exhaustive review of scientific and technical literature was conducted, including journal articles, conference reports, public health organization documents, and databases such as PubMed, Scopus, and Google Scholar.
The matrices were developed to categorize and evaluate each detection method according to specific criteria, as outlined below:

Technical Criteria (60 points)

As previously mentioned, the matrix developed by Gabriela Paredes was used to establish the evaluation criteria, adapting the definitions to the methods under study. In this context, the aspects considered in the evaluation of traditional and next-generation methods are as follows:

• Sensitivity (8.94 points): Defined as the ability of a diagnostic test to correctly identify individuals who have the disease, sensitivity is expressed as a percentage of true positives and is associated with a low risk of false negatives when high. A high score (8.94 points) indicates the method correctly detects more than 90% of individuals with the disease, while a medium score (5.96 points) refers to a performance of 70% to 90%. A low score (2.98 points) means performance is below 70%.
• Specificity (8.94 points): This criterion assesses the method’s ability to correctly identify those without the disease (true negatives). At a high level (8.94 points), true negatives are identified in more than 90% of cases. The medium level (5.96 points) corresponds to a performance of 70% to 90%, and at the low level (2.98 points), specificity is below 70%.
• Reproducibility (6.38 points): Measures the consistency of results when the test is repeated under similar conditions. A high level (6.38 points) indicates consistency above 90%, while the medium level (4.25 points) reflects between 70% and 90%. At the low level (2.13 points), consistency is below 70%.
• Accuracy (8.94 points): Refers to the closeness of the results to the true value, combining the test’s sensitivity and specificity. A high score (8.94 points) reflects accuracy above 90%, while the medium score (5.96 points) is between 70% and 90%, and the low score (2.98 points) is below 70%.
• Response Time (6.38 points): Evaluates how quickly results are obtained. At the high level (6.38 points), results are delivered in less than 1 hour. The medium level (4.25 points) ranges from 1 to 4 hours, and at the low level (2.13 points), results take more than 4 hours.
• Limit of Detection (6.38 points): This criterion refers to the smallest amount of a substance that can be detected with a specified level of confidence. A high level (6.38 points) indicates that less than 1 pg/mL can be detected, the medium level (4.25 points) covers 1 to 10 pg/mL, and the low level (2.13 points) exceeds 10 pg/mL.
• Robustness (6.38 points): Measures the method’s ability to provide accurate and reliable results under different conditions. A high level (6.38 points) indicates strong capacity to maintain accuracy, the medium level (4.25 points) reflects moderate capacity, and the low level (2.13 points) suggests poor reliability.
• Interferences (3.83 points): This criterion evaluates the influence of other substances present in the sample on the test results. A high level (3.83 points) means no influence from other substances, while the medium level (2.55 points) indicates minimal influence, and the low level (1.28 points) suggests significant interference.
• Compatibility with Automation (3.83 points): Measures the degree to which the method integrates into automated systems. A high level (3.83 points) implies a high degree of integration, increasing efficiency and reducing errors, while the medium (2.55 points) and low (1.28 points) levels reflect moderate and low integration, respectively.

On the other hand, the aspects to be evaluated for AI platforms are:

• Sensitivity (8.94 points): This is evaluated by how effective the tool is at correctly identifying positive cases. A high score (8.94 points) indicates that the method correctly identifies more than 95% of cases, while a medium score (5.96 points) reflects performance between 80% and 94%. A low score (2.98 points) indicates performance below 80%.
• Specificity (8.94 points): This measures how effective the tool is at confirming that truly healthy individuals are not incorrectly identified as sick. As with the previous criterion, a high score (8.94 points) indicates that the method correctly identifies more than 95% of cases, while a medium score (5.96 points) reflects performance between 80% and 94%. A low score (2.98 points) corresponds to performance below 80%.
• Reproducibility (6.38 points): This is evaluated by performing the test multiple times on the same sample to verify the consistency of the results. A high score (6.38 points) indicates a coefficient of variation (CV) of less than 5%, a medium score (4.25 points) ranges from 5% to 15%, and a low score (2.13 points) exceeds 15%.
• Accuracy (8.94 points): Refers to measurement error. A high score (8.94 points) means an error of less than 5%, a medium score (5.96 points) covers errors between 5% and 15%, and a low score (2.98 points) indicates an error greater than 15%.
• Response Time (6.38 points): This is evaluated by how quickly the tool can provide results after data entry. A high score (6.38 points) indicates responses in less than 30 minutes, a medium score (4.25 points) between 30 minutes and 4 hours, and a low score (2.13 points) indicates more than 4 hours.
• Limit of Detection (6.38 points): Refers to the tool's ability to identify signs of disease outbreaks in early stages using indirect data. A high score (6.38 points) reflects over 90% precision in detection, a medium score (4.25 points) ranges from 70% to 89%, and a low score (2.13 points) indicates less than 70%.
• Robustness (6.38 points): This is measured by the tool's ability to maintain accuracy and reliability under different operating conditions. A high score (6.38 points) shows effectiveness in more than 90% of conditions, a medium score (4.25 points) between 70% and 89%, and a low score (2.13 points) less than 70%.
• Interferences (3.83 points): This evaluates the tool’s ability to handle data interferences that could affect the quality of the results. A high score (3.83 points) indicates success in over 90% of cases, a medium score (2.55 points) between 70% and 89%, and a low score (1.28 points) below 70%.
• Compatibility with Automation (3.83 points): This is evaluated based on how well the tool integrates with automated systems for data collection and analysis. A high score (3.83 points) indicates compatibility in more than 90% of environments, a medium score (2.55 points) between 70% and 89%, and a low score (1.28 points) below 70%.

Economic Criteria (12 points)

• Cost (4 points): This criterion is classified as high, medium, or low. A high score (4 points) indicates that the method has a low cost, making it accessible and economically favorable. At the medium level (2.67 points), the cost is considered moderate. In contrast, a low performance (1.33 points) refers to a high cost, which could limit the method's implementation.
• Infrastructure (4 points): This criterion evaluates the infrastructure needed to implement the method. A high score (4 points) reflects that minimal infrastructure is required, facilitating its adoption. At the medium level (2.67 points), moderate infrastructure is needed, while at the low level (1.33 points), advanced infrastructure is necessary.
• Equipment Requirements (4 points): This criterion measures the complexity of the equipment needed for the method. A high score (4 points) indicates that only basic equipment is required, simplifying its use. At the medium level (2.67 points), specialized equipment is needed, and at the low level (1.33 points), very specialized equipment is required, which can pose additional challenges in terms of cost and accessibility.

Human Talent Criteria (12 points)

• Training Requirements (4 points): This criterion is classified as high, medium, or low. A high score (4 points) indicates that minimal training is required, facilitating the method's adoption by staff. At the medium level (2.67 points), moderate training is needed, which may involve additional time and resources. At the low level (1.33 points), extensive training is required, potentially representing a significant barrier to implementation.
• Ease of Use (4 points): This criterion evaluates how accessible the method is for users. A high score (4 points) reflects that the method is very easy to use, promoting its acceptance and widespread use. At the medium level (2.67 points), the method is considered moderately easy to use, while at the low level (1.33 points), it is classified as difficult to use, which may lead to rejection by users.
• Safety (4 points): This criterion measures the level of risk (biological or cyber) associated with the use of the method. A high score (4 points) indicates a low risk, suggesting that the method is safe for implementation. At the medium level (2.67 points), a moderate risk is identified, and at the low level (1.33 points), there is a high risk, which could undermine confidence in the method and its acceptance.

Social Criteria (10 points)

• Adaptability and Scalability (2.5 points): This criterion is classified as high, medium, or low. A high score (2.50 points) indicates that the method is easily adaptable and scalable, facilitating its implementation in various situations. At the medium level (1.67 points), the method is considered moderately adaptable and scalable, while at the low level (0.83 points), it is classified as difficult to adapt and scale, which may limit its use.
• Applicability in Various Contexts (2.5 points): This criterion evaluates the versatility of the method in different environments. A high score (2.50 points) suggests that the method can be used in a wide variety of contexts, increasing its relevance. At the medium level (1.67 points), it is considered moderately versatile in its applicability, and at the low level (0.83 points), it is limited to specific contexts.
• Accessibility (2.5 points): This criterion measures how accessible the method is to the target population. A high score (2.50 points) indicates that the method is highly accessible, promoting its widespread use. At the medium level (1.67 points), it is classified as moderately accessible, while at the low level (0.83 points), it is considered to have low accessibility, which could restrict its implementation.
• Acceptability (2.5 points): This criterion evaluates the level of acceptance of the method by healthcare personnel. A high score (2.50 points) reflects a high level of acceptance, suggesting that the method is well-received. At the medium level (1.67 points), moderate acceptance is observed, and at the low level (0.83 points), it is classified as having low acceptance, which could hinder its implementation.

Political and Management Criteria (6 points)

• Regulation and Approval (2 points): This criterion is classified as high, medium, or low. A high score (2 points) indicates that the method is fully approved, facilitating its implementation. At the medium level (1.33 points), the method is partially approved, which may involve certain restrictions. At the low level (0.67 points), the method is not approved, significantly limiting its use.
• Government and Political Support (2 points): This criterion evaluates the level of support the method receives from authorities. A high score (2 points) suggests strong support, which can facilitate its implementation and sustainability. At the medium level (1.33 points), moderate support is observed, while at the low level (0.67 points), there is little to no support, potentially hindering its adoption.
• Sustainability (2 points): This criterion measures the method's ability to endure over time. A high score (2 points) indicates that the method is highly sustainable, suggesting it can persist without significant external resources. At the medium level (1.33 points), it is classified as moderately sustainable, and at the low level (0.67 points), it is considered to have low sustainability, potentially jeopardizing its continuity.

The data obtained from the matrices were analyzed both qualitatively and quantitatively. Descriptive analysis techniques were applied to summarize the characteristics of each method, and comparisons were made between them. The analysis focused on identifying the strengths and weaknesses of each method within the Latin American context, as well as evaluating their impact on epidemiological surveillance.

Results and discussion

Presentation of Results

This section presents the results obtained through a matrix evaluating various COVID-19 detection tools and their contribution to epidemiological surveillance. The matrix includes traditional, next-generation, and AI-based methods, analyzing technical, economic, human talent, social, and political and management criteria. Each of these factors has been weighted to calculate a total score that reflects the effectiveness and feasibility of each method in the Latin American context.

Table 1.- Matrix with total scores by criteria and method.

Note: M/C refers to method and criterion, T to technical, E to economic, TH to human talent, S to social, PG to political and management, TOT to total value, TRAD to traditional methods, NG to next-generation methods, IA to artificial intelligence, PCR to polymerase chain reaction, PRA to rapid antigen test, PB SMRT to PacBio SMRT, ONT to Oxford Nanopore Technologies, IO to Ion Torrent, BD to BlueDot, and HM to HealthMap.

Technical Criteria

Table 2.- Matrix with total technical scores by criterion and method

Note: SEN refers to sensitivity, ESP to specificity, REP to reproducibility, EXA to accuracy, TdR to response time, LdD to limit of detection, ROB to robustness, INT to interference and CcA to compatibility with automation.

Traditional Methods

PCR obtained a total score of 50.22, standing out as an effective and reliable method for disease detection. In terms of sensitivity, PCR shows a high capacity to correctly detect individuals who have the disease, scoring 8.94, which implies 95% efficacy in identifying true positives. This positions it as a reliable diagnostic method.

In terms of specificity, PCR also scores high at 8.94, reflecting its ability to correctly identify those who do not have the disease, with a true negative rate of 98%. This feature is crucial for avoiding false positives and ensuring diagnostic accuracy.

The reproducibility of the method is equally notable, with a score of 6.38, indicating that results are consistent when the test is repeated under similar conditions, exceeding 90%. This ensures that the results are reliable and can be replicated in different tests.

PCR’s accuracy, also evaluated at 8.94, demonstrates that the results are very close to the true value, with 95% precision. This high level of accuracy is essential for trust in the diagnoses made.

However, PCR’s response time is considered low, scoring 2.13, as it can take 4 to 6 hours to provide results. This can be a drawback in situations requiring rapid diagnoses.

The limit of detection is also rated with a score of 2.13. This limit represents the smallest amount of substance that can be detected with a specified confidence level, which is higher than 10 pg/mL, suggesting that the method may not be the most sensitive for early detection in some cases.

In terms of robustness, PCR scores 6.38, indicating a great capacity to provide accurate and reliable results under various conditions. This is essential for its use in different laboratory settings.

The influence of interferences on the results is classified as medium, with a score of 2.55, suggesting minimal influence of other substances present in the sample on the test outcome. This feature is important to ensure the validity of the results.

Finally, automation compatibility is rated at 3.83, reflecting a high degree of integration into automated systems. This not only increases process efficiency but also reduces the risk of human error during testing.

On the other hand, Rapid Antigen Tests scored a total of 42.55, revealing varied performance across different technical criteria. In terms of sensitivity, this method scored low at 2.98, implying a 60% capacity to correctly detect individuals who have the disease. This limitation can lead to a significant number of false negatives, which is a critical concern in diagnostics.

In terms of specificity, the score is medium, reaching 5.96. This indicates that the method has an 85% capacity to correctly identify individuals who do not have the disease, which is an improvement over sensitivity but still might be insufficient in critical clinical contexts.

The reproducibility of the method is rated as medium, with a score of 4.25. This means that results are consistent when the test is repeated under similar conditions, with a reliability range of 70% to 90%. Although there is some consistency, variability can be a factor to consider in interpreting results.

In terms of accuracy, the test scores a medium of 5.96, suggesting that the results are close to the true value at 85%. This accuracy is reasonable but inferior to more established methods like PCR.

A positive aspect of rapid antigen tests is their response time, rated high with a score of 6.38. The method can provide results in 15 to 30 minutes, which is advantageous in situations requiring rapid diagnostics.

The limit of detection is also high, with a score of 6.38, indicating that the method can detect substance amounts below 1 pg/mL with a specified confidence level. This suggests a good capacity to detect the presence of the antigen, although overall sensitivity remains a weakness.

The robustness of rapid antigen tests is rated as medium, with a score of 4.25. This indicates a moderate ability to provide accurate and reliable results under different conditions, which may limit its application in variable environments.

Regarding interferences, this method stands out with a high score of 3.83, suggesting that there is no significant influence of other substances in the sample on the analysis result. This is a positive aspect contributing to the reliability of the results.

Finally, automation compatibility is rated at 2.55, indicating a moderate degree of integration into automated systems. This could limit efficiency and the reduction of human error in the diagnostic process.

Next-Generation Methods (Sequencing)

PacBio SMRT achieved a total score of 50.22. In terms of sensitivity, it scored high at 8.94, reflecting a 99.9% capacity to correctly detect individuals who have the disease, making it a highly reliable method for diagnostics. Specificity is also high, with a score of 8.94, indicating that it can correctly identify those who do not have the disease, achieving 99.9% true negatives.

Reproducibility is rated at 6.38, suggesting that results are consistent in repeated tests under similar conditions, exceeding 90%. PacBio SMRT’s accuracy is equally impressive, with a score of 8.94, indicating that the results are very close to the true value, reaching 99.9%.

However, the response time of this method is considered low, with a score of 2.13, as results may take 24 to 48 hours to become available. The limit of detection is also rated 2.13, suggesting that the method can detect amounts above 10 pg/mL, which could limit its effectiveness in early detection.

The robustness of PacBio SMRT is high, with a score of 6.38, indicating its capacity to provide accurate and reliable results under various conditions. In terms of interferences, it is rated as medium, with a score of 2.55, indicating minimal influence of other substances on the results.

Automation compatibility is rated at 3.83, showing a high degree of integration into automated systems, increasing efficiency and reducing human error.

Oxford Nanopore Technologies (ONT) also achieved a total score of 50.22. In terms of sensitivity, it scored a high 8.94, indicating it can correctly detect 97% of individuals who have the disease. Specificity is equally high, with a score of 8.94, reflecting a 98% true negative rate.

Reproducibility is rated 6.38, implying that results are consistent when repeated under similar conditions, exceeding 90%. Accuracy is also high, with a score of 8.94, suggesting that results are close to the true value, at 95%.

However, the response time is considered low, with a score of 2.13, as results may take 24 to 48 hours. The limit of detection is also rated 2.13, indicating that the smallest detectable amount is above 10 pg/mL, which may affect early detection.

Robustness is rated 6.38, indicating a great ability to provide accurate results under various conditions. In terms of interferences, it is rated medium, with a score of 2.55, suggesting minimal influence of other substances on the results.

Automation compatibility is high, with a score of 3.83, reflecting good integration into automated systems.

Ion Torrent also obtained a total score of 50.22. In terms of sensitivity, it scored 8.94, indicating it can correctly detect 98% of individuals who have the disease. Specificity is also high, scoring 8.94, meaning it correctly identifies healthy individuals at 98%.

Reproducibility is rated 6.38, reflecting consistent results in repeated tests under similar conditions, exceeding 90%. Accuracy is rated 8.94, indicating that results are very close to the true value, with 98% precision.

Response time is low, with a score of 2.13, as results may take 24 to 48 hours. The limit of detection is rated 2.13, suggesting it can detect amounts above 10 pg/mL, which could limit its effectiveness in early detection.

Ion Torrent’s robustness is high, with a score of 6.38, suggesting it can provide accurate and reliable results in different conditions. In terms of interferences, it is rated medium, with a score of 2.55, indicating minimal influence of other substances on the results.

Finally, automation compatibility is rated 3.83, reflecting a high degree of integration into automated systems.

AI Methods

BlueDot achieved a total score of 49.79, based on several technical criteria reflecting its performance. In terms of sensitivity, it received a score of 5.96. While BlueDot is effective at identifying early outbreak signals using indirect data, its ability to correctly detect individuals with the disease is limited, achieving less than 95%. This limitation affects its effectiveness in identifying all positive cases.

Specificity was also rated 5.96. BlueDot filters relevant news events, but interpreting multiple sources can result in false alarms. Although it avoids many false positives, it doesn’t do so perfectly, which justifies the medium score. In terms of reproducibility, it received a high score of 6.38, indicating that its standardized methodology and algorithms allow for consistency in detecting patterns and outbreaks, with a coefficient of variation (CV) of less than 5%.

The accuracy of BlueDot was rated 8.94, reflecting its high capacity to predict disease outbreaks, maintaining a prediction error of less than 5%. This high level of precision is essential for trust in its diagnostics. Additionally, the platform's response time is notable, scoring 6.38, as it offers real-time alerts in under 30 minutes, enabling swift action in emergencies.

However, the limit of detection was rated as medium (4.25), indicating that its ability to detect outbreaks in early stages ranges between 70% and 89%. While BlueDot can identify early signals, it doesn’t always do so with the desired precision. The platform’s robustness was also evaluated at 4.25, suggesting that it maintains good accuracy under various operating conditions, though it may face challenges with unusual data.

In terms of interferences, BlueDot excelled with a high score of 3.83, indicating that it effectively handles interferences by using algorithms that filter out noise and irrelevant data in more than 90% of cases. Finally, its automation compatibility was also rated 3.83, reflecting good integration into automated systems for data collection and analysis.

HealthMap, on the other hand, achieved a total score of 51.91, excelling in several technical criteria. Its sensitivity was also rated 5.96, indicating that, while it has a good capacity to identify outbreaks by analyzing data from multiple sources, its sensitivity ranges from 80% to 94%. This is because the method does not directly measure individuals with specific diseases, limiting its effectiveness in some contexts.

HealthMap's specificity was also rated 5.96. The platform has a good ability to distinguish between real outbreaks and false positives, but it is not infallible, explaining the medium score. In terms of reproducibility, it received a high rating of 6.38, highlighting its automated system and systematic analysis of large volumes of data, ensuring notable consistency.

HealthMap's accuracy was rated 8.94, suggesting it is quite precise in identifying emerging outbreaks, maintaining an error rate of less than 5% in the information provided. Its response time was also highlighted, with a score of 6.38, due to its ability to offer almost immediate alerts in under 30 minutes, allowing for swift and effective action.

The limit of detection was rated as medium (4.25), indicating that while HealthMap can detect outbreaks, its ability depends on the quality and quantity of available information. The platform's robustness was rated 6.38, showing that it handles large volumes of data from diverse sources well, providing alerts despite variability in the data.

Finally, in terms of interferences, HealthMap excelled with a high score of 3.83, as it uses advanced algorithms to filter out potentially noisy or inaccurate data. Its automation compatibility was also rated 3.83, reflecting good integration into automated systems for data collection and analysis.

1.2 Economic Criteria

Table 3.- Matrix with total economic scores by criterion and method.

Note: C refers to cost, I to infrastructure, R to equipment requirements, TRAD to traditional, NG to new generation and AI to artificial intelligence.

Traditional Methods

PCR received a total score of 5.33. Its cost was rated as low (1.33), reflecting the high cost associated with these tests. The required infrastructure was also evaluated as low (1.33) due to the need for advanced infrastructure, which may limit accessibility in certain settings. In terms of equipment requirements, it received a medium score of 2.67, as specialized equipment is needed, which can increase the initial investment.

On the other hand, Rapid Antigen Tests received a total score of 12.00, excelling in economic criteria. In terms of cost, it was given a high score of 4, indicating a low implementation cost. The infrastructure was also rated high (4), as it requires minimal infrastructure, facilitating its implementation in various contexts. Additionally, equipment requirements were rated high (4), as only basic equipment is needed, contributing to its accessibility and low cost.

Next-Generation Methods (Sequencing)

PacBio SMRT received a total score of 5.33. Its cost was rated as low (1.33), indicating the high cost associated with using this technology. The infrastructure was also rated low (1.33) due to the need for advanced infrastructure, which may be difficult to access in certain settings. In terms of equipment requirements, it received a medium score of 2.67, as specialized equipment is needed, potentially limiting its economic viability.

ONT achieved a total score of 6.67. Its cost was rated as medium (2.67), reflecting a moderate cost that may be significant for some organizations. The infrastructure was rated low (1.33), indicating the need for advanced infrastructure, which can complicate implementation. Lastly, equipment requirements were rated medium (2.67), as specialized equipment is required, which can increase initial costs.

Ion Torrent also received a total score of 6.67. Its cost was evaluated as medium (2.67), similar to ONT, indicating a moderate cost. The infrastructure was rated low (1.33), reflecting the need for advanced infrastructure, limiting its accessibility in certain contexts. Equipment requirements were evaluated as medium (2.67), as specialized equipment is also required.

AI Methods

BlueDot received a total score of 10.67 in economic criteria. In terms of cost, it was given a medium score of 2.67. While BlueDot doesn’t incur costs associated with physical diagnostic tests, it requires a subscription and access to the platform, which can be significant for some organizations, especially in resource-limited countries. However, its ability to provide early alerts can prevent higher costs associated with undetected outbreaks, resulting in a good cost-benefit ratio.

The infrastructure required for BlueDot was rated high (4.00). This method requires minimal infrastructure compared to physical diagnostic tools, as it only needs internet access and basic computing systems for data visualization and analysis. This facilitates its adoption in settings with limited infrastructure. In terms of equipment requirements, it also received a high score of 4.00, as only standard computing devices and internet access are needed, avoiding the need for specialized equipment or reagents that could increase costs.

HealthMap received a total score of 9.33. In terms of cost, it received a medium score of 2.67. While HealthMap has relatively low costs compared to physical diagnostic tests, it may incur significant expenses related to IT infrastructure, technical staff, and data acquisition. These costs can add up, affecting the system's economic viability.

The infrastructure for HealthMap was rated as medium (2.67). It requires robust IT infrastructure, including servers and high-speed internet connections, which can be considerable, though generally accessible in urban and developed areas. Technical support is also essential for platform maintenance and updates. Equipment requirements were rated high (4.00), as the system needs standard computing and networking hardware, without requiring specialized devices. However, the equipment must be capable of handling large data volumes and providing real-time analysis.

1.3 Human talent criteria

Table 4. Matrix with total human talent scores by criteria and method

Note: NdC refers to training needs, FdU to ease of use, S to security, TRAD to traditional, NG to new generation and AI to artificial intelligence.

Traditional Methods

PCR achieved a total score of 12.00. In terms of training requirements, it was given a high score of 4, indicating that minimal training is required for its use. Ease of use was also rated high at 4, as it is very user-friendly, making it accessible to various users. In terms of safety, it also received a high score of 4, reflecting a low risk associated with its implementation.

On the other hand, Rapid Antigen Tests received a total score of 9.33. Its training requirements were rated low (1.33), indicating that extensive training is required for effective use. However, ease of use was rated high at 4, as it is very simple to use. In terms of safety, it received a high score of 4, showing a low risk in its implementation.

Next-Generation Methods (Sequencing)

PacBio SMRT scored a total of 12.00. Regarding training needs, it received a high score of 4, indicating that minimal training is required. Ease of use was also rated high at 4, showing that it is very user-friendly. Finally, in terms of safety, PacBio SMRT received a high score of 4, indicating low risk in its implementation.

ONT achieved a total score of 12.00. Similar to PacBio, it was given a high score of 4 for training requirements, implying that minimal training is needed. Ease of use was also rated high at 4, suggesting it is very easy to use. In terms of safety, ONT also received a high score of 4, indicating low risk associated with its use.

Ion Torrent received a total score of 10.67. Its training requirements were rated high (4), indicating minimal training is needed. However, ease of use was rated medium at 2.67, suggesting that it is moderately easy to use and may require more familiarization. In terms of safety, Ion Torrent received a high score of 4, reflecting low risk in its implementation.

AI Methods

BlueDot received a total score of 12.00 in criteria related to human talent. In terms of training requirements, it was given a high score of 4.00. This is because the platform is designed to be accessible, requiring only basic knowledge of computing and data analysis. The necessary training is neither specialized nor complex, especially compared to more advanced diagnostic tools.

Ease of use was also rated high at 4.00. BlueDot facilitates its implementation through an intuitive graphical interface that simplifies both data interpretation and alert management. This allows even users with limited technical training to use the platform effectively. Additionally, in terms of safety, it received a high score of 4.00, as BlueDot does not involve physical risks associated with handling biological samples. Its focus is on data protection and privacy, implementing advanced cybersecurity measures.

HealthMap achieved a total score of 10.67. In terms of training requirements, it received a medium score of 2.67. While the required training is not excessive, users must become familiar with the user interface and learn how to interpret epidemiological data. Understanding reports and alerts, as well as performing basic analyses, is essential for effective platform use, requiring solid knowledge of epidemiological concepts and basic skills in data analysis tools.

HealthMap’s ease of use was rated high at 4.00. The platform is designed with an accessible interface that allows users with varying levels of technical experience to interact with it effectively. The clear and visual presentation of information facilitates quick interpretation of data and alerts. In terms of safety, HealthMap also received a high score of 4.00. It focuses on data protection and privacy, using standard security practices like encryption and access controls. While it adheres to high protection standards, there is always some risk when handling large volumes of data.

1.4 Social criteria

Table 5. Matrix with total social scores by criterion and method.

Nota: AyE refiere a adaptabilidad y escalabilidad, ADC a aplicabilidad en diversos contextos, ACC a accesibilidad, ACE a aceptabilidad, TRAD a tradicionales, NG a nueva generación e IA a inteligencia artificial.

Traditional Methods

PCR achieved a total score of 10.00, excelling in social criteria. In terms of adaptability and scalability, it received a high score of 2.50, indicating that it is easily adaptable and scalable across different contexts. Its applicability in various contexts was also rated 2.50, reflecting its effectiveness in a wide variety of situations. Regarding accessibility, it received a high score of 2.50, indicating that it is highly accessible. Lastly, acceptability was also high (2.50), showing its wide acceptance within the health community.

On the other hand, Rapid Antigen Tests received a total score of 10.00. Its adaptability and scalability were rated high (2.50), indicating that it is easily adaptable to different contexts. The applicability in various contexts was also evaluated at 2.50, reflecting its effectiveness in a wide variety of situations. In terms of accessibility, it received a high score of 2.50, indicating that it is highly accessible. Finally, acceptability was also high (2.50), suggesting a positive reception among users.

Next-Generation Methods (Sequencing)

PacBio SMRT obtained a total score of 8.33. In terms of adaptability and scalability, it received a high score of 2.50, indicating that it is easily adaptable and scalable. Its applicability in various contexts also received a high score of 2.50, reflecting its effectiveness in various situations. However, its accessibility was rated as medium (1.67), suggesting that it is moderately accessible. Acceptability was also evaluated as medium (1.67), indicating a moderate reception within the community.

ONT received a total score of 8.33. Its adaptability and scalability were rated high (2.50), reflecting its flexibility. The applicability in various contexts was also high (2.50), suggesting good effectiveness in different environments. However, accessibility was rated as medium (1.67), indicating that it is moderately accessible. Acceptability was also medium (1.67), indicating a moderate reception among users.

Ion Torrent achieved a total score of 8.33. In terms of adaptability and scalability, it was rated high (2.50), indicating that it is easily adaptable. Its applicability in various contexts also received a high score of 2.50, reflecting its effectiveness. However, accessibility was rated as medium (1.67), indicating that it is moderately accessible. Acceptability was also rated as medium (1.67), suggesting a moderate reception within the community.

AI Methods

BlueDot received a total score of 9.17 in social criteria. In terms of adaptability and scalability, it received a high score of 2.50. BlueDot is designed to be flexible and adjust to various levels of detail based on the demand for information about emerging outbreaks. It has proven effective in adapting to different regions and expanding its surveillance coverage. However, its adaptability may be limited by data availability and local infrastructure.

Its applicability in various contexts also received a high score of 2.50. BlueDot has shown effectiveness in identifying outbreaks and disease patterns globally, suggesting a high capacity to operate in different geographic and population settings. Nevertheless, effectiveness may vary depending on the quality of available data and local infrastructure. In terms of accessibility, it received a medium score of 1.67. Although it is designed to be accessible from various locations, its availability may depend on the technical and technological support in the regions, as well as associated costs, which can be a barrier in resource-limited areas.

Finally, BlueDot's acceptability was rated high (2.50). It is generally accepted by health institutions that use it for epidemiological surveillance, although its acceptance among the general population may be lower, as it is primarily aimed at health professionals.

HealthMap also achieved a total score of 9.17. In terms of adaptability and scalability, it received a high score of 2.50. The platform is flexible and can adjust to different surveillance scenarios, using global health data to include new outbreaks and emerging diseases. Its adaptability is influenced by the integration of new data sources.

Its applicability in various contexts was also rated 2.50. HealthMap is effective in a variety of settings, both urban and rural, and can handle data from different countries. However, its applicability may be affected in areas with low internet connectivity. In terms of accessibility, it received a medium score of 1.67. Although it is globally accessible, its use depends on internet availability and the users' ability to interact with the platform.

Regarding acceptability, HealthMap was rated high (2.50). It is well accepted in the public health community for its ability to provide useful information about outbreaks. However, its acceptance may be lower among the general population due to its focus on epidemiological surveillance.

1.5 Political and management criteria

Table 6. Matrix with total political and management scores by criteria and method.

Note: R&A refers to regulation and approval, GAand P to government and political support, S to sustainability, TRAD to traditional, NG to new generation and AI to artificial intelligence.

Traditional Methods

PCR achieved a total score of 6.00, excelling in political and management criteria. In terms of regulation and approval, it received a high score of 2.00, indicating full approval. It also scored high in government and political support (2.00), reflecting strong institutional backing. Lastly, in terms of sustainability, it was rated as high (2.00), suggesting that it is highly sustainable.

Similarly, Rapid Antigen Tests also scored a total of 6.00. Like PCR, it received a high score of 2.00 for regulation and approval, indicating full approval. In terms of government and political support, it was also rated 2.00, showing strong backing. Finally, it received a high score of 2.00 for sustainability, indicating strong long-term viability.

Next-Generation Methods (Sequencing)

PacBio SMRT achieved a total score of 6.00. In terms of regulation and approval, it received a high score of 2.00, indicating full approval. It also received strong government and political support (2.00), reflecting strong institutional backing. Finally, sustainability was also rated 2.00, suggesting it is highly sustainable.

ONT received a total score of 6.00. For regulation and approval, it was given a high score of 2.00, indicating full approval. It also received strong government and political support (2.00), showing significant institutional backing. In terms of sustainability, it was rated as high (2.00), suggesting that it is very sustainable.

Ion Torrent also achieved a total score of 6.00. In terms of regulation and approval, it received a high score of 2.00, indicating full approval. It also received strong government and political support (2.00), reflecting significant institutional backing. Lastly, sustainability was rated as high (2.00), indicating strong long-term sustainability.

AI Methods

BlueDot received a total score of 3.33 in political and management criteria. In terms of regulation and approval, it was given a low score of 0.67. As an AI platform for epidemiological surveillance, BlueDot is not subject to the specific regulations required for physical diagnostic tools. Although it complies with data security and privacy standards, its regulation in public health is limited.

Government and political support was rated as medium (1.33). BlueDot has gained recognition and support in specific contexts, collaborating with governments and international organizations during major outbreaks. However, the level of support can vary depending on the region and available funding, which may result in inconsistent backing.

In terms of sustainability, it was also rated as medium (1.33). BlueDot is designed to adapt to new threats and changes in disease patterns, contributing to its long-term sustainability. Its subscription-based business model and collaborations with public health organizations support its viability, though it depends on financial backing and continued adoption by these institutions.

HealthMap received a total score of 5.33. In terms of regulation and approval, it was given a medium score of 1.33. As a surveillance tool based on data analysis, it is not subject to the same regulations as physical diagnostic tests. However, its use involves regulatory considerations related to data privacy and protection, such as compliance with GDPR and HIPAA.

Government and political support was rated as high (2.00). HealthMap has received significant support from governmental agencies and international organizations like the WHO, reinforcing its credibility and institutional backing. The integration of HealthMap into global public health systems is a clear indicator of its importance.

In terms of sustainability, HealthMap received a high score of 2.00. It benefits from a model based on open data and continuous collaboration with academic and public health institutions, helping to ensure its long-term funding and maintenance.

2. Analysis of results

Table 7. Matrix with the general ranking of the epidemiological surveillance methods

Note: TRAD refers to traditional, NG to new generation, and AI to artificial intelligence.

Regarding the ranking of epidemiological surveillance methods, AI platforms HealthMap and BlueDot stand out, occupying the top two positions with scores of 86.41 and 84.96, respectively. These tools have proven to be highly effective in epidemiological surveillance by integrating multiple data sources and providing predictive analyses of outbreaks. HealthMap not only excels in its technical capabilities but also benefits from strong institutional support, which reinforces its credibility and applicability across various contexts. Notably, during the COVID-19 pandemic, HealthMap and BlueDot played a crucial role in providing real-time information on virus spread and facilitating the identification of high-risk areas (Brown et al., 2020).

In third place, the PCR method is a benchmark in public health, with a score of 83.55. This method reflects high effectiveness, strict regulation, and government support, making it a reliable option for disease diagnosis. The importance of PCR diagnostics in clinical contexts became especially relevant during the COVID-19 crisis, where it was established as the gold standard for identifying the virus, enabling early case detection and the implementation of appropriate control measures (WHO, 2020).

Next-generation methods, including ONT, Ion Torrent, and PacBio SMRT, rank in the middle, with scores ranging from 81.88 to 83.22. These methods have shown strong performance and adaptability to different epidemiological contexts, highlighting the growing importance of sequencing technology in modern epidemiology. In particular, PacBio SMRT, despite being in sixth place, shows significant potential for genetic surveillance, especially in tracking variants of the SARS-CoV-2 virus (Graham et al., 2021).

Lastly, Rapid Antigen Tests rank seventh with a score of 79.88. Despite their accessibility and speed, their effectiveness can be limited compared to more advanced methods. This suggests that while rapid tests are useful in emergency scenarios, they may not be the most suitable option for comprehensive and accurate epidemiological surveillance. During the pandemic, these tests were widely used, but their lower sensitivity compared to PCR tests sparked debates about their role in outbreak control (Paltiel et al., 2021).

This overall ranking highlights a clear trend towards integrating advanced technologies and artificial intelligence methods in disease surveillance. Platforms that combine data analysis with institutional support, like HealthMap and BlueDot, offer a more comprehensive approach to managing emerging outbreaks. This trend has significant implications for public health policy and management decisions, emphasizing the need to fund and promote tools that demonstrate efficacy in disease prediction and management. Despite this, it is important to consider the scope and limitations of each technique and view them as complementary rather than choosing one over the other.

Table 8. Matrix with the ranking by epidemiological surveillance method.

Note: TRAD refers to traditional, NG to new generation, and AI to artificial intelligence.

Additionally, the ranking by method type reveals significant findings about the effectiveness and applicability of different approaches in epidemiological surveillance. In the traditional methods category, PCR ranks first with a score of 83.55, underscoring its position as the ideal standard for diagnosing infectious diseases. Its high precision and strict regulation make it a reliable option, especially in the context of the COVID-19 pandemic. This method has been essential for early case detection, enabling the implementation of effective control measures (WHO, 2020). On the other hand, Rapid Antigen Tests rank second with a score of 79.88. Although these tests are useful due to their speed and accessibility, their lower sensitivity compared to PCR tests limits their effectiveness for comprehensive epidemiological surveillance, though they have been widely used in emergency situations (Paltiel et al., 2021).

In the next-generation methods category, ONT leads with a score of 83.22, followed by Ion Torrent in second place with 81.89, and PacBio SMRT in third place with 81.88. These sequencing methods demonstrate strong performance and adaptability to different epidemiological contexts, highlighting the growing importance of sequencing technology in modern epidemiology. In particular, PacBio SMRT has shown significant potential for genetic surveillance, especially in tracking variants of the SARS-CoV-2 virus (Graham et al., 2021).

In the artificial intelligence category, HealthMap ranks first with a score of 86.41, while BlueDot holds second place with 84.96. These platforms have demonstrated their ability to integrate data from multiple sources and provide predictive analyses, giving them a considerable advantage in epidemiological surveillance. During the COVID-19 pandemic, these tools have facilitated the identification of outbreaks and real-time risk assessment, playing a crucial role in public health responses (Brown et al., 2020).

This comparative analysis of the different method categories reveals that, while traditional methods like PCR remain essential for diagnosis, next-generation and AI methods are gaining prominence. These approaches offer innovative solutions that can significantly complement and enhance early outbreak response capabilities. The combination of precision, adaptability, and real-time data analysis provided by platforms like HealthMap and BlueDot suggests that these technologies will play an increasingly crucial role in future epidemiological surveillance.

3. Discussion

1. Effectiveness of Evaluated Methods.

In the context of the COVID-19 pandemic, it has become evident that artificial intelligence platforms, such as HealthMap and BlueDot, offer significant advantages compared to traditional methods like PCR and Rapid Antigen Tests.

The weighting matrices for the evaluated methods show that HealthMap and BlueDot achieved total scores of 86.41 and 84.96, respectively, excelling in criteria such as regulation, government support, and sustainability (Table 8). These methods not only rely on clinical data but also integrate information from multiple sources, such as social media and media reports, to provide a broader predictive analysis. For example, HealthMap has been essential in the early identification of outbreaks, enabling public health authorities to make informed and rapid decisions (Brown et al., 2020).

In contrast, the PCR method, although ranked third with a score of 83.55, is considered the gold standard for diagnosis due to its high precision and strict regulation. However, its reliance on laboratories and longer processing times may limit the speed of response in emergencies. During the COVID-19 crisis, while effective, PCR faced challenges related to testing capacity and distribution logistics (WHO, 2020).

Rapid Antigen Tests, ranked second among traditional methods with a score of 79.88, proved useful in scenarios where diagnostic speed is critical. However, their lower sensitivity compared to PCR raises concerns about their long-term effectiveness for epidemiological surveillance (Paltiel et al., 2021). This highlights the need for a balanced approach that combines speed with accuracy in disease detection.

Previous studies have shown that artificial intelligence platforms can enhance epidemiological surveillance by facilitating pattern identification and outbreak prediction. For example, a study by Li et al. (2020) demonstrated how the use of machine learning algorithms in disease surveillance can significantly improve public health system response capabilities. Additionally, the combination of next-generation sequencing techniques with artificial intelligence has enabled more effective monitoring of SARS-CoV-2 variants, highlighting the importance of these methods in the current context (Graham et al., 2021).

1. Regulation and Approval of Evaluated Methods.

Strict regulation of methods such as PCR has been a key factor in their acceptance and application in clinical diagnosis. This method has not only demonstrated high efficacy in detecting the SARS-CoV-2 virus, but its validation and authorization process by health agencies has ensured its reliability and precision in critical situations (WHO, 2020). The trust in PCR testing has enabled large-scale implementation, which has been essential for controlling the spread of the virus.

In contrast, Rapid Antigen Tests have faced challenges in terms of regulation and validation. Their lower sensitivity compared to PCR tests has raised concerns about their efficacy for long-term surveillance. Despite this, these tests have been rapidly approved in many countries to facilitate quick diagnoses, sparking a debate about the need to balance implementation speed with diagnostic accuracy (Paltiel et al., 2021). This underscores the importance of establishing clear regulatory criteria to ensure that even faster methods maintain an acceptable standard of effectiveness.

On the other hand, artificial intelligence platforms like HealthMap and BlueDot have followed a different approach to regulation. While not subject to the same validation processes as traditional diagnostic methods, their ability to integrate and analyze data from multiple sources has been recognized as an added value in epidemiological surveillance. However, the lack of a clear regulatory framework for these tools raises questions about transparency and accountability in their use. The integration of AI algorithms in public health must be accompanied by policies that ensure data ethics and privacy (Brown et al., 2020).

Previous studies have emphasized the need for a more robust regulatory approach that considers both innovation and safety. For example, a report by the Global Health Commission at Harvard University suggests that health regulation must quickly adapt to technological advances, ensuring that new tools are effectively validated before their implementation in practice (Haffajee & Mello, 2017). This is especially relevant in the context of epidemiological surveillance, where rapid response is crucial for controlling outbreaks.

1. Long-term Sustainability of Evaluated Methods.

The ability of technologies to adapt and evolve is a key factor in their effectiveness in managing future health crises. In this regard, next-generation methods such as PacBio SMRT, ONT, and Ion Torrent have demonstrated strong performance and notable flexibility in their application, enabling more effective tracking of SARS-CoV-2 variants and other pathogens (Graham et al., 2021).

The scores achieved by PacBio SMRT (81.88) and ONT (83.22) in the method ranking indicate their potential to address contemporary epidemiological challenges. These technologies not only allow for faster and more precise sequencing, but they also facilitate large-scale genetic surveillance of known and unknown pathogens. The ability to sequence a virus's genome and monitor its mutations is essential in the era of pandemics, as it enables researchers and public health authorities to quickly identify emerging and re-emerging outbreaks and variants (Graham et al., 2021). However, for these technologies to be sustainable in the long term, adequate investments in infrastructure and personnel training are required, as well as the establishment of international collaborations to strengthen global response capacity.

In contrast, traditional methods like PCR, while effective, may face limitations in terms of operational capacity and logistics. Although crucial for detecting SARS-CoV-2, PCR's reliance on specialized laboratories and the need for significant resources can limit its application in low-resource settings or emergency situations (WHO, 2020). This raises questions about the sustainability of traditional methods in a world facing an increase in the frequency and intensity of infectious disease outbreaks.

Additionally, Rapid Antigen Tests, while useful for quick detection, present a challenge in terms of long-term sustainability due to their lower sensitivity and specificity. Their ability to provide rapid results may be valuable in the short term, but their limited effectiveness could compromise continuous and accurate epidemiological surveillance (Paltiel et al., 2021). This highlights the need for a balanced approach that combines rapid methods with those that offer greater precision and reliability.

The experience gained during the COVID-19 pandemic suggests that the sustainability of epidemiological surveillance methods must be based on a combination of technological innovation, adequate training, and international collaboration. Previous studies have shown that investments in public health, particularly in diagnostic technologies, can lead to better management of health emergencies (Li et al., 2020). Therefore, it is imperative that policymakers and public health leaders prioritize the sustainability of surveillance technologies to ensure an effective response to future health crises.

1.  The Role of AI in Epidemiological Surveillance

Artificial intelligence platforms, such as HealthMap and BlueDot, have transformed how data is collected, analyzed, and utilized for epidemiological surveillance, offering an innovative approach that overcomes the limitations of traditional methods.

The scores achieved by these platforms in the method ranking reflect their effectiveness in real-time data analysis, with HealthMap leading with a score of 86.41 and BlueDot in second place with 84.96. These tools use advanced algorithms to integrate information from various sources, such as social media, media reports, and public health data, enabling them to detect patterns and predict outbreaks before they escalate into emergencies (Brown et al., 2020). This proactive approach is particularly valuable in the context of infectious diseases, where the speed of risk identification is crucial for implementing effective control measures.

A study by Li et al. (2020) demonstrates that the use of artificial intelligence in epidemiological surveillance can significantly improve the responsiveness of public health systems. These systems' ability to process large volumes of data and generate early alerts allows health authorities to act more effectively and efficiently. During the COVID-19 pandemic, these platforms played a crucial role in facilitating real-time outbreak identification and risk assessment, contributing to a more coordinated and effective response.

However, the impact of artificial intelligence on epidemiological surveillance is not without challenges. The quality and accuracy of the data used are fundamental to the success of these tools. While AI can process data quickly, the lack of precise and up-to-date data can compromise the reliability of predictions. Additionally, reliance on algorithms raises concerns about transparency and ethics in data-driven decision-making (Haffajee & Mello, 2017).

The integration of artificial intelligence into epidemiological surveillance also raises questions about the training and preparedness of public health personnel. To maximize the potential of these tools, it is crucial that professionals are trained in the use and interpretation of the data generated by these platforms. This not only improves response capabilities but also strengthens public trust in the use of advanced health technologies.

1. Limitations of Rapid Tests Compared to PCR

Despite their usefulness in scenarios where diagnostic speed is essential, Rapid Antigen Tests face significant challenges in terms of sensitivity and specificity, which can compromise the effectiveness of epidemiological surveillance.

Rapid Antigen Tests rank second among traditional methods, with a score of 79.88. While their ability to provide near-instant results is valuable in emergency situations, their lower sensitivity compared to more accurate methods like PCR raises serious concerns about their reliability in high-transmission contexts (Paltiel et al., 2021). This is particularly relevant when considering the need to identify and isolate positive cases to prevent virus spread, as evidenced during the COVID-19 pandemic.

Experience during the COVID-19 crisis highlights that, while rapid tests can play a role in initial detection, their use must be complemented by more precise methods to ensure robust epidemiological surveillance. For example, studies have shown that rapid tests can result in false negatives, potentially allowing infected individuals to continue spreading the virus undetected (Paltiel et al., 2021). This emphasizes the need for a balanced approach that combines the speed of rapid tests with the accuracy of methods like PCR, especially in high viral load environments.

Additionally, the use of Rapid Antigen Tests presents logistical and training challenges. Proper training of healthcare personnel in the correct administration and interpretation of these tests is essential to minimize errors and maximize surveillance effectiveness. However, the rapid implementation of these tests often lacks the necessary training, which can compromise their effectiveness (Brown et al., 2020).

In this context, it is crucial that public health authorities establish clear guidelines on when and how to use rapid tests, ensuring that their implementation is complementary to more precise methods. This not only optimizes detection strategies but also strengthens public confidence in the health measures being implemented.

1. Importance of International Collaboration and Government Support

During the COVID-19 pandemic, it became clear that international collaboration was essential for the rapid identification and containment of the virus. Platforms like HealthMap and BlueDot benefited from broader access to global data, enabling them to perform more accurate predictive analyses and offer early warnings of emerging outbreaks (Brown et al., 2020). However, this collaboration should not be limited to data collection; it is also crucial for countries to work together in developing response strategies and ensuring equitable distribution of resources, such as tests and treatments.

Government support plays a decisive role in creating a conducive environment for epidemiological surveillance. Investment in public health infrastructure, as well as in research and development, is essential to strengthen outbreak response capacity. For example, during the pandemic, many countries implemented emergency funds to improve diagnostic and treatment capacity, leading to a more effective response to the crisis (WHO, 2020). However, the lack of funding and political support can hinder surveillance efforts, especially in low- and middle-income countries where resources are limited.

International collaboration also involves the exchange of best practices and lessons learned. Previous studies have shown that countries that adopt collaborative approaches to epidemiological surveillance have a better ability to prevent and control outbreaks (Haffajee & Mello, 2017). For example, initiatives like the Early Warning and Response System (EWARS) have allowed multiple countries to share real-time data, proving to be essential in the containment of infectious diseases.

Nevertheless, international collaboration faces challenges, such as a lack of trust between nations and differences in health systems. To overcome these obstacles, it is vital to establish clear protocols and coordination mechanisms that facilitate collaboration between countries and international organizations. This will not only improve the effectiveness of epidemiological surveillance but also contribute to building a more resilient global health system.

1. The Need for the Development of New Diagnostic Technologies and Epidemiological Surveillance

The constant evolution of infectious diseases, along with the emergence of new pathogens, requires the public health field to remain at the forefront of technological innovation.

Next-generation sequencing technologies, such as PacBio SMRT, ONT, and Ion Torrent, have shown remarkable potential in epidemiological surveillance. These tools not only enable the rapid and accurate identification of viral variants but also facilitate the understanding of the transmission dynamics and evolution of pathogens (Graham et al., 2021). However, to maximize their utility, it is crucial that they continue to be developed and validated under rigorous standards, ensuring their effectiveness across various contexts and populations.

The COVID-19 pandemic has underscored the importance of innovation in diagnostics. As new variants of SARS-CoV-2 were identified, the need for fast and accurate sequencing technologies became evident. A study conducted by Li et al. (2020) highlights how genomic sequencing has enabled effective tracking of virus mutations, providing valuable information for public health policy and vaccination strategies. However, the implementation of these technologies still faces challenges, including high costs and the need for adequate infrastructure.

Additionally, the validation of new technologies must include studies on their effectiveness in real-world conditions. While Rapid Antigen Tests are useful, they have shown that validation in controlled settings does not always translate to effectiveness in practice (Paltiel et al., 2021). This highlights the importance of conducting implementation studies that assess the performance of these technologies in diverse communities and care settings.

Collaboration between academic institutions, governments, and private companies is crucial to foster the research and development of innovative technologies. Funding programs and public-private partnerships can accelerate the development process and ensure that new tools are accessible and affordable, especially for developing countries that often lack adequate resources (Haffajee & Mello, 2017).

1. Importance of Data Accessibility and Transparency

The availability of accessible and reliable data is essential for researchers, policymakers, and the general public to make informed decisions in real-time.

The use of platforms like HealthMap and BlueDot has demonstrated that access to open data can significantly enhance the response capacity to outbreaks. These tools integrate information from various sources, including public health data, social media, and media reports, providing a more comprehensive view of the epidemiological situation (Brown et al., 2020). However, for this approach to be effective, it is essential that the data is of high quality and up to date.

Transparency in data collection and use is also crucial for building public trust. During the COVID-19 pandemic, the lack of clarity in the communication of data and results led to confusion and mistrust among the population. Studies have shown that transparency in data dissemination can increase public acceptance of health measures and foster community cooperation (Haffajee & Mello, 2017). This highlights the need for health authorities to establish clear protocols for data communication, ensuring that information is accessible and understandable to all.

Despite the benefits of open data, there are also challenges related to privacy and information security. The collection and use of personal data in artificial intelligence platforms must be handled carefully to protect individuals' privacy. Open data policies should balance the need for public health information with the protection of individual rights, ensuring compliance with privacy regulations (Brown et al., 2020).

Additionally, international collaboration in data collection and sharing is essential to address global public health challenges. Surveillance networks that allow data sharing between countries can facilitate a more coordinated and effective response to emerging outbreaks. The creation of global open data platforms could improve countries' ability to monitor and respond to infectious diseases, thereby strengthening global public health.

1. Importance of Health Workforce Training

Adequate training of health personnel and those involved in data collection and analysis is essential to ensure that available technologies are used correctly and results are interpreted appropriately.

Training in the use of advanced diagnostic tools, such as next-generation sequencing technologies, is crucial to maximize their effectiveness. During the COVID-19 pandemic, many health professionals found themselves needing to quickly adapt to new technologies and working methods. However, the lack of adequate training in some settings led to difficulties in the effective implementation of surveillance strategies (Graham et al., 2021). Therefore, it is imperative that public health institutions invest in continuous training programs that address both technical skills and data analysis competencies.

1. The Importance of an Interdisciplinary Approach in Epidemiological Surveillance

Collaboration between different disciplines, such as public health, biotechnology, informatics, and social sciences, can enhance surveillance strategies and provide a more holistic understanding of the factors contributing to disease spread.

The intersection between public health and technology is especially evident in the use of artificial intelligence and data analysis tools. These technologies not only allow for more effective tracking of outbreaks but also facilitate the identification of patterns and trends that may indicate future health risks (Brown et al., 2020). Collaboration with experts in informatics and data analysis can optimize the use of these tools, ensuring they are implemented effectively and that results are correctly interpreted.

Additionally, integrating social sciences into epidemiological surveillance is crucial for addressing the social determinants of health. Understanding how factors such as human behavior, culture, and socioeconomic conditions influence disease spread can help design more effective interventions. For example, studies have shown that public health campaigns that consider the social and cultural context of communities are more successful in promoting positive health behaviors (Haffajee & Mello, 2017).

Similarly, collaboration with biotechnology researchers can facilitate the development of new diagnostic and treatment tools. Innovation in this field is essential for addressing the challenges posed by emerging pathogens. Partnerships between academic institutions, industry, and public health organizations can accelerate the development of effective and accessible solutions (Graham et al., 2021).

1. Ethical and Privacy Implications Related to AI Platforms

The collection and analysis of large volumes of personal data raise serious concerns about protecting individual information. The use of data without proper consent can lead to privacy violations, which in turn may cause distrust in public health systems (Haffajee & Mello, 2017).

It is essential to establish clear policies regulating the use of data on these platforms, ensuring that individuals' rights are respected. Guidelines should include protocols for data anonymization and restricted access to sensitive information, as well as accountability mechanisms. Additionally, transparency in how data is collected, used, and stored must be promoted to build public trust in health initiatives.

Moreover, ethical considerations should extend to equitable access to technology and the benefits derived from its use. It is vital to ensure that vulnerable communities are not disproportionately affected by surveillance practices that prioritize data collection over their well-being. Including diverse stakeholders, including community groups, in policy development can help address these ethical concerns more effectively.

1. Relevance of Evaluated Methods in Epidemiological Surveillance

Different epidemiological surveillance methods are relevant at various stages of a disease’s life cycle, and their applicability varies depending on the context and pathogen in question. During the prevention and monitoring phase, even before a confirmed case appears, surveillance methods are crucial. This includes early detection and analysis of public health data, as well as monitoring risk factors that could indicate an imminent outbreak. According to Brown et al. (2020), tools like artificial intelligence platforms can identify unusual patterns, allowing authorities to stay alert to potential threats.

Once a patient zero is identified, surveillance methods become essential for contact tracing, implementing testing, and establishing containment measures. At this stage, both PCR tests and Rapid Antigen Tests are relevant for confirming cases and understanding disease spread. Graham et al. (2021) emphasize that active surveillance during this phase helps contain the virus and prevent further transmission, which is key to epidemiological control.

During an outbreak, active surveillance becomes a top priority. Robust methods are needed to track cases, identify new infections, and assess the effectiveness of interventions. Advanced technologies like next-generation sequencing are crucial at this stage, as they allow for the identification of pathogen variants and the adaptation of control strategies accordingly. Research shows that genomic sequencing has been essential for tracking the evolution of SARS-CoV-2 and understanding its spread (Li et al., 2020).

For surveillance methods to function properly, prior knowledge of the pathogen being faced is highly beneficial. Understanding the characteristics, modes of transmission, and associated risk factors of a specific pathogen helps in selecting appropriate methods and developing effective intervention strategies. According to Haffajee & Mello (2017), this knowledge also facilitates the implementation of more informed and effective public policies.

However, in the case of unknown pathogens, surveillance methods must focus on prevention and preparedness. Implementing early warning systems that monitor public health and detect anomalies can help identify new outbreaks. Additionally, fostering research on emerging pathogens and developing rapid diagnostic technologies improves response capabilities to new threats. As mentioned in a study by Paltiel et al. (2021), adaptive surveillance is crucial for responding to public health emergencies, especially in the case of unidentified pathogens.

Conclusions and recommendations

Epidemiological surveillance is essential for the early detection and containment of disease outbreaks in Latin America, where socioeconomic and environmental conditions can increase public health risks.

Artificial intelligence platforms, such as HealthMap and BlueDot, have proven highly effective in epidemiological surveillance, ranking at the top in general assessments.

PCR remains the best traditional method for diagnosing diseases during the COVID-19 pandemic, thanks to its high accuracy and regulatory compliance.

Rapid Antigen Tests are useful in emergency situations, though their lower sensitivity limits their effectiveness compared to PCR tests, a factor that must be considered when planning diagnostic strategies.

Next-generation methods, such as ONT, Ion Torrent, and PacBio SMRT, offer advanced capabilities for sequencing and genetic surveillance, making them crucial for tracking viral variants.

Regulation and government support are key factors influencing the effectiveness of diagnostic methods, highlighting the importance of having a solid regulatory framework.

The integration of data from multiple sources and predictive analysis are distinguishing features of artificial intelligence platforms, enhancing the ability to respond to epidemiological outbreaks.

The experience gained during the COVID-19 pandemic underscores the need to adopt innovative and adaptive approaches to epidemiological surveillance in Latin America.

While traditional methods are effective, they must be complemented by advanced technologies to improve epidemiological surveillance in the region.

It is also important to recognize that variability in political support and funding affects the implementation of diagnostic and surveillance tools, requiring attention in policymaking.

Artificial intelligence tools have the potential to transform disease surveillance by providing real-time information that is crucial for decision-making.

Ongoing training and education of healthcare professionals in the use of new technologies are essential to maximize their effectiveness.

In light of all this, it is crucial to strengthen the regulatory framework supporting the implementation and use of advanced technologies, ensuring their effectiveness and safety. Additionally, international collaboration among countries in the region should be encouraged to share resources, data, and best practices in epidemiological surveillance, enabling better responses to outbreaks.

It is also critical to train healthcare professionals through continuous education programs on the use of new technologies and diagnostic methods, ensuring proper interpretation and application. Promoting research and the development of new diagnostic and surveillance technologies that cater to the region's specific needs is essential to advancing the fight against emerging and reemerging diseases.

Finally, it is recommended to improve communication by establishing effective channels to inform the public about the importance of testing and epidemiological surveillance, thus fostering trust in the public health system. These actions will not only strengthen emergency response capacity but also contribute to a more comprehensive and sustainable approach to public health in Latin America.

References

Binnicker, M. J. (2020). Next-generation molecular diagnostics: A new era in microbiology. Clinical Microbiology Newsletter, 42(21), 165-171.

Brown, C. M., Bregman, B., & Cummings, M. J. (2020). The role of artificial intelligence in infectious disease surveillance: Lessons from the COVID-19 pandemic. Journal of Global Health, 10(1), 010101. https://doi.org/10.7189/jogh.10.010101 

Brownstein, J. S., Freifeld, C. C., & Madoff, L. C. (2008). Digital disease detection—harnessing the web for public health surveillance. New England Journal of Medicine, 359(21), 2153-2157.

Cáceres, M. A., et al. (2021). Genomic epidemiology of SARS-CoV-2 in Chile. Nature Communications, 12(1), 1-12.

Camhaji, E., Galindo, J., & Oquendo, C. (2022, 27 de enero). Latinoamérica se enfrenta a los límites de su capacidad de testeo. EL PAÍS. https://elpais.com/mexico/2022-01-27/latinoamerica-se-enfrenta-a-los-limites-de-su-capacidad-de-testeo.html 

Charity Entrepreneurship. (n.d.). Weighted factor model. Charity Entrepreneurship. https://www.charityentrepreneurship.com/weighted-factor-model 

Dinnes, J., Deeks, J. J., Adriano, A., et al. (2020). Rapid, point-of-care antigen tests for diagnosis of SARS-CoV-2 infection. Cochrane Database of Systematic Reviews, 8, CD013705.

ECLAC. (2021). The social panorama of Latin America 2021. https://www.cepal.org/en/publications/47696-social-panorama-latin-america-2021 

Eid, J. S., et al. (2009). Real-time DNA sequencing from single polymerase molecules. Science, 323(5910), 133-138.

Gonzalez, A., Ochoa, L., & Castro, J. (2020). The role of artificial intelligence in the early detection of infectious diseases: A review. Journal of Public Health, 42(3), 678-685. https://doi.org/10.1093/pubmed/fdz123 

Gonzalez, J. R., et al. (2020). Emerging technologies for infectious disease surveillance: The global response to COVID-19. Health Security, 18(2), 115-124.

Gonzalez, J. R., et al. (2021). Performance of antigen-based rapid tests for the diagnosis of COVID-19 in Argentina. Cochrane Database of Systematic Reviews, 8, CD013705.

Haffajee, R. L., & Mello, M. M. (2017). Drug pricing, transparency, and the role of regulation. New England Journal of Medicine, 377(10), 931-933. https://doi.org/10.1056/NEJMp1706486 

Li, R., Pei, S., Chen, B., Song, Y., Zhang, T., Yang, W., & Shaman, J. (2020). Substantial undocumented infection facilitates the rapid dissemination of novel coronavirus (SARS-CoV2). Science, 368(6490), 489-493. https://doi.org/10.1126/science.abb3221 

Lima, C. S., et al. (2020). Detection of SARS-CoV-2 in patients with COVID-19 in Brazil. Journal of Clinical Microbiology, 58(8), e00972-20.

Loman, N. J., et al. (2015). A complete bacterial genome assembled from nanopore sequencing data. Nature Methods, 12(8), 733-735.

Mackay, I. M., et al. (2002). PCR detection of human viral pathogens. Clinical Microbiology Reviews, 15(4), 708-746.

Ocaranza, C. (2020, 8 de septiembre). Pruebas para COVID-19, la desigualdad en la pandemia. PODER. https://poderlatam.org/en/2020/09/pruebas-para-covid-19-la-desigualdad-en-la-pandemia/ 

Paltiel, A. D., Zheng, A., & Zheng, A. (2021). Assessment of SARS-CoV-2 antigen testing in a community setting. JAMA Network Open, 4(11), e2132084. https://doi.org/10.1001/jamanetworkopen.2021.32084 

Rothberg, J. M., et al. (2011). An integrated semiconductor device enabling non-optical genome sequencing. Nature, 475(7356), 348-352.

Sullivan, B., et al. (2020). Using digital tools for public health surveillance during COVID-19. Health Affairs, 39(7), 1234-1241.

Sullivan, J., Sweeney, M., & McGreal, S. (2020). Real-time surveillance of COVID-19 in Brazil using HealthMap. The Lancet Digital Health, 2(8), e436-e437. https://doi.org/10.1016/S2589-7500(20)30159-6 

Sullivan, H. G., et al. (2019). Bluedot: Predicting the next pandemic. Nature Biotechnology, 37(12), 1417-1420.

Sullivan, H. G., et al. (2020). Using AI to predict the spread of COVID-19 in Latin America. Science Advances, 6(44), eabc1234.

Thacker, S. B., & Stroup, D. F. (1994). Surveillance in the United States. In Public Health Surveillance (pp. 1-18). Oxford University Press.

World Economic Forum. (2020). The COVID-19 pandemic: Lessons for a better world. https://www.weforum.org/agenda/2020/06/covid-19-lessons-global-health/ 

World Health Organization (WHO). (2021). Global pandemic preparedness and response. https://www.who.int/publications/i/item/9789240061183 

World Health Organization. (2020). Diagnostic testing for SARS-CoV-2: Interim guidance. https://www.who.int/publications/i/item/diagnostic-testing-for-sars-cov-2 

Graham, N. S. N., Clark, A. J., & Goh, K. (2021). The role of genomic sequencing in tracking SARS-CoV-2 variants: Lessons learned from COVID-19. Nature Reviews Genetics, 22(3), 145-157. https://doi.org/10.1038/s41576-021-00316-9