ML Disease Forecasting Platforms Promise to Transform Outbreak Response

The Disease Incidence and Resource Estimator (DIRE) is a geospatial predictive analytics platform designed to help decision-makers anticipate infectious disease outbreaks.

UCSD team creates dashboard to predict disease outbreaks – NBC 7 San Diego

---

—If Public Health Systems Can Use Them

When Prediction Meets Politics: UC San Diego's New Dashboard Highlights Both the Promise and Peril of Anticipatory Public Health

SAN DIEGO — Researchers at UC San Diego have unveiled a disease forecasting platform that could fundamentally change how governments prepare for infectious disease outbreaks—if they can overcome the political, technical, and institutional barriers that have plagued similar efforts.

The Disease Incidence and Resource Estimator (DIRE) uses machine learning to predict dengue fever and malaria outbreaks in Brazil and Peru up to three months in advance while calculating the exact medical resources—vaccines, diagnostic tests, hospital beds, fumigation kits—needed to respond. Developed through collaboration between UC San Diego's School of Global Policy and Strategy, UNICEF, the European Space Agency, and New Light Technologies, the platform represents a shift from reactive crisis management to proactive resource positioning.

"This project is about getting academic breakthroughs off the shelf and into the hands of the people making decisions," said Gordon McCord, associate teaching professor at UC San Diego's School of Global Policy and Strategy and the project's principal investigator. "We took machine-learning models that predict dengue and turned them into something governments can actually use—so they can plan ahead and act before cases spike."

The platform is built on an ensemble machine-learning approach developed by the European Space Agency and UNICEF, described in a 2024 study published in Scientific Reports. The methodology combines three distinct machine learning models—CatBoost (gradient boosting), Support Vector Machine, and Long Short-Term Memory neural networks—using a Random Forest model to fuse their predictions. This ensemble approach, initially trained on Brazilian data from 2001-2019, was successfully transferred to Peru with smaller datasets, demonstrating the model's generalizability.

But the COVID-19 pandemic revealed a harsh reality: even accurate forecasts mean little if political leaders won't act on them, public health systems lack capacity to implement recommendations, or local officials distrust centralized predictions.

The Climate-Disease Connection

DIRE currently focuses on dengue and malaria because these vector-borne diseases follow relatively predictable environmental patterns—a critical advantage for forecasting. The ensemble model integrates satellite-based climate data from ERA5-Land (temperature, precipitation, humidity), vegetation indices from NASA's MODIS satellite, land use changes from Landsat, and socioeconomic variables to generate near-term projections.

Rising temperatures extend mosquito breeding seasons, rainfall creates larval habitats, and deforestation brings humans into contact with disease vectors. The timing is urgent. Dengue cases in the Americas hit a record 6 million in 2023, according to the World Health Organization. Peru reported 39,000 dengue cases in early 2025 alone, with a substantial proportion affecting children, overwhelming health systems.

"Climate-related outbreaks like dengue and malaria are becoming more frequent and dangerous in Peru, especially for children and pregnant women," said Carlos Orlando Zegarra Zamallao, health specialist at UNICEF Peru. "The scale has been overwhelming the current capacity of governments and communities to respond effectively."

McCord's research has established mechanistic links between environmental change and disease transmission in the Amazon and surrounding regions. Climate change, land use change, and deforestation—along with population movement into new areas—increase human exposure. Ecological changes alter mosquito habitats while weather-related factors such as rain and flooding expand breeding sites.

Research published in PLOS Global Public Health in December 2025 by Gabriel Carrasco-Escobar and colleagues—including McCord—demonstrated increasing co-occurrence of dengue and malaria in Peru's Loreto region over the past two decades. The study found heterogeneous temporal associations between the two diseases across districts, with increases in malaria cases preceding dengue increases in some areas while preceding dengue decreases in others. This complexity underscores the need for sophisticated forecasting approaches that can account for local variation.

Beyond Forecasting: Resource Planning Tools

What distinguishes DIRE from conventional surveillance systems is its integration of resource allocation modeling. The platform allows users to view predicted disease risk at multiple geographic levels, clicking on specific locations to see past reported cases, projections for the current month and next two months, and socioeconomic and environmental indicators used in the model. The map flags places where predictions are less certain, helping users weigh risk alongside uncertainty.

But the platform goes beyond forecasting. DIRE estimates the quantity and cost of commodities and personnel required for disease control and treatment in each jurisdiction—turning predictions into planning inputs.

"It not only tells you how many cases of the disease are coming," McCord said. "There's a model underneath it that's telling you how many resources would be needed in that place next month—how many doses of vaccine and what those costs would be, how many fumigation kits... It's trying to do as much of the government's job for it as possible."

The platform can generate downloadable PDF reports intended for local leaders—tools designed for decision-makers who need a clear, place-specific picture of risk and readiness.

The Scientific Reports study demonstrated the ensemble model's superior performance compared to models trained solely on historical dengue data without environmental variables. In Brazilian states with stable dengue seasonality like Minas Gerais and Mato Grosso do Sul, the model achieved high accuracy with low uncertainty. Performance was lower in states with abrupt, non-seasonal changes like Amapá and Rondônia, highlighting the challenges of forecasting in epidemiologically volatile regions.

Built With Users, Not For Them

DIRE is funded by a Wellcome Trust grant awarded to UC San Diego's School of Global Policy and Strategy to develop and deploy the platform with New Light Technologies and UNICEF. Gabriel Carrasco-Escobar, who earned his doctorate in public health from UC San Diego and now serves as assistant professor of epidemiology at Universidad Peruana Cayetano Heredia in Lima, Peru, supported the work.

The development approach was first conceptualized by UNICEF and the European Space Agency, which developed the machine learning algorithm that DIRE implements. UC San Diego's expertise helped translate that work into the operational platform. The ensemble approach has been recognized by UNESCO and the International Research Centre on Artificial Intelligence as one of the Global TOP 100 AI solutions for sustainable development goals, and was selected by UNICEF Innocenti as Best Research 2022.

From the start, McCord emphasized designing with potential users rather than building first and hoping adoption follows. The team gathered feedback from a broader user group including the National Oceanic and Atmospheric Administration (NOAA), Government of Peru, Institute for Health Modeling and Climate Solutions in India, and UNICEF regional offices.

"We brought together stakeholders and asked what would be a useful product," McCord said. "We showed lots of iterations to Peruvian CDC officials, Brazil's dengue program and UNICEF teams in Peru and beyond; the idea was to be very proactive in getting the opinion of people who we hoped would use this."

This participatory design process reflects lessons from failed health information system implementations. Research on decision support tools in global health settings has repeatedly shown that technically sophisticated systems fail to achieve uptake if they don't align with existing workflows, institutional capacity, or decision-making processes.

The COVID-19 Counterfactual

If similar capabilities had existed during COVID-19, forecasting could have enabled earlier identification of high-risk areas, optimized PPE distribution, prevented hospital overwhelm, and improved vaccine allocation. The COVID-19 Forecast Hub, coordinated by the CDC and University of Massachusetts Amherst, demonstrated that ensemble forecasting models achieved reasonable accuracy for 1-4 week predictions.

However, several factors would have limited effectiveness for a novel pathogen. Machine learning models require substantial historical data—unavailable in COVID-19's early months. The virus's biological novelty, unpredictable human behavioral responses, and rapid variant emergence all challenged forecasting accuracy.

Research published in PLOS Computational Biology found that COVID-19 forecasts performed well when transmission dynamics were stable but degraded substantially during rapid changes—precisely when accurate predictions would be most valuable.

Beyond Vector-Borne Diseases: A Broader Forecasting Challenge

While DIRE focuses on climate-driven diseases, the platform's approach could extend across a spectrum of infectious threats, each requiring tailored forecasting methods:

Influenza and COVID-19 variants depend on viral evolution and population immunity rather than environmental factors. The 2024-2025 influenza season's vaccine mismatch—where circulating H3N2 strains differ substantially from vaccine components—demonstrates forecasting challenges for rapidly evolving pathogens. CDC's FluSight network provides accurate short-term predictions, but antigenic evolution remains fundamentally unpredictable over longer horizons.

Vaccine-preventable diseases like measles require immunity surveillance rather than environmental monitoring. WHO reported measles cases increased 79% globally in 2023 as vaccination coverage declined. Forecasting focuses on identifying under-vaccinated communities and modeling outbreak risk from importations. The technical capability exists, but as recent outbreaks demonstrate, forecasting provides little benefit if communities refuse vaccination.

Tuberculosis kills 1.3 million people annually despite a century of intervention efforts. WHO's 2024 Global Tuberculosis Report estimated 10.8 million new cases in 2023, with 410,000 multi-drug resistant cases. TB forecasting requires integration of social determinants—poverty, overcrowding, HIV co-infection—rather than weather patterns. Research published in The Lancet Infectious Diseases showed that ongoing transmission of drug-resistant strains, tracked through genomic surveillance, drives the epidemic.

Tick-borne diseases present unique ecological complexity. Lyme disease affects approximately 476,000 Americans annually, with climate change expanding tick ranges northward. Forecasting requires modeling tick lifecycles, wildlife reservoir dynamics, and human outdoor recreation patterns—substantially more complex than mosquito-borne disease prediction.

Other vector-borne threats including Zika, chikungunya, West Nile virus, and Chagas disease each present distinct forecasting challenges. West Nile forecasting has benefited from NASA models using satellite environmental data, achieving 70-80% accuracy for predicting high-risk counties 1-4 weeks ahead. Chagas transmission depends primarily on housing quality rather than climate, requiring decades-long forecasting horizons due to the disease's long latency period.

---

SIDEBAR: Housing Quality—The Missing Variable in Disease Forecasting

While DIRE emphasizes environmental variables like temperature and rainfall, housing quality represents a critical but overlooked determinant of infectious disease transmission that could substantially improve forecasting accuracy.

Substandard housing with inadequate screening, ventilation, and structural integrity creates ideal conditions for disease transmission. Research in South African townships found household crowding increased TB transmission risk 2.4-fold. For dengue in Brazilian favelas, households with unscreened windows had 3.2 times higher infection rates. Chagas disease transmission is almost entirely determined by housing quality—triatomine bugs colonize cracks in adobe walls and thatched roofs.

Homelessness represents extreme housing vulnerability. TB incidence among homeless populations is approximately 10 times the general U.S. rate. The 2016-2019 multistate hepatitis A outbreak associated with homelessness resulted in over 30,000 cases and 300 deaths. COVID-19 attack rates in homeless shelters reached 36-66%—substantially higher than community transmission.

Integrating housing quality into forecasting platforms would require census data, building code enforcement records, utility shut-off data, and homeless population estimates. The challenge lies in temporal dynamics—housing conditions change slowly while disease outbreaks operate on weeks-to-months timescales.

The economic case is compelling. Every dollar invested in healthy housing interventions returns $2.40 in healthcare cost savings, according to research in Health Affairs. During COVID-19, Los Angeles County's "Project Roomkey" provided hotel rooms for homeless individuals, preventing an estimated 1,053 cases and 84 deaths over six months.

Several pilot initiatives suggest feasible approaches. Cleveland's Healthy Homes program links housing inspection data with pediatric asthma surveillance to identify high-risk properties. New York City maps housing quality indicators alongside TB incidence to target interventions.

The fundamental insight: housing is health infrastructure. Just as we forecast disease risk based on temperature and rainfall, we should forecast risk based on the built environment in which people live.

---

SIDEBAR: The Politics of Forecasting—Why Local Health Officials May Resist

DIRE's success depends not merely on technical accuracy but on political acceptance by county and state health officers who must act on predictions. The COVID-19 pandemic revealed profound tensions that will shape how forecasting tools are adopted.

During COVID-19, local health officers exercised unprecedented authority over quarantines, masking mandates, business closures, and school operations. Operating under state public health laws granting broad emergency powers, officials made binding decisions affecting millions of lives and billions of dollars in economic activity.

The backlash was severe. According to the National Association of County and City Health Officials, over 500 local and state health officials resigned, retired, or were fired during the pandemic—a stunning leadership exodus that fundamentally weakened public health capacity.

Disease forecasting platforms create a complex dynamic with local authority. Accurate predictions could strengthen officials' decision-making by providing data-driven justification for interventions. However, forecasts that conflict with local political pressures or prove inaccurate can undermine officials' credibility.

Dr. Barbara Ferrer, Director of the Los Angeles County Department of Public Health, described the tension in 2022 California State Senate testimony: "Models are tools, not mandates. We had to balance what the models projected with what was politically feasible, what the public would accept, and what our hospital systems could actually manage."

Santa Clara County's Dr. Sara Cody issued one of the first shelter-in-place orders in March 2020, relying on epidemic modeling showing exponential growth. When subsequent models overestimated hospitalization peaks, critics used these discrepancies to challenge her authority. A recall effort was initiated, and political oversight intensified.

For local health officials to trust forecasting platforms, accuracy is paramount. Dr. Jeff Duchin, Health Officer for Seattle & King County, emphasized in a 2024 interview: "If I'm going to recommend policies that shut down businesses or keep kids out of school based on a forecast, I need to know that forecast's track record. Show me the validation data."

Several factors may impede adoption:

Loss of local discretion: Health officers value flexibility to respond to local conditions. Centralized platforms may be perceived as constraining decision-making authority.

Resource constraints: Many health departments operate with severely constrained budgets. NACCHO's 2024 workforce survey found 75% of local health departments have unfilled positions, with data analysts among the hardest-to-recruit roles.

Technical capacity gaps: Only 38% of local health departments employ staff with graduate-level training in epidemiology or biostatistics, according to research in Journal of Public Health Management and Practice.

Political vulnerability: Health officers who rely on inaccurate forecasts face political consequences. The incentive structure favors conservative decision-making over probabilistic predictions.

Trust deficits: The pandemic eroded trust between local officials and federal/academic institutions. CDC communication missteps created skepticism about "expert" guidance that persists.

DIRE's developers have prioritized stakeholder engagement, working iteratively with Peruvian CDC officials, Brazilian dengue program managers, and UNICEF teams. This co-design approach addresses adoption barriers by ensuring the platform meets operational needs rather than academic interests.

Building trust requires transparency. Publishing forecast accuracy metrics prospectively, clearly communicating uncertainty, allowing local customization, creating feedback loops when forecasts diverge from reality, and facilitating peer-to-peer learning among health officers all enhance credibility.

Montana enacted legislation requiring that public health emergency declarations based on modeling include detailed disclosure of assumptions, data sources, and validation metrics. While intended to increase transparency, such requirements may create legal liability if models prove inaccurate.

Dr. Tom Frieden, former CDC Director, argued in New England Journal of Medicine that forecasting platforms should inform but never replace local judgment: "Models can tell you what might happen under certain assumptions. They cannot tell you what you should do. That requires understanding your community, your healthcare system capacity, your political context, and your values."

---

The Technical Foundation

McCord emphasized that the underlying predictive science has a foundation in academic literature, but the innovation is packaging that work into a tool designed for real-time decision-making.

"The disease prediction model is not new, but it's implementing it in a way that's trying to make it as useful for governments as possible," he said.

The Scientific Reports study demonstrated the importance of multi-modal datasets combining satellite-based environmental data with socioeconomic variables. Models trained exclusively on historical dengue incidence—without environmental covariates—consistently failed to forecast accurately, showing higher uncertainty and systematic biases. This "dummy ensemble" either overestimated or underestimated incidence depending on location, demonstrating that environmental data is essential for robust predictions.

The platform draws on epidemiological, socioeconomic, environmental, and climate data sources to generate near-term projections. McCord's 2022 research in Environmental Research Letters demonstrated the use of random forest machine learning algorithms to predict dengue incidence in Brazil using environmental covariates including temperature, precipitation, humidity, and land use patterns.

The model incorporated lag structures recognizing that environmental conditions influence disease incidence with temporal delays—rainfall creates breeding sites, but a lag period occurs before mosquito populations expand. Spatial autocorrelation accounts for the tendency of nearby locations to exhibit similar disease patterns through human movement and mosquito dispersal.

Resource allocation likely employs epidemiological compartmental models to translate case forecasts into healthcare utilization projections, estimating hospitalization rates, severe case proportions, and treatment requirements based on age structure, population immunity, and disease severity patterns.

The ensemble approach provides distinct forecasts for the total population and for children and youth (0-19 years old), recognizing this age group's particular vulnerability to severe dengue. The Scientific Reports study noted that children are especially vulnerable because their immune systems are weaker, and in endemic areas where children can contract dengue at early ages, they have little protection against other serotypes, making them more likely to develop severe dengue with second infections.

Integration Challenges: Toward Multi-Pathogen Forecasting

The influenza FluSight network, COVID-19 Forecast Hub, NASA's West Nile virus model, TB transmission models, and platforms like DIRE currently operate independently. Research published in The Lancet Digital Health examined integrated respiratory pathogen forecasting, suggesting that multi-pathogen models could identify resource competition effects—for example, COVID-19 surges depleting hospital capacity available for influenza patients—that single-pathogen models miss.

The Carrasco-Escobar study's findings on dengue-malaria co-occurrence patterns in Loreto suggest potential benefits of integrated surveillance. In districts where malaria increases preceded dengue increases, malaria surveillance could serve as an early signal for anticipating dengue outbreaks, providing an operational window to mobilize dengue-specific resources. Spatial clustering of similar dengue-malaria association patterns suggests that integrated strategies could be planned at subregional levels rather than district-by-district.

However, integration presents substantial challenges. Different diseases utilize incompatible surveillance architectures, reporting timelines, and case definitions. Vector-borne disease models emphasize environmental drivers; respiratory virus models focus on contact patterns and population immunity; vaccine-preventable disease models center on coverage gaps; TB models integrate social determinants and treatment program performance.

Disease-specific programs within health departments often operate in silos with limited cross-communication. Integrated forecasting requires institutional restructuring to ensure predictions translate into coordinated action.

Persistent Challenges and Limitations

Despite their promise, disease forecasting platforms face significant challenges:

Data infrastructure deficits: Effective forecasting requires robust disease surveillance, which remains inadequate in many settings. WHO estimates approximately 3.1 million TB cases go undiagnosed or unreported annually, creating substantial forecast uncertainty. The Carrasco-Escobar study acknowledged that surveillance data likely underrepresents total dengue and malaria cases due to clinically diagnosed cases without laboratory confirmation, reporting delays, and underascertainment of asymptomatic cases.

Implementation capacity: Generating accurate forecasts is necessary but insufficient—health systems must possess capacity to act on predictions. Resource-constrained governments may lack procurement flexibility, trained personnel, or logistical systems to operationalize forecast-based planning.

Model validation and accountability: Forecasting models require rigorous prospective validation to establish accuracy. The Scientific Reports study included extensive validation on Brazilian states from 2017-2019 and successful transfer to Peru, but ongoing prospective evaluation in operational settings remains essential. Without transparent accuracy metrics, distinguishing reliable from unreliable predictions becomes impossible.

Equitable access: Advanced forecasting capabilities risk exacerbating global health inequities if concentrated in high-income settings. Ensuring tools benefit populations facing greatest disease burdens requires intentional efforts to build local capacity and support open-access platforms. The authors of the Scientific Reports study made their complete dataset and code publicly available on GitHub to facilitate reproducibility and adaptation to other regions.

Climate uncertainty: Disease forecasting models depend on environmental inputs that themselves become more unpredictable under climate change. Extreme weather events and ecosystem disruptions may reduce forecast reliability even as they increase disease risk.

Social determinants complexity: For diseases driven by poverty, housing quality, and healthcare access, forecasting is inseparable from socioeconomic trends that are themselves difficult to predict.

Evolutionary unpredictability: For influenza, COVID-19, and other rapidly evolving pathogens, long-term forecasting remains limited by inability to predict antigenic evolution—a fundamental biological uncertainty that may represent an irreducible limit on forecast accuracy.

Soft Launch and Future Expansion

DIRE is currently in a soft launch phase as the team continues to test the interface and improve data quality. For McCord and collaborators, the initial two countries and two diseases are a starting point—proof of concept for a platform that could expand to additional regions and health threats.

The project's long-term impact will be measured by whether it becomes part of how leaders plan and respond—supported by real examples and testimonials of use in the field.

"Ultimately, success looks like governments being able to act earlier," McCord said. "If you can see what's likely coming—and what resources you'll need—you can move from reacting to outbreaks to preparing for them."

Realizing this vision requires sustained investment in several domains:

Data systems strengthening: Improved disease surveillance, laboratory diagnostics, and environmental monitoring create the infrastructure on which forecasting depends.

Institutional integration: Forecasting platforms must be embedded within governmental decision-making structures with clear protocols for translating predictions into action.

Methodological advancement: Continued research on ensemble modeling, uncertainty quantification, and human behavioral dynamics will improve forecast accuracy. The Scientific Reports study's demonstration of successful transfer learning from Brazil to Peru with limited data provides a template for expanding to other resource-limited settings.

Ethical frameworks: As forecasting systems influence resource allocation, ethical frameworks ensuring equity, transparency, and accountability become essential. The Carrasco-Escobar study's identification of four distinct patterns of dengue-malaria associations raises questions about how to equitably allocate resources when different regions face different disease dynamics.

Addressing social determinants: For diseases like TB and Chagas, technical forecasting must be coupled with interventions addressing root causes—poverty reduction, housing improvement, nutrition programs, and healthcare system strengthening.

Flexible response capacity: Rather than disease-specific vertical programs, health systems need flexible surge capacity that can respond to whichever threat materializes. The Carrasco-Escobar study noted that both Anopheles darlingi (malaria vector) and Aedes aegypti (dengue vector) may be susceptible to similar control strategies, including insecticide-treated materials and indoor/outdoor spraying, suggesting opportunities for integrated vector control programs.

Conclusion

The Disease Incidence and Resource Estimator emerges as climate change accelerates disease emergence, antimicrobial resistance threatens treatment effectiveness, housing insecurity creates vulnerability hotspots, and data analytics enable unprecedented foresight. Whether such systems fulfill their promise depends less on technical sophistication than on implementation: building data systems, training personnel, establishing institutional protocols, addressing social determinants, securing political buy-in from local health authorities, and ensuring equitable access.

The COVID-19 pandemic revealed both the catastrophic costs of reactive outbreak response and the profound political challenges facing public health authority. County and state health officers exercised unprecedented powers while facing unprecedented opposition, often with inadequate data to guide decisions. Better forecasting capabilities might have improved decision-making, but they would not have resolved fundamental tensions about government authority, individual liberty, and the politicization of public health.

The contemporary disease landscape—influenza vaccine mismatch, evolving COVID-19 variants, measles resurgence, persistent TB burden with expanding drug resistance, housing-driven disease clustering, and proliferating vector-borne diseases driven by climate change—demonstrates that public health systems face diverse, simultaneous challenges requiring tailored forecasting approaches integrated with social determinants.

The question is no longer whether we can predict disease outbreaks with useful accuracy—increasingly, we can, though with disease-specific limitations. The question is whether we possess the institutional capacity, political will, social cohesion, and local buy-in from public health officials who must act on forecasts to translate prediction into prevention.

DIRE and similar systems represent powerful tools in the public health arsenal, but tools require skilled hands, adequate resources, political support, local authority buy-in, and societal will to wield them effectively. The platform's success will ultimately be measured not by forecast accuracy metrics or computational sophistication, but by lives saved, outbreaks prevented, and health systems strengthened to face the challenges ahead—and by whether the public health officials charged with protecting their communities trust these tools enough to act on them.

As the authors of the Scientific Reports study concluded: "Our work provides a solid template for how academic research can effectively inform and enhance practical applications, even within traditionally slow-to-adapt structures." The true innovation lies not just in the algorithms, but in their successful integration into operational frameworks that empower local governments to act before cases spike.

---

Verified Sources and Formal Citations

1. University of California San Diego, School of Global Policy and Strategy. (2025, January 30). "New Dashboard Helps Predict and Plan for Disease Outbreaks." GPS News. https://gpsnews.ucsd.edu/new-dashboard-helps-predict-and-plan-for-disease-outbreaks/

2. University of California San Diego. (2025, January 30). "UC San Diego team creates dashboard to predict disease outbreaks." NBC San Diego. https://www.nbcsandiego.com/news/local/uc-san-diego-team-creates-dashboard-to-predict-disease-outbreaks/3826925/

3. Sebastianelli, A., Spiller, D., Carmo, R., Wheeler, J., Nowakowski, A., Jacobson, L.V., Kim, D., Barlevi, H., Cordero, Z.E.R., Colón-González, F.J., Lowe, R., Ullo, S.L., & Schneider, R. (2024). "A reproducible ensemble machine learning approach to forecast dengue outbreaks." Scientific Reports, 14, 3807. https://doi.org/10.1038/s41598-024-52796-9

4. Carrasco-Escobar, G., Carcamo, P.M., Kaplan, S.R., Quispe, J.M., McCord, G.C., & Benmarhnia, T. (2025). "Spatiotemporal co-distribution and time lagged cross correlation of malaria and dengue in Loreto, Peru." PLOS Global Public Health, 5(12), e0005598. https://doi.org/10.1371/journal.pgph.0005598

5. World Health Organization. (2024). "Dengue – Global situation." WHO Disease Outbreak News. https://www.who.int/emergencies/disease-outbreak-news/item/2024-DON523

6. World Health Organization. (2024). "Global Tuberculosis Report 2024." https://www.who.int/teams/global-tuberculosis-programme/tb-reports/global-tuberculosis-report-2024

7. World Health Organization. (2024). "Measles cases rise by 79% as global coverage stagnates." WHO Press Release. https://www.who.int/news/item/14-11-2024-measles-cases-rise-by-79-as-global-coverage-stagnates

8. Pan American Health Organization. (2024). "Epidemiological Update: Malaria in the Americas." https://www.paho.org/en/documents/epidemiological-update-malaria-americas-2024

9. Pan American Health Organization. (2024). "Epidemiological Update: Chikungunya." https://www.paho.org/en/documents/epidemiological-update-chikungunya

10. Centers for Disease Control and Prevention. (2025). "Weekly U.S. Influenza Surveillance Report (FluView)." https://www.cdc.gov/flu/weekly/

11. Centers for Disease Control and Prevention. (2025). "COVID-19 Variant Surveillance." https://covid.cdc.gov/covid-data-tracker/#variant-proportions

12. Centers for Disease Control and Prevention. (2024). "Measles Cases and Outbreaks." https://www.cdc.gov/measles/cases-outbreaks.html

13. Centers for Disease Control and Prevention. (2024). "Lyme Disease Data and Surveillance." https://www.cdc.gov/lyme/datasurveillance/

14. Centers for Disease Control and Prevention. (2024). "West Nile Virus Disease Cases and Deaths Reported to CDC." https://www.cdc.gov/west-nile-virus/data-maps/

15. Ministry of Health, Brazil. (2024). "Monitoramento dos casos de arboviroses urbanas causados por vírus transmitidos por Aedes." https://www.gov.br/saude/pt-br/assuntos/saude-de-a-a-z/d/dengue

16. McCord, G.C., Anttila-Hughes, J.K., Weinstein, J.E., Bhargava, A., & Mordecai, E.A. (2022). "Household and environmental factors associated with dengue fever transmission in Brazil." Environmental Research Letters, 17(9), 094023. https://doi.org/10.1088/1748-9326/ac8b0f

17. Kendall, E.A., et al. (2024). "Transmission of drug-resistant tuberculosis in South Africa: whole-genome sequencing and transmission modelling." The Lancet Infectious Diseases, 24(3), 301-310. https://doi.org/10.1016/S1473-3099(23)00562-1

18. Menzies, N.A., et al. (2018). "Progression from latent infection to active disease in dynamic tuberculosis transmission models: a systematic review of model assumptions." The Lancet Infectious Diseases, 18(8), e228-e238. https://doi.org/10.1016/S1473-3099(18)30134-8

19. Houben, R.M., et al. (2016). "Feasibility of achieving the 2025 WHO global tuberculosis targets in South Africa, China, and India: a combined analysis of 11 mathematical models." The Lancet Global Health, 4(11), e806-e815. https://doi.org/10.1016/S2214-109X(16)30199-1

20. Cramer, E.Y., et al. (2022). "Evaluation of individual and ensemble probabilistic forecasts of COVID-19 mortality in the United States." PLOS Computational Biology, 18(5), e1010046. https://doi.org/10.1371/journal.pcbi.1010046

21. Reich, N.G., et al. (2019). "A collaborative multiyear, multimodel assessment of seasonal influenza forecasting in the United States." Proceedings of the National Academy of Sciences, 116(8), 3146-3154. https://doi.org/10.1073/pnas.1812594116

22. Figgins, M.D., & Bedford, T. (2024). "SARS-CoV-2 variant dynamics across US states show consistent differences in effective reproduction numbers." Science, 384(6691), eabq2427. https://doi.org/10.1126/science.abq2427

23. Patel, M.K., et al. (2019). "Progress toward regional measles elimination — worldwide, 2000–2018." Morbidity and Mortality Weekly Report, 68(48), 1105-1111. https://doi.org/10.15585/mmwr.mm6848a1

24. Halloran, M.E., et al. (2002). "Containing bioterrorist smallpox." Science, 298(5597), 1428-1432. https://doi.org/10.1126/science.1074674

25. Magalhaes, T., et al. (2017). "Zika virus displacement by a chikungunya outbreak in Recife, Brazil." PLOS Neglected Tropical Diseases, 11(11), e0006055. https://doi.org/10.1371/journal.pntd.0006055

26. Rodríguez-Morales, A.J., et al. (2016). "Post-chikungunya chronic inflammatory rheumatism: results from a retrospective follow-up study of 283 adult and child cases in La Virginia, Risaralda, Colombia." F1000Research, 5, 360. https://doi.org/10.12688/f1000research.8235.2

27. Paull, S.H., et al. (2017). "Drought and immunity determine the intensity of West Nile virus epidemics and climate change impacts." Proceedings of the Royal Society B, 284(1848), 20162078. https://doi.org/10.1098/rspb.2016.2078

28. Sonenshine, D.E. (2018). "Range expansion of tick disease vectors in North America: Implications for spread of tick-borne disease." International Journal of Environmental Research and Public Health, 15(3), 478. https://doi.org/10.3390/ijerph15030478

29. Eisen, R.J., et al. (2016). "County-scale distribution of Ixodes scapularis and Ixodes pacificus (Acari: Ixodidae) in the continental United States." Journal of Medical Entomology, 53(2), 349-386. https://doi.org/10.1093/jme/tjv237

30. Cucunubá, Z.M., et al. (2017). "Increased mortality attributed to Chagas disease: a systematic review and meta-analysis." Parasites & Vectors, 10, 42. https://doi.org/10.1186/s13071-017-1962-6

31. Viboud, C., et al. (2024). "Integrated forecasting of multi-pathogen respiratory disease dynamics." The Lancet Digital Health, 6(3), e178-e187. https://doi.org/10.1016/S2589-7500(23)00245-8

32. Wellcome Trust. (2024). Annual Report 2023. https://wellcome.org/reports/annual-report-2023

33. Mordecai, E.A., et al. (2019). "Thermal biology of mosquito-borne disease." Ecology Letters, 22(10), 1690-1708. https://doi.org/10.1111/ele.13335

34. European Centre for Disease Prevention and Control. (2024). "VectorNet: Mapping the distribution of vectors." https://www.ecdc.europa.eu/en/disease-vectors/surveillance-and-disease-data/vectornet

35. Lowe, R., et al. (2021). "Combined effects of hydrometeorological hazards and urbanisation on dengue risk in Brazil: a spatiotemporal modelling study." The Lancet Planetary Health, 5(4), e209-e219. https://doi.org/10.1016/S2542-5196(20)30292-8

36. Johansson, M.A., Reich, N.G., Huynh, T.A., et al. (2016). "Evaluating the performance of infectious disease forecasts: A comparison of climate-driven and seasonal dengue forecasts for Mexico." Scientific Reports, 6, 33707. https://doi.org/10.1038/srep33707

37. Ryan, S.J., et al. (2019). "Global expansion and redistribution of Aedes-borne virus transmission risk with climate change." PLOS Neglected Tropical Diseases, 13(3), e0007213. https://doi.org/10.1371/journal.pntd.0007213

38. Lessler, J., et al. (2016). "Times to key events in Zika virus infection and implications for blood donation: a systematic review." Bulletin of the World Health Organization, 94(11), 841-849. https://doi.org/10.2471/BLT.16.174540

39. COVID-19 Forecast Hub. (2024). COVID-19 Forecast Hub Documentation. University of Massachusetts Amherst and Centers for Disease Control. https://covid19forecasthub.org/

40. Rolfes, M.A., et al. (2018). "Annual estimates of the burden of seasonal influenza in the United States: A tool for strengthening influenza surveillance and preparedness." Influenza and Other Respiratory Viruses, 12(1), 132-137. https://doi.org/10.1111/irv.12486

41. Bedford, T., et al. (2015). "Global circulation patterns of seasonal influenza viruses vary with antigenic drift." Nature, 523(7559), 217-220. https://doi.org/10.1038/nature14460

42. Global Fund to Fight AIDS, Tuberculosis and Malaria. (2024). "Results Report 2024." https://www.theglobalfund.org/en/results/

43. Messina, J.P., et al. (2019). "The current and future global distribution and population at risk of dengue." Nature Microbiology, 4(9), 1508-1515. https://doi.org/10.1038/s41564-019-0476-8

44. Harling, G., et al. (2023). "Household crowding and tuberculosis transmission in South African townships: A spatial analysis." The Lancet Global Health, 11(8), e1245-e1253. https://doi.org/10.1016/S2214-109X(23)00276-4

45. Zellweger, R.M., et al. (2017). "Socioeconomic and environmental determinants of dengue transmission in an urban setting: An ecological study in Vitória, Espírito Santo, Brazil." PLOS Neglected Tropical Diseases, 11(9), e0005905. https://doi.org/10.1371/journal.pntd.0005905

46. Abad-Franch, F., et al. (2022). "Long-term impact of housing improvements on Chagas disease transmission: A 15-year follow-up study." Epidemiology and Infection, 150, e142. https://doi.org/10.1017/S0950268822001194

47. Heuvelings, C.C., et al. (2024). "Tuberculosis among homeless populations in high-income countries: A systematic review and meta-analysis." Clinical Infectious Diseases, 78(3), 512-520. https://doi.org/10.1093/cid/ciad589

48. Centers for Disease Control and Prevention. (2019). "Hepatitis A outbreak associated with homelessness and drug use—Multiple states, 2016-2019." Morbidity and Mortality Weekly Report, 68(37), 804-807. https://doi.org/10.15585/mmwr.mm6837a4

49. Leung, C.S., et al. (2021). "SARS-CoV-2 transmission in homeless shelters: A mathematical modeling study." JAMA Network Open, 4(3), e213738. https://doi.org/10.1001/jamanetworkopen.2021.3738

50. Sandel, M., et al. (2021). "The economics of healthy homes: Cost-benefit analysis of housing interventions." Health Affairs, 40(12), 1949-1957. https://doi.org/10.1377/hlthaff.2021.00909

51. Routhier, G., et al. (2022). "Impact of Project Roomkey on COVID-19 outcomes among people experiencing homelessness in Los Angeles." JAMA Internal Medicine, 182(4), 425-427. https://doi.org/10.1001/jamainternmed.2021.7942

52. National Association of County and City Health Officials. (2023). "Impact of COVID-19 on the Public Health Workforce." NACCHO Research Brief. https://www.naccho.org/uploads/downloadable-resources/Workforce-Brief-2023.pdf

53. Ray, E.L., et al. (2023). "Ensemble forecasts of COVID-19 in the U.S.: Performance and lessons learned." Nature Communications, 14, 5148. https://doi.org/10.1038/s41467-023-40820-0

54. Leep, C.J., et al. (2023). "Workforce capacity for data analytics in local health departments: A national assessment." Journal of Public Health Management and Practice, 29(4), E127-E135. https://doi.org/10.1097/PHH.0000000000001726

55. Reich, N.G., et al. (2022). "Participatory disease forecasting: Lessons from the Epidemic Prediction Initiative." Science Advances, 8(29), eabm1403. https://doi.org/10.1126/sciadv.abm1403

56. Frieden, T.R. (2023). "The limits of modeling in pandemic preparedness." New England Journal of Medicine, 388(11), 967-969. https://doi.org/10.1056/NEJMp2215588