Abstract

Introduction and purpose

To monitor pandemic pathogens effectively, modern surveillance systems should make use of the growing wealth of routine data from the health sector and from a range of other sources with new applications to disease surveillance such as mobility, internet searches, and wastewater. Public health experts must integrate these data in new ways that increase their value. We need purpose-built systems to detect new and evolving threats and to provide information as quickly as possible about those threats. What are the characteristics of the new pathogens or new variants of existing pathogens? What is their incidence and prevalence? What is the vulnerability of the population to infection and disease? What is the impact of our efforts to respond to these threats?

Systems to generate, integrate, and interpret these data should be designed and built with the explicit purpose of providing timely evidence to inform decisions about disease control and mitigation. First, they will provide direct input into decision making. For example, evidence of low vaccine effectiveness may prompt efforts to boost or change formulations or doses. As another example, real-time lab order data for diagnostic tests may prompt adjustments to resource allocation. Second, these data will parameterize scenario and forecasting models (1–3). For instance, estimates of per-case severity of a new variant, incorporated into forecasts or other models of case burden, may influence planning for hospital capacity and supply stockpiling and distribution.

This document reflects the framing ideas and the discussions held at a symposium organized by Harvard T.H. Chan School of Public Health entitled “Quantitative Tools and Data Opportunities for Pandemic Surveillance and Response,” held June 29–30, 2022, involving a range of public health and public officials, surveillance experts and other epidemiologists, and epidemic modelers. We first aim to identify the most important decisions for disease control and mitigation and the evidence that is needed to inform them. We then describe a set of surveillance activities designed to provide timely, reliable, and appropriately scaled data to inform these decisions. Our focus in this report is limited to domestic detection, characterization, and estimation of the burden of a pandemic pathogen in terms of direct health effects. Although we note the importance of monitoring economic, social, and indirect public health impacts of a disease control measures, we do not offer a comprehensive treatment of this element of pandemic monitoring and response activities. This paper builds on earlier efforts (4) while incorporating both the new possibilities that technology now provides, as well as the lessons of COVID-19.

We differentiate between two related but distinct goals of surveillance, as this document will focus on only one. The first goal is to provide early warning about a potential pandemic, and so this type of surveillance includes global monitoring and rapid identification of domestic introductions. The second goal is to provide support for decision making during an ongoing pandemic, including tracking incidence, prevalence, and the pathogen’s properties. While we will briefly remark on the former–surveillance for early warning–we will primarily focus on the latter–surveillance for decision making.

Detecting a jurisdiction’s first cases of a new disease

The first set of decisions faced by a domestic public health jurisdiction, following the appearance of a pandemic threat somewhere else in the world, concerns the questions of whether, how, and to what extent to scale up a response to reduce the risk of importation or, if importation has happened, to control its spread within the jurisdiction. Measures to reduce importation via restriction or testing of inbound travelers may buy limited time to prepare (5, 6), though such measures lose relevance once local transmission is established (7).

To inform decisions about how to balance scarce public health resources between preventing importation vs. controlling local spread, it is critical to assess the risk that the infection has already arrived and started spreading within the jurisdiction. Testing and sequencing of specimens from international travelers at airports and analysis of wastewater from international flights may provide evidence of pathogen importation (8). Early evidence of local spread may come from informal communications among health care providers, reporting systems such as ProMED-mail (9), and “pre-health care” data (e.g., absenteeism, internet search queries). Signals may arise from monitoring of syndromes compatible with infections, or the volume, distribution, and results from clinical laboratory tests. Increasingly, wastewater monitoring can be the site of early detection of new threats or variants (10) and from routine programs or enhanced efforts at sequencing of clinical samples (11). They may also come from anomalous findings in sentinel and research efforts [e.g., the Seattle Flu Study at the start of COVID-19 (12)].

With each type of monitoring, there is first the question of what defines the signal we are looking for and then there is a tradeoff between having a highly sensitive and timely system capable of sounding an alarm early on one hand and producing too many false alarms on the other. In most situations, a high positive predictive value for such systems will be essential because the cost of responding to frequent false alarms is high. Much more work is needed to assess how to use and combine complementary monitoring signals to identify points at which an alarm should be escalated into a response.

Surveillance for decision making during a pandemic

A comprehensive list of the decisions and guidance required in a pandemic would fill a much longer document than this. Still, based on the combined experience of the emergency response phase of COVID-19 (2020–2023) and H1N1 influenza (2009–10) pandemics, we propose a set of consistent themes that capture many of the major types of decisions arising both in the early days of each pandemic and throughout.

A brief, necessarily incomplete, list of such decisions follows, adapted from the list in (4):

1. Public health goals of a response (elimination, control, protection of high-risk groups, protection of health care functioning, or a combination) and overall scale of response needed to meet these goals.

2. Timing of scale-up and scale-down of response.

3. Choice of nonpharmaceutical countermeasures (individual-targeted such as quarantine, isolation, and personal protection; population-targeted such as closures). This includes decisions about the timing, magnitude, and geographic range of protective measures that may be socioeconomically costly. A related set of decisions concerns how such measures should be prioritized, i.e., who should receive protective equipment when it is scarce, and how closures should be targeted to reduce economic and social disruption.

4. Choice of medical countermeasures, including diagnostics, therapeutics, and vaccines. This includes decisions about development, stockpiling, procurement, expanding capacity (e.g., building alternative care sites), and more. Here too, questions of allocation and prioritization are central. This also includes planning for potential surges.

5. Specific policies for each of the issues above in special populations including vulnerable communities, and settings such as health care, schools, congregate settings, transport, etc. (See Centering Equity below)

6. Balance between community countermeasures to reduce severe disease or reduce transmission (e.g., allocation of resources to those at high risk of complications or high risk of transmission).

7. Design and implementation of staged alert systems to provide real-time risk awareness and trigger policy changes (13, 14).

8. Imposition and removal of international travel screening and restrictions.

9. Choice of public health communication strategies.

Each of these decisions requires specific data to decide how to improve health equitably, effectively, and efficiently while minimizing social and economic disruption. For example, decisions on testing, isolation, and quarantine policies require evidence on the natural history of infectiousness (or at least a proxy such as viral load), test sensitivity at different levels of viral shedding, the relationship between symptoms and infectiousness, and the potential economic and social consequences for various communities of the policies under consideration. In contrast, decisions about the timing of vaccine boosters require evidence on the effectiveness of existing vaccines against infection, transmission, and severe disease endpoints, stratified by such factors as pathogen variant, time since vaccination, and age, as well as understanding of how vaccine protection is distributed across demographic groups.

Decisions faced by state, territorial, local, and tribal authorities

In the federal system in the United States, public health is decentralized and typically not coordinated among states. State, territorial, local, and tribal (STLT) governments are responsible for nearly all binding policy decisions in public health, with governance health structures varying by state (15). The purview of these bodies includes (16) prescribing and enforcing isolation, quarantine, mask mandates, and restrictions on businesses and gatherings; vaccine prioritization and distribution; and (to a degree) diagnostic testing. They also hold responsibility for closely related areas, such as public education. STLT governments all have a desire for similar types of data, but vary in how much they need, how quickly they need it, and how they use it.

Many decisions involve procurement and distribution of countermeasures. Because STLT authorities are making allocation decisions within their jurisdictions (e.g., for counties, cities, hospitals, schools), jurisdiction-wide measures of disease activity are rarely sufficient; instead, more geographically granular numbers are required (Table 1).

Data needs for decision support: the COVID-19 experience

A range of data sources could and, during the COVID-19 pandemic, did provide evidence to support decisions by health authorities. Following initial social media reports of clusters of pneumonia, some of the earliest specific data to characterize the COVID-19 threat came from traditional sources, such as from case reports posted on Chinese public health department websites (17). A key challenge was the repeated change in the syndromic case definition in the early days (18). But other early data came from unexpected sources, such as cruise ships (19), restaurants (20), and fishing vessels (21), where conditions allowed inference of the path of transmission and thereby provided evidence about the degree and mechanisms of spread. Specifically, these provided some of the earliest evidence of asymptomatic and aerosol spread, which, when properly interpreted, aided in the design and prioritization of testing and other control measures. As had been true in the 2009 influenza pandemic (22, 23), sampling of travelers provided early estimates of the extent of global spread, growth rates, and likely under-detection (24, 25).

As the pandemic spread, the strengths and limitations of each data source became evident. Multiple data types were required to provide even an incomplete picture of trends in incidence and prevalence and behavior (26, 27). For example, case counts were used as an important indicator of disease burden. However, the relationship between new cases and true incidence varies as a function of numerous factors, including test availability, test reporting requirements by jurisdiction (which did not always include reports of negative tests), rates of testing through clinical facilities (which declined with the growth of rapid antigen testing), and incentives to get or avoid testing (Figure 1). Some of these limitations can be mitigated by breaking out case counts by the reason why an individual was tested (symptoms, travel, surveillance), but this was not consistently done in the US. As a result of these limitations, hospitalizations and even deaths were increasingly used as the more reliable indicators of case numbers, sacrificing some timeliness for a more consistent relationship to the underlying incidence of infection. Random sampling approaches (described below) can overcome these limitations and provide more consistent and reliable estimates of incidence and prevalence and how these change over time. Only the United Kingdom and Luxembourg used random sampling on a large scale, perhaps because of the cost and logistical challenges. Notwithstanding their limitations, case counts were the major early data source in the United States and provided critical evidence especially when linked to demographic information. Syndromic surveillance–done routinely as part of monitoring influenza trends–from emergency room visits and hospital admissions were also valuable data sources, particularly when testing was limited. However, interpretation of syndromic surveillance was complicated by changes in healthcare seeking behavior and the increased use of telemedicine.

Open in a separate window

Figure 1

Testing patterns that vary in space and time as a result of individual incentives (left) and barriers (right) determine a changing relationship between epidemiological quantities (top left) and reported case counts, making these counts an uncertain source of evidence for current case burden and for calibration of transmission models.

Novel data streams provided confirmatory evidence as well as early warnings of trends that might not be evident in case counts. For example, wastewater surveillance for SARS-CoV-2 was adopted in numerous jurisdictions from 2020 to 2022 and provided evidence on local epidemic trends, although the precise relationship between wastewater abundance and the number of infected persons depends on the wastewater sampling scheme and on shedding patterns (among other issues), and thus difficult to quantify (28). Moreover, by its nature, wastewater cannot indicate who has been infected, thus leaving the demographic profile of infected persons uncertain. Finally, wastewater surveillance as currently applied will miss infections in areas with high reliance on septic systems, which serve roughly one fifth of the households in the U.S. with heavy concentrations in certain geographies, inducing inequities in whose infections are tracked by this approach (29).

Another novel data stream is the use of population-wide distributions of viral load measured from PCR testing, which, in the aggregate, can provide information on the trajectory of the epidemic, even from a single cross-sectional analysis (30). This approach has reached proof of concept and has the advantage that it may be less sensitive to trends in testing behaviors than measures of incidence based on case counts, and unlike wastewater surveillance, it can provide some information on the demography and precise location of cases. However, further work is needed to see how a transition to non-PCR testing for many new cases, the halting of pre-procedural and asymptomatic testing, and the shifts in viral kinetics that come with immunity (from vaccination and infection) affects the nature of this signal (31–33). Moreover, the identifiability of time-since-infection from viral load, which is needed for the approach to work, depends on the asymmetry in viral load over time [fast rise, slower decline (30)], which may or may not be a feature of future infectious diseases.

Digital data can also be used for surveillance and to inform on epidemic trajectory. ProMED-mail (9) and HealthMap (34) are valuable for flagging and disseminating reporting and information on events known or suspected to be infections and outbreaks. Data from search engines, social media, and news reports data can also inform epidemic dynamics and for forecasting (35–37).

Finally, testing for antibodies in sera collected either for the purpose of serologic surveillance or in convenience samples (e.g., blood banks, discarded clinical samples) was used to characterize both the landscape of population immunity (i.e., who was and wasn’t vulnerable to reinfection) and to distinguish between those who had acquired immunity via vaccination vs. infection (38). Secondary analyses of COVID-19 vaccine studies identified complexities in answering the latter question, finding that infection does not reliably induce antibodies to non-vaccine antigens in vaccinated individuals (39).

An important conclusion is that no one data source or surveillance tactic is sufficient. In a setting like the U.S., multiple surveillance approaches are needed at scale. Beyond the obvious need to combine data sources, several points stand out.

The first is the value of data completeness and of linking data types to produce evidence that is greater than the sum of the parts. For instance, while counts of cases and hospitalizations are valuable, missing race/ethnicity, geographic, and other patient characteristics have impeded efforts to improve services to groups that are underserved or experience high disease burden and to improve equity in health-related outcomes. Similarly, meticulously linking sequence data from patient isolates with demographic and clinical predictors of severe outcomes, including vaccination history, and clinical outcomes can help to evaluate the threat posed by novel variants (40). Unfortunately, despite prodigious amounts of SARS-CoV-2 sequencing in the U.S., this form of linkage has been relatively rare to date.

Second is the value of clear and accessible data dashboards with transparent data sources to make the state of the epidemic locally evident to the public. The same data should also be available to analysts in public health departments, academia, and other sectors via application programming interfaces (APIs) to facilitate rapid data analysis. This can facilitate shared decision making and help to increase public support for control measures. For example, the city of Austin, Texas developed a COVID-19 staged alert system that guided local policy between May 2020 and March 2022 (13, 14). The public-facing dashboard featured a single graph that tracked COVID-19 hospital admissions and clearly indicated thresholds between the red, orange, yellow, green, and blue risk levels that were linked to specific actions. The county judge, city mayor, and public health authority cited the dashboard almost daily to communicate risks, explain changes in policy, and cultivate adherence via news outlets and social media. This system was only possible because local authorities required area hospitals to report daily admissions beginning in April 2020, long before such data were generally available.

Finally, discussion at the Symposium emphasized the value of metrics that could be compared across jurisdictions. Decision-makers expressed a desire for objective criteria by which their performance can be judged. Comparisons across states, for example, were hampered by differential testing rates that affected case counts in ways not reflecting actual prevalence. A CDC-supported academic effort called covidestim (41) used Bayesian evidence synthesis to harmonize estimates of current and cumulative infections across states and counties, providing an example of what could be done by health authorities. However, this effort was also hampered by unanticipated changes in reporting tempo, as well as ‘data dumps’ and data backfilling. Different definitions of COVID-19 hospitalization across states and over time impeded comparisons of outcomes that would have provided indications to elected leaders of the quality of their responses and informed improved responses.

Surveillance inputs to forecasts, scenario projections, and analytic models

As noted above, many aspects of pandemic decision-making can directly incorporate evidence from surveillance, and will also make use of

• Nowcasts: estimates of current burden of cases, hospitalizations, deaths, and other quantities that account for delays in reporting (42);

• Forecasts: relatively short-term projections using time-series and other statistical modeling techniques, sometimes supplemented with transmission-dynamic approaches to estimate future case, morbidity, and mortality burden, typically on the scale of days to weeks (1, 3);

• Scenario models: longer-term estimates of pandemic dynamics using transmission-dynamic modeling approaches to anticipate multiple possible futures under stated assumptions about behavior, viral evolution, vaccine durability, etc., typically on the scale of months to years (43);

• Results from analytic models: estimates about different characteristics of the pathogen or a population of concern that are specifically designed to inform a decision or guidance, such as school- or nursing-home-based testing policies (44), border restrictions or contact tracing procedures (45), or quarantine approaches (46).

These categories of decision-support tools require estimates of input quantities that represent the assumptions of the models–for example, in scenario models, estimates of per-case severity or vaccine effectiveness. These estimates need to be timely, representative, and specific to the pathogen or variants circulating or anticipated to circulate (for example, due to importation). In addition, scenario and forecast model output must be calibrated to existing measurements of disease burden: incidence of infection, diagnosed cases, hospitalizations, deaths, and other relevant metrics, as well as against cumulative measures such as seroprevalence. The categories of input and output are somewhat fluid, as a model with sufficient data to calibrate outputs may be able to estimate the values of some of the quantities described here as inputs. In a fully Bayesian framework, both external estimates (as priors) and calibration to output data may contribute to posterior parameter estimates.

For forecasts, evaluation can be performed quickly due to the short-term horizon of the predictions made, with results that can provide feedback to modelers about places where models are mis-specified. Evaluating scenario projections is more complicated, as multiple sets of counterfactual projections are made under different assumptions about how a pandemic situation will evolve over the course of months or years (47). Most (or perhaps all) of the scenarios will not be realized exactly as assumed, making evaluation less straightforward.

Together the quality and timeliness of these input parameters and output calibrations are important determinants of how useful a model is for decision making. While there are techniques to adjust for incomplete or lagged information, the absence of certain ingredients–especially model output calibration targets such as numbers of cases or hospitalizations–can critically compromise the ability to generate models that reflect reality to the point of hampering basic situational awareness. Data systems that support modeling and in turn decision-making during pandemics should be considered vital national security capabilities and prioritized accordingly.

A list of the key needs for model inputs is as follows, many of which may change as a pathogen evolves (referred to below by their letters):

1. Pathogen kinetics/epidemiological parameters (e.g., incubation period, latent period, infectious period, infection fatality ratio). Estimation of these inputs may itself require simple models, particularly at the early stages of a pandemic [ref: Gostic paper].

2. Transmissibility and efficiency of various transmission mechanisms

3. Risk factors for infection and severity

4. Individual and population immunity (including effects of infection, vaccination, and waning)

5. Diagnostic test characteristics, including specificity and sensitivity for active (acute) and past infection

6. Vaccine effectiveness and waning of effectiveness, for infection, severe disease, and mortality endpoints.

7. Treatment effectiveness

8. Policies, uptake, and effectiveness of nonpharmaceutical interventions

9. Population mobility and interactions: contact networks and patterns by setting

10. Importation risk

11. Other co-circulating pathogens of concern (e.g., if concurrent with significant influenza transmission)

12. Capacity and utilization of healthcare resources (including hospital beds, therapeutics, and vaccines)

Key additional data requirements for fitting models– as well as for general situational awareness – include:

1. Geographically and demographically stratified incidence, duration and prevalence of infection, hospitalization, ICU admission, death, and other relevant metrics associated with the pathogen, ideally by variant.

2. Strain-specific incidence.

Meeting these needs

This draft framework is a preliminary attempt to scope a system that could meet the needs listed above for situational awareness, decision support, and inputs and outputs for modeling and analytics for a new variant or a new pandemic. Capacity to achieve these would also be applicable to other pathogens, especially, but not only, respiratory ones.

Strengthening personnel and research ties, including globally

In response to the COVID-19 emergency and the need for expertise to gather, analyze, and interpret evidence around the pandemic and the clinical and public health responses, many academics put aside their usual research programs to engage directly in public health activities and research. The close interactions between academics and local, state, and national public health officials were often productive and important for guiding the pandemic response but raised issues that should be addressed before the next pandemic. These include the ad hoc way in which these academic-public health collaborations came into being, the lack of uniformity of access to academics with appropriate expertise across states, and the misalignment of incentives between public health and academic work.

Ideally, academia-public health collaborations can be rapidly scaled up in times of need through established pathways. One idea is to create a “rotator” program, in which academics (and potentially those in training, including doctoral students and postdoctoral fellows) are embedded within public health agencies–and similarly public health officials are embedded within academic groups–for intervals (such as 3 or 6months) that build familiarity, collegiality, and accessibility. The LEAP fellowship through the Infectious Disease Society of America (78) and the joint Infectious Diseases/EIS fellowship (79) programs are efforts in this direction. Another approach is to establish an academic career path in which some fraction of time and effort are based in public health activities, analogous to academic medicine paths in which researchers spend some fraction of their time doing clinical work. Cooperative agreements established in 2023 between the CDC’s Center for Forecasting and Outbreak Analytics (CFA) and academic and other groups include a surge provision whereby the performers on these agreements would provide scientific assistance in times of crisis. Relatedly, an official “public health reserve corps” of analysts and modelers could provide a workforce available to be called up to prepare for and respond to emergencies. Formal recognition of these paths as prestigious and vital, and placing value on these activities within the academic systems of rewards and incentives, will be key to success (Box 1).

Often tools developed for one public health jurisdiction solve common problems and could be readily implemented in other jurisdictions, underscoring the importance of making code open access and ideally making tools generalizable. This would have the benefits of “not reinventing the wheel” and allowing those jurisdictions lacking local expertise access to useful tools. A curated clearinghouse of such tools, organized by research problem and perhaps hosted by CFA offers one strategy for providing access. Broad efforts to create and maintain state-of-the-art tools for epidemiologic modeling, such as Epiverse (82) and Recon are encouraging developments in this space (83).

We have focused on domestic systems in the United States, but that international cooperation is essential for multiple reasons. Maintaining systems to de new pathogens at multiple locations in a globalized world will speed detection, facilitating timely responses including development of countermeasures (25) and providing early warning to other jurisdictions (84). For pathogen characterization, as we note in Box 2, each jurisdiction can benefit from findings in other jurisdictions on quantities that are relatively similar across populations, including for example the effectiveness of vaccines and treatments, as well as certain features of infection natural history (e.g., viral load kinetics). Strengthening these capacities globally means that locations that have exceptional data and study infrastructure and/or early experience with a pathogen or variant, can contribute to the global store of knowledge of pathogen characteristics (49, 54, 85–87).

Educating the consumers

Tools are best deployed by those who understand how they work, how they are limited, and how they can be modified to improve their applicability to specific situations. Workforce development mandates to build subject matter expertise within public health departments, such as through CDC efforts via the Office of Science and the Office of Advanced Molecular Detection and through fellowships such as the CDC/Association of Public Health Laboratories (APHL) program, are critical efforts. While waiting for these workforce programs to get up and running, and since public health officials may not stay current with the frontiers of analytical and modeling methods, opportunities for regular formal trainings should be developed. For example, meetings such as the Council of State and Territorial Epidemiologists Annual Conference could provide a forum for workshops on advances in modeling, genomic epidemiology, and other fields. Relatedly, encouraging public health officials to attend field-specific meetings (e.g., Epidemics, Applied Bioinformatics in Public Health Microbiology) could provide opportunities for knowledge sharing, relationship building, and networking across sectors and disciplines.

Improving knowledge flow

Successful communication of a health agency’s current understanding of a pandemic and outlook for its future requires a combination of approaches to communicate different kinds of data and outlooks, for different audiences. It has been suggested that principles for such communications include: thematic structure related to informing key decisions, synthesis of evidence from multiple sources, quantification of uncertainty, inclusion of visualizations as well as text and tables, and inclusion of forward-looking material (outlooks for the future) (91); another important principle is open access to the data underlying figures in these reports. CDC’s Technical Reports on the Mpox epidemic in 2022 (92) sought to put these principles into practice, explicitly emulating aspects of the United Kingdom Health Security Agency (UKHSA) Technical Briefings from COVID-19 (93). Creating a regular cadence for such reports during an emergency, as was the case in the United Kingdom during the height of the COVID-19 pandemic, may help develop an audience and facilitate knowledge flow.

Expanding the range of data types

As described above, any health system, but particularly one as decentralized as that of the U.S., benefits from the ability to ingest and synthesize multiple types of data. Increased use of wastewater data (106, 107) has contributed to early warning of rising infection incidence and to surveillance for new variants. Further work to standardize collection and better define the quantitative relationships between true infection incidence and total and variant-specific concentrations of viral genomes in wastewater is needed to improve the value of such data, as well as a clear mapping of where it will not be informative, such as areas using septic systems. Likewise, mobility data from various sources (58) can be useful in informing strategies for disease monitoring and surveillance, modeling disease spread, and guiding interventions. Immune measures from serology provide a window onto past infection and a lens onto the landscape of risk (108). Here, further work is needed to ensure data standardization and accuracy as well as routine and frequent updating to capture important temporal variations. Such new forms of data may also raise privacy considerations that have not entirely been solved (109, 110).

Crowdsourced and survey data (111, 112) can provide important insights into behaviors that affect the interpretations of other data; for example, the increasing prevalence of self-testing using antigen tests for COVID-19 reduces the utility of PCR-positive case counts.

A key to making use of this expanded range of data types is solving the problem of how to synthesize multiple data types into a single estimate of a quantity of interest, accounting for the different properties of each data type (76), including understanding the different biases that will affect each data stream. Significant further work is needed to advance the ability to do this in real time. A related but distinct problem is how to link data across data systems to understand the continuum of care and otherwise improve inference about the course of individual cases.

Expanding the range of data sources

The use of claims data from health care payers (insurers) and electronic medical records from providers has exploded in many areas of health services research. There have been some notable examples of such data for surveillance to address the questions described in this report (53, 72–74), but in the United States there remains untapped potential to expand such efforts and improve their timeliness. This will require building relationships between public health entities and health care systems in their jurisdictions, including relationships between scientific investigators in each sector with regular discussions for bidirectional learning. In the spirit of administrative preparedness above, this will require up-front planning of master agreements to move resources in a timely fashion to address pressing questions. Health providers and public health have suffered from a “two cultures” challenge that results in the need to expand public health training of investigators and other personnel in health systems, acknowledge the contributions of health systems to community benefits, and find ways to produce incentives so that contributing to public health surveillance aligns with the business interests of health systems. Medical examiners and coroners are another group that has been disconnected from public health but with whom cooperation can enhance and help to calibrate surveillance for pathogen-specific deaths, as illustrated by some examples both domestically (113) and abroad (114).

As we described above, new data sources become useful in proportion to our understanding of their “normal” behavior. As we expand the range of data types, it will be essential to monitor new data streams and continue to monitor old ones outside of epidemic periods to establish a baseline that can be used to calibrate signals of new outbreaks and estimate the exceedance caused by the ongoing transmission of novel pathogens (115).

Conclusion

Data and modeling needs change over the course of a pandemic and vary by the jurisdictional dimensions, requiring anticipatory, rapid, dynamic, and locally adapted and scaled activities to optimize pandemic management and population health. Here, we have sought to describe concepts, tools, and strategies to address those needs, building on those enacted during the COVID-19 pandemic and those that could have facilitated this work. While not a comprehensive list, we hope that the ideas we propose and envision serve as a useful resource and guide in efforts to manage ongoing infectious diseases challenges and preparedness for the inevitable next pandemic.

Author contributions

ML: Conceptualization, Methodology, Project administration, Writing – original draft, Writing – review & editing. MB: Investigation, Writing – review & editing. JB: Investigation, Writing – review & editing. PE: Investigation, Writing – review & editing. DE: Investigation, Writing – review & editing. MG: Investigation, Writing – review & editing. JH: Investigation, Writing – review & editing. MJ: Investigation, Writing – review & editing. SK: Investigation, Writing – review & editing. DL: Investigation, Writing – review & editing. JeL: Investigation, Writing – review & editing. JuL: Investigation, Writing – review & editing. RL: Investigation, Writing – review & editing. DM: Investigation, Writing – review & editing. LCM: Investigation, Writing – review & editing. CM: Investigation, Writing – review & editing. LAM: Investigation, Writing – review & editing. SO: Investigation, Writing – review & editing. CQ: Investigation, Writing – review & editing. AB: Investigation, Writing – review & editing. NR: Investigation, Writing – review & editing. SR: Investigation, Writing – review & editing. RR: Investigation, Writing – review & editing. MS: Investigation, Writing – review & editing. RS: Investigation, Writing – review & editing. RBS: Investigation, Writing – review & editing. DS: Investigation, Writing – review & editing. ST: Investigation, Writing – review & editing. JV: Investigation, Writing – review & editing. YG: Funding acquisition, Investigation, Project administration, Writing – original draft, Writing – review & editing, Conceptualization.

Funding Statement

The author(s) declare that financial support was received for the research, authorship, and/or publication of this article. This project has been funded (in part) by contract 200–2016-91779 with the Centers for Disease Control and Prevention (CDC).

Footnotes

References