Megaregions and COVID-19: a call for an innovative governing structure in the United States

ABSTRACT In the United States, megaregions, which are networks of urban centres and their hinterlands, have a greater prevalence of COVID-19, but have been overlooked as a geographical unit for multi-jurisdictional governance for pandemic response. Existing multi-metropolitan planning organization (MPO) collaborations and state-level COVID-19 coalitions in the United States demonstrate the utility of the megaregion as an effective framework for regional collaboration and operations. Using multilevel modelling, we explore the significance of megaregions in explaining disease occurrence. The results suggest that the megaregion can be an appropriate geographical scale for multi-jurisdictional operations, governance and pandemic response.


INTRODUCTION
Globalization and the advancement of technology have produced hyper-connectivity at diverse scales of geography.According to the DHL Global Connected Index report (Altman & Bastian, 2019), global connectedness has been increasing since 2001 in terms of trade, capital, information and people connections.Increasing connectivity is not restricted to the global level.The number of intraregional, international visitations increased from 0.6 billion in 2006 to 1.1 billion in 2018 (United States Travel Association, 2019).The growing interdependence and hyper-connectivity between different geographies bring disadvantages and benefits, as connectivity 'increases the likelihood of inherently unquantifiable extreme events, such as financial crises, a nuclear holocaust, hostile artificial intelligence, global warming, destructive biotechnology, and pandemics' (Derviş & Strauss, 2020).Others have linked the ability of the megaregion to promote collaboration across various levels of government and spatial geographies to provide an appropriate regional structure to meet challenges in the globalized world.The COVID-19 pandemic is one such example that has been accelerated and intensified with hyper-connectivity, illustrating the need for expansive regional geographies such as megaregions to respond to and contain such threats.Previous case study research (Rouse & Ross, 2019) has shown that multi-jurisdictional collaboration at the megaregional scale enables several efficiencies such as strategic resource-sharing, readiness to handle emerging threats, preventing redundancies in investment decisions and presenting a unified voice to seek resources at the federal level.These noted advantages form the impetus for us to investigate the contribution of the megaregion in explaining COVID-19 occurrence via an empirical framework and advocating for it as an appropriate governance model for pandemic management.
In a recent report, the Organization for Economic Cooperation and Development (OECD) stated 'current issues' like climate change, viruses, and international competition need an agile governance framework able to transcend traditional views of governance and geography, making planning at the megaregional scale increasingly important (OECD, 2021).The report delineates the process for creating a megaregion and references the need for expanded planning capacity and authority across local, national and regional boundaries.Adler et al. (2020) investigate the role of the megaregion in the transmission of COVID-19 from potentially higher exposure risk.Their findings show that once the virus emerged, it spread much faster in megaregions with the highest number of deaths in comparison with counties not located in megaregions.They call for more empirical studies since pandemics could be added to the list of urban externalities that occur as a result of living in megaregions.Megaregional governance could mount a more coordinated response to the appropriate scale at which transmission occurs.Yaro et al. (2022) call for the development of megaregion governance with large-scale investments.They assert the capacity of megaregions to meet the challenges of climate change, traffic congestion and growing inequality while enhancing economic opportunities, quality of life and service delivery.The authors caution that without these investments we risk these geographies becoming centres of poverty, growing inequality, limited services and diminishing housing opportunities.Khanna (2016, p. 15) asserts: Mega-infrastructures overcome the hurdles of both natural and political geography, and mapping them reveals that the era of organizing the world according to political space (how we legally subdivide the globe) is giving way to organizing it according to functional space (how we actually use it).In this new era, the de jure world of political borders is giving way to the de facto world of functional connections.
With the emergence of COVID-19, emerging technologies, work from home and changing commute patterns, the megaregion becomes increasingly important to accommodate related flexibility in governance structures.This paper addresses the following primary research questions: . Do regional scale frameworks, for example, states or megaregions, provide substantial insight into understanding or responding to the pandemic? .Why do we need regional governance in response to pandemics?
We investigate the prevalence and transmission patterns of COVID-19 through the lens of the megaregion structure in comparison with existing county and state geographies.A primary focus is to employ an empirical approach to explore potential contributions regarding disease progression under these different spatial geographies and offer an alternate spatial governance scale that might promote more coordinated effort towards pandemic management.
The remainder of the paper is structured as follows.The literature review discusses multi-jurisdictional collaborations, megaregions, COVID-19 and research contributions.The materials and methods section discusses the use of multilevel linear modelling to examine comparative contributions of state and megaregion boundaries to explaining COVID-19 case rates.The results interpret the findings from the models.The discussion section focuses on outlining the foundation for use of the megaregion as a governance framework for disease management and response.Lastly, the conclusions and policy implications of the findings are discussed.

State of regional collaboration in the United States
In the United States, regional planning is conducted at the state, metropolitan and city levels.Metropolitan planning organizations (MPOs) can be created by agreement between its state governor and local governments for any urbanized areas with a population greater than 50,000 (Sciara, 2017).Since the purpose of creating MPOs was to promote regional transportation planning, their roles are limited and do not provide other services unless they are also designated as regional commissions or councils (Angel & Blei, 2021;Sciara, 2017).
Newer forms of regional, multi-jurisdictional planning are emerging including multi-MPO planning.Transportation issues, environmental issues and economic issues are the three main categories of regional concerns motivating neighbouring jurisdictions to form partnerships (state or MPO) (Morley et al., 2019).Multi-MPO planning is encouraged and not mandated by the federal government, and collaboration agreements are not necessarily legally binding.Therefore, interregional issues, shared interests and benefits, and compatible perspectives are preconditions for sustaining multi-MPO collaborations in the long term.
Given that these collaborations are voluntary, what then are the benefits of multi-jurisdictional collaborations?According to the US Department of Transportation's (USDOT) (2017) multi-MPO guidebook, multi-jurisdictional planning, that is, multi-MPO planning, can be beneficial to several stakeholders.State departments of transportation (DOTs) can be less burdened from resolving conflicts among MPOs or coordinating with individual MPOs, as multi-MPO planning requires mutual understanding and consent among MPOs as a precondition.For MPOs, it is an opportunity to address issues that transcend their planning boundaries and respond collectively to those issues more efficiently.Local communities such as cities and neighbourhoods can directly benefit from the outcomes of these collaborations, including improved accessibility, environmental quality and economic opportunities.Several successful cases are introduced in the guidebook, including the San Joaquin Valley Regional Policy Council (SJVRPC) and New York Metropolitan Area Planning Forum (MAP Forum).The SJVRPC coordination area comprises eight MPOs in the southern half of California's Central Valley.The area has regional air-quality issues, goods movement and growth management challenges that necessitate a collective response from member MPOs to collaboratively design transportation model improvements and community engagement efforts to address these issues.By acting collaboratively, the member MPOs could extend their collective capacity to overcome the individual limitations of lower-resourced MPOs compared with 'the more populous, affluent, and heavily urbanized MPOs in their state' (Morley et al., 2019, p. 31).The MAP Forum is another example of multi-state and multi-MPO collaboration of 10 MPOs in the states of New York, New Jersey, Connecticut and Pennsylvania.The members work on a consensus that each MPO's plans and programmes can affect other member MPOs as they are intimately related by commuting patterns, transportation infrastructure and economic relationships which motivate strong collaboration.By participating in the MAP Forum, the member MPOs benefit from sharing experience, enhanced capacity from being supported by higher capacity MPOs and more opportunities from information-sharing.
These examples suggest that multi-MPO planning has been motivated by the need and benefits of collaboration shared among MPOs that have common regional issues.In some instances, the area and scale of collaborations can cross state boundaries, having mega-scale population and area.For example, the MAP Forum coordination area has a population of over 20 million and over 10,000 square miles (about the area of New Jersey).From this perspective, megaregion-level governance without formal governance structures is already being practiced motivated by need, suggesting that the megaregion framework is a natural extension of multi-jurisdictional collaboration to respond to new regional issues.The Federal Highway Administration (FHWA) recognizes the value of regional models of cooperation and defines 13 megaregions (Figure 1).Liu and Arnosti (2018) outline significant advantages to collaborations across cities and regions particularly regarding shared problems, for example, their reference to the City of San Francisco and its lack of affordable housing.City officials realized that housing needs cannot be met in the city alone and they rely on collaborations with adjacent jurisdictions (Liu & Arnosti, 2018).A well-functioning health system is increasingly reliant on multi-jurisdictional governance.Sagan et al. assert that 'Governance is also increasingly conducted across levels, from local to global, with multilevel governance becoming increasingly important' (Sagan et al., 2021, p. 11).They further suggest that governance has been a primary factor in good responses to COVID-19.Collaboration across regions allows different areas to solve problems and deficiencies that they could not solve or address on their own.

Megaregions and COVID-19
Megaregions are urban agglomerations linked through infrastructure systems, intensive activities, large populations, economic and industrial production, technology hubs, and the movement of people and goods (Dewar & Epstein, 2007;Nelson, 2017;Ross & Woo, 2009).They mirror the economic geography of the United States composed of a network of regional and metropolitan economies, producing a majority of employment opportunities and economic output (Cisneros et al., 2021, p. 4).According to the USDOT (2017), more than 70% of the US population, employment and related economic activity are forecast to be concentrated within 13 existing megaregions, despite occupying only 28% of land area.Table 1 presents selected characteristics of megaregions.In addition, they are connected by airports and interstates, with more than 85% of air passengers using airports in megaregions.Among the top 50 busiest airports in the country, 44 are in megaregions.
Reflecting on the future growth of megaregions, Barnett (2020) outlines how the redesign of regions can be driven by development of megaregional governance focused on a few primary initiatives including adapting development suitable to the environment, reorganizing the transportation infrastructure and developing policies that target inequality.Megaregions present an appropriate geospatial structure in response to the increasing awareness that many systems and their impacts, from transportation networks to watersheds, transcend metropolitan area boundaries and require synchronized, multi-jurisdictional solutions (Harrison & Hoyler, 2015;Moeckel, 2017;Mollanejad et al., 2015;Ross et al., 2016).
COVID-19 is a type of emerging infection where 'many types of host heterogeneity influence host/pathogen interactions at the scale of the individual, such as genetics, age, sexual activity, location, and typical movement patterns' (Riley, 2007(Riley, , p. 1298)).The literature on infectious disease also identifies several relevant factors of transmissibility, including an individual's health, socio-economic factors, spatial factors and behavioural factors.
The COVID-19 pandemic has generated renewed interest in the direct influence of the social determinants of health (SDOH) on infectious disease transmission, severity and recovery, in addition to the indirect risk for severe COVID-19 outcomes from chronic disease comorbidities (Abrams & Szefler, 2020;Baidal et al., 2020;Nayak et al., 2020;Rollston & Galea, 2020).Low socioeconomic status (SES) neighbourhoods often lack access to essential resources and services, for example, healthy foods, educational and employment opportunities, transportation options, and adequate and timely healthcare (Bambra et al., 2020;Emeruwa et al., 2020;Lima et al., 2020;Singu et al., 2020).Low SES populations are often employed in low (hourly) wage jobs, which preclude working from home and the ability to follow social distancing recommendations.Often, these coincide with transitdependent populations, cumulatively increasing the risk of contracting COVID-19 due to increased exposure (Rollston & Galea, 2020).Social indicators tend to cluster in geographical locations which do not necessarily conform to political boundaries (Messer et al., 2006) and health disparities closely follow socio-economic gradients since they are associated with modifiable social disadvantages (risk factors).The impact of factors outside of the health sector on COVID-19 outcomes necessitates a coordinated response from multiple agencies and spatial scales of government.
Since COVID-19 can be transmitted from human to human, factors such as population density that can influence in-person interaction are relevant (Acuto, 2020, p. 317).However, findings on the relationship between COVID-19 transmission and population density are mixed, as some studies have found no relationship (Carozzi et al., 2020;Hamidi et al., 2020), whereas others report a positive relationship (Wheaton & Kinsella Thompson, 2020).Similarly, public transit usage and other factors became significant factors of transmission, particularly after state government-imposed restrictions, for example, shelter-in-place, social distancing (Musselwhite et al., 2020;Zheng et al., 2020).
From the literature, we can infer that megaregions require more attention in understanding the pattern of infectious diseases such as COVID-19.Adler et al. (2020) suggest that megaregions are more likely to be infected earlier and are more vulnerable in the early stage of pandemic events.Adda (2016) finds that economic activities are closely related to the spread of infectious diseases.In an extensive review of the relationship between urbanization and infectious diseases, Connolly et al. (2021) identify three criteria that most affect disease spread: governance, socio-demographic changes and infrastructure.Houston and Ruming (2014) suggest the greater connectivity between urban, suburban or peri-urban regions have marshalled in a need for fresh approaches to use the benefits of increased connectivity.Major cities and urban areas constitute the core of megaregions.There is a greater risk of disease transmission from 'the heterogeneity in the health of urban dwellers, increased rates of contact, and mobility of people' (Alirol et al., 2011, p. 131).Moreover, hyper-connectivity, population, the abundance of connected transport infrastructure (Figure 1) and resulting mobility patterns may expose residents of megaregions, as diseases 'can spread rapidly between cities through infrastructures, e.g., global air networks' (Connolly et al., 2021, p. 10).For example, the severe acute respiratory syndrome (SARS) epidemic and the 2009 H1N1 (Swine flu) pandemic demonstrated that air travel was a significant factor for transmitting disease, especially on the worldwide scale (Tizzoni et al., 2014).These characteristics of megaregions imply that megaregions can be an appropriate framework for explaining the spread of and managing the COVID-19 pandemic.

Regional responses to COVID-19
From a research perspective, Agostinis and Parthenay (2021) note that while there is an abundance of literature that discusses global health governance, regions located between the global and local scales remain largely underinvestigated.In addition, their potential to create shared policy to tackle multi-jurisdictional health challenges has been underused.From a governance perspective, some of  the main critiques around health systems and COVID-19 response include siloed preparedness and response efforts, lack of health leadership as well as poor systems integration to reflect more contemporary public health approaches such as 'health in all policies' (World Health Organization (WHO), 2021).Ansell et al. (2021, p. 950) characterize the COVID-19 pandemic as a 'complex and turbulent problem' akin to 'climate chaos, globalized terrorism, the US opioid crisis and huge oil spills' which require multi-jurisdictional collaboration coupled with robust and innovative governance structures that perform significantly better than existing, archaic bureaucracies.Recognized characteristics of effective global and domestic governance responses to managing the pandemic include collaborative efforts among a diverse set of actors that went beyond the traditional public health sector and a coordinated response between national, local and other intermediate levels of government as well as multiple stakeholders (Choi, 2020;Grizzle et al., 2020;Huang, 2020).In a critique of the US approach to pandemic response, Kettl (2020) describes the fractured governance landscape where the federal government deferred most decisionmaking to the states.However, state governments were divided and inconsistent in their governance approaches undergirded by their political differences, embracing of national health and social policy, and differential ability to compete for scarce medical supplies.
On the contrary, currently functional regional governance structures (multi-MPO and megaregions) have tremendous potential to support such coordinated efforts by bridging the gap between federal and state response, since they represent diverse stakeholders mobilized around common objectives and already comply with collaborative governance principles.Regional collaboration to mitigate the pandemic and ensuing disruptions to global value chains are transferable concepts to support interjurisdictional collaboration within the United States where there was a heavy reliance on health-supporting resources (personal protective equipment (PPE), hospital beds) across state borders.While it is too early to assess their effectiveness, existing multi-state pacts such as the Eastern States Multi-State Council, the Western States Pact and the Midwest Regional Partnership demonstrate that COVID-19 has already been a major driver in the formation of multi-jurisdictional governance structures for pandemic response.These collaborations build on existing economic and infrastructure connectivity.Activities have included establishing regional supply chains to manufacture and distribute medical equipment, joint lobbying for financial resources from congress and strategic plans for reopening their economies.Megaregions provide a more refined delineation of interconnected geographies to promote resource-sharing and governance-related policy to prevent and manage pandemics.

Research significance
COVID-19 has significantly impacted life in the United States, and a plethora of research has brought insight into the transmission mechanism of COVID-19, its relationship with the built environment (Carozzi et al., 2020;Hamidi et al., 2020;Wheaton & Kinsella Thompson, 2020;Yoo & Ross, 2021) and human behaviour (Chernozhukov et al., 2021;Zheng et al., 2020).Nevertheless, examining the COVID-19 pandemic within a regional context, especially with the megaregion, has rarely been undertaken.Most of the studies that focus on a regional perspective examined effectiveness of regional level policy interventions.For instance, Bourdin et al. (2021) analysed the effectiveness of lockdown strategy in Italy by using the local Moran's I and spatial autoregressive model.The results highlighted the importance of spatial contiguity in the spread of the disease.However, the study did not consider multilevel (local and regional) effects.In the United States, Berry et al. (2021) examined whether the state governments' shelter-in-place (SIP) orders by analysing county-level mobility data and found no significant impact of SIP orders.Similarly, Feyman et al. (2020) analysed mobility data and found varying effectiveness of SIP orders among states.These studies focused on the effectiveness of SIP orders rather than the necessity of regional-scale policy interventions.
Some exceptional studies called for a new scale of regional framework including Adler et al. (2020) and Angel and Blei (2021).Using analysis of covariance (ANCOVA) models, Adler et al. (2020) analysed 11 megaregions and found that megaregions had higher COVID-19 mortality at the early stage of the pandemic.It is the only study that explicitly applies the concept of megaregion in investigating COVID-19.Nevertheless, the authors offered few recommendations for enhancing multi-jurisdictional governance.Angel and Blei (2021) analysed data on county attributes that are associated with COVID-19 and cases and deaths in US counties and metropolitan statistical areas (MSAs) to suggest MSA as an appropriate spatial framework for responding to COVID-19.The study found that COVID-19 cases and deaths in MSAs exceeded the number of cases and deaths in their representation in the general population.It also found that the number of cases and deaths in MSAs were more highly correlated to the levels in their adjacent counties than was the case for smaller areas and rural counties.The study highlights the importance of MSAs in understanding the pandemic and urges to expand the role of existing MSAs to embrace pandemic response.Nevertheless, the study is primarily descriptive as it categorized counties into four groupsall, multicounty MSAs, small cities and rural countiesand compared the difference of average number of cases and deaths.
In summary, the existing literature mostly considers only one level of geography (e.g., Adler et al., 2020;Angel & Blei, 2021;Hamidi et al., 2020;Yoo & Ross, 2021), such as county or state, overlooking the hierarchical nature of different scales of geographical units (e.g., counties are nested in states).However, there can be a significant impact from higher levels of geographical units (context) on outcome variables at lower levels of geography.Many studies also examine the effectiveness of current governance frameworks rather than articulate new ones (e.g., Berry et al., 2021;Bourdin et al., 2021;Feyman et al., 2020).Our study incorporates a multilevel approach to empirically test the value of megaregion boundaries on explaining COVID-19 case rates.Our approach bridges interdisciplinary methods from regional planning and social epidemiology by using multilevel modelling to study the influence of place (megaregion) on health and statistically account for spatial nesting.

MATERIALS AND METHODS
It is essential to clarify the term 'megaregions' used in this study.Our study focuses on megaregions in the United States.As Harrison and Hoyler (2015) point out, there are different conceptualizations with divergent emphases on large-scale urban-regional configurations.For instance, there is a significant difference between North American and European perspectives on megaregions, as the former is on spatial form whereas the latter is on functional aspects.They further argued that these different conceptualizations are spatially different that can be categorized into three types, including one urban system, multiple interrelated urban systems, and the global urban system across all geographical space.The differences in conceptualizations, emphases, and delineations of megaregions exist even among studies within the United States (e.g., Lang & Dhavale, 2005; Regional Plan Association (RPA), 2006; Ross et al., 2009).
In accordance with our geographical focus, the megaregion in this study is explicitly the one conceptualized in the United States for planning mega-scale interrelated urban systems.Specifically, this study uses the FHWA's megaregions that defines megaregion as 'networks of urban centers and their surrounding areas, connected by existing environmental, economic, cultural, and infrastructure relationships'.As the only federal agency that officially uses the concept of megaregions, the FHWA identifies and recognizes 13 megaregions for transportation planning.Using the already defined and operationalized megaregions, our study attempts to explain the importance of a regional-scale governance framework in the light of the COVID-19 pandemic.
We explore the applicability of regional scale governance in enhancing understanding and response to the COVID-19 pandemic using the megaregion framework.We also investigate whether a megaregion's significance varies by time.We apply multilevel linear modelling (MLM)an approach used in social epidemiologyto examine whether megaregions contributions are significant in explaining disease prevalence (COVID-19 case rates) after controlling for other county-level factors.We construct similar models using state boundaries and compare the results with the megaregion models.The comparative analysis demonstrates the relative importance of these regional boundaries (used as a proxy for the scale of governance) in explaining county-level COVID-19 case rates.Social epidemiologists extend the statistical concept of clustering/nesting to understand the impact of environmental factors (place) on health (Merlo et al., 2005a;Subramanian et al., 2003).In our dataset, counties are nested within larger geographical units (states, megaregions).MLMs are beneficial since they allow us to partition the variance to examine how much of the variation in COVID-19 case rates can be attributed to the counties (local), and how much can be attributed to the Megaregion (region).Furthermore, by applying MLMs, we can avoid oversimplification of COVID-19 prevalence by considering the regional differences, which can be captured with larger geographical unit variables.
Our models test if the megaregion has a significant and independent influence on the occurrence of COVID-19 cases over and above other county-level predictors.The MLM in this study incorporated two levels of datalevel 1 contains the county-level variables, and level 2 is the megaregion.Referring to Diez and Aiello (2005), the model can be specified as follows: where Y ij is COVID-19 case rates (dependent variable) for the ith county in the jth megaregion; B 0j is the group (megaregion or state)-specific intercept; B 1j is the group (megaregion or state)-specific effect of the county level variable; I ij is the county level variable for the ith county in the jth group (megaregion or state); and ε ij is the county level error.
We fit a random intercept, fixed slope model to examine the impact of the megaregion coefficients on COVID-19 case rates.Through this model, every megaregion group has an intercept that can vary, but the slopes of the county-level coefficients stay constant across all groups.Essentially, the model tests if factors at the megaregion level either increase, decrease or have no influence on county-level case rates as measured by the intercept.We used R package 'lme4' for MLM on confirmed case rates per 100,000 population with county-level predictors while controlling for clustering at the megaregion level.
In interpreting MLM, we particularly focus on the intraclass correlation coefficient (ICC).ICC is 'the fraction of the total variation in the data that is accounted for by between-group variation' (Musca et al., 2011, p. 2), which indicates the extent to which clustering at the megaregion scale explains the COVID-19 cases standardized by population.Through analysis of the ICC values, we can infer whether megaregion boundaries can be a meaningful spatial framework in understanding the pattern of the pandemic.ICC can be calculated from the following equation: where t 2 is between-group variance (between megaregions or states); and σ 2 is within-group variance (within megaregions or states).
In addition, we examined daily variance of ICC values acquired from fitting MLMs every day for two reasons.First, the COVID-19 case rate is sensitive to time and can differ based on the dates selected.Thus, selecting a particular date can bias the results.Second, the policy interventions from state governments started and stopped at different times.Therefore, creating comparative models for a series of days reduces the bias that could result from picking an arbitrary date to compare models.The assumption here is that differential timing and duration of policy interventions may have resulted in significant variation in COVID-19 case rates.For these reasons, we analyse and compare variance on a daily basis rather than a particular date's ICC to explore whether megaregions' significance varies over time compared with state boundaries.
We also fit MLMs using the state boundary and compare their ICC values with the ICC values from the models using the megaregion boundary.MLMs were fit from 22 March 2020, which is the first day when more than 30% of the counties have at least one confirmed case, to 22 March 2022.For each day, we ran a series of three models with increasing complexity to check if there is persistent variation in COVID-19 case rates that is explained by the megaregion boundary even after adding county-level variables.Model 1 is a null model with no covariates.Model 2 has level 2 (megaregion or state) random intercepts but no county-level variables.Model 3 is a full model with the level 2 random intercepts and countylevel variables.
Table 2 contains the list of outcome and independent variables, the data sources, and descriptive statistics.We constructed a national dataset of 3108 contiguous US counties combining risk factors, assets, the strength of COVID-19 restrictions and other relevant control variables.As the dependent variable, we calculated COVID-19 case rates per 100,000 population using confirmed case counts from Johns Hopkins University Center for Systems Science and Engineering (JHU CSSE).The data provide daily cumulative confirmed cases.Our model's relevant independent and control variables were identified through a literature review of predictors used in other studies to model risk for COVID-19 morbidity.
We obtained county-level data for health, socio-economics, behaviour, environment and healthcare access from the 2020 County Health Rankings dataset.These capture the breadth of the SDOH variables used to model COVID-related outcomes.Based on theoretical relationships and underlying statistical correlations, we combined variables into domain-specific indices to enable effective data reduction and avoid multi-collinearity issues in our model (see Appendix A in the supplemental data online).As a result, we created a health index (HI), behaviour index (BI), medical index (MI) and socio-economic index (SI).Higher values on these indices indicate that the county has poor health outcomes, higher rates of unhealthy behaviours, poor access to healthcare, and low socio-economic status.Other control variables such as population density, transit population, rurality, and employment characteristics are from the American Community Survey (ACS) 2018 five-year estimates.
In addition, we created a restriction index (RI) to represent the strength of restrictions passed by each state and assigned that value to each county in the same state because restrictions on population movement can have a significant impact on the transmission of COVID-19.For this index, we used data from the Institute of Health Metrics and Evaluation (IHME) (2020) that documented the different containment and closure policies imposed in each state, and the dates they were implemented and began to be eased.We also included data on mask mandates from The American Association of Retired Persons (AARP) website.Seven policies were compiled into one index using the methodology used by the Oxford COVID-19 Government Response Tracker (n.d.) (see Appendix A in the supplemental data online).We also calculated an additional metric called Duration, which counts the number of days from the first restriction goes into place to when the first restriction eases.Figure 2 shows the different values of the RI for each state where the red values indicate elevated levels of restrictions while the blue values indicate low levels of restriction.Specifically, we note that while restrictions were implemented at the state level, the RI has only five levels and tend to cluster spatially (neighbouring states have similar RI scores in some regions), providing the basis for a hypothesis that regional policy coordination might have an influence on disease transmission, incidence, and prevalence.MLM was thus a good fit for this analysis as it allowed us to account for the hierarchical nature of counties nested within Megaregions and states.As noted earlier, we fit a random intercept, fixed slope model due to the limited availability of sample-size and policy data at county level.
In the future, we will explore random slopes for the county-level variables based on sample-size availability and refinements to our RI as more policy data becomes available for counties.

RESULTS
The modelling framework tests the impact of location within megaregions on COVID-19 case rates.We fit multilevel models using megaregion boundaries and state boundaries, respectively, and added county-level covariates.Irrespective of dates, full models that include county-level covariates and megaregion variables have a better model fit than null models or megaregion level intercept-only models as fit statistics including -2 log likelihood, Akaike information criterion corrected for a finite sample (AICc), and Bayesian information criterion (BIC) reduce in magnitude.Since there are 731 full models that have both megaregion random effect and county-level fixed effects, we present the results of two significant days which are the first day of the analysis (Table 3) and 1 July 2020 (Table 4) which is the first day when the state model's ICC becomes greater than the megaregion model's ICC.On 22 March 2020, the ICC value of the megaregion boundary in full model (model 3) is 0.19, showing that the megaregions explain 19% of the variation in the COVID-19 case rates after controlling for county-level covariates (Table 3).As anticipated, the SI has a significant and positive relationship with the case rates, making the case for improving the SDOH (education, employment, housing conditions).On the other hand, the BI has a negative relationship with the case rates which is counter to what we expected.The proportion of population using transit and population density also have significant and positive relationships with the case rates as expected.However, the RI and the Duration variable are not significant because at that point in time, only a few states imposed restrictions and the duration of restrictions was too short to yield any impact.
On 1 July 2020, the ICC value of the megaregion boundary in full model (model 3) decreases to 0.13, suggesting that the megaregions explain 13% of the variation in the COVID-19 case rates after controlling for county-level covariates (Table 4).Although the ICC value is not high in magnitude, there is a potential indication that clustering at the megaregion level does have a statistically significant association with COVID-19 rates at the county-level, over and above county-level fixed effects.The significance of this finding is consonant with the social epidemiology literature where a non-zero intraclass correlation is considered an imperative to further investigate the combined impact of upper-level units (neighbourhoods, counties, states, etc.) and lower-level units (individuals, etc.) on health outcomes (Merlo et al., 2005a(Merlo et al., , 2005b(Merlo et al., , 2018)).While the SI still has a significant and positive relationship with the dependent variable, several variables become insignificant including the HI, BI, population density and proportion of the population using transit.Interestingly, the RI and duration become significant.The RI now has a negative relationship with the case rates, indicating that stronger restriction policies are associated with lower case rates.On the other hand, the duration has a positive relationship with the case rates.This implies that in the states where the COVID was severe have a longer duration of restrictions.In fact, several states postponed their reopening plans amid the rise of COVID cases including California, Michigan and Indiana.
Temporal change of ICC values from the megaregion models and state models are also examined.Figure 3 shows the longitudinal change of ICC values of megaregion and state boundaries.It suggests that the megaregion  values up until 1 July 2020, after which they start to diverge.This pattern indicates that the megaregion boundary has better explanatory capacity compared with the state boundary during this period.The difference in ICCs between the two boundaries becomes greater on 24 April and 1 May 2020.The pattern continues until 1 July 2020, then the pattern is reversed as the state boundary has greater ICC values.At the beginning of the analysis period, the megaregion models had a greater ICC value, suggesting a better explanation of COVID-19 case rates.
At the end of our analysis period, the state models have greater ICC values than the megaregion models.

DISCUSSION
Recently, several researchers have examined the relationship between infectious disease and urban enclaves.In a study of the vulnerability of vast geographies, for example, suburbs, satellite communities, peri-urbanization to infectious diseases, Connolly et al. (2021, p. 1) conclude that these places 'may result in increased vulnerability to infectious disease'.Litman (2020, p. 15) asserts, 'risks vary by location: urban residents have more exposure to infectious disease while rural residents are more likely to die if infected'; and Hsu (2020) makes the argument that infection rates increase with crowding not density.We employed an empirical approach to analyse the relationship between the spread of COVID-19 within the context of the megaregions.This investigation creates a premise for megaregion pandemic management.The analysis supports the efficacy and values of the megaregion as both a governance and operations structure to potentially reduce the impacts of COVID-19 by revealing the following: . The multilevel investigation suggests megaregions are associated with COVID-19 case rates independently, above county effects. .The temporal change of the megaregion boundary's ICC values shows that megaregions could explain more variance of COVID-19 case rates, at least at the beginning of the pandemic before policy actions from states become effective.
The temporal change of the megaregion and state boundary's ICC values suggests there is a need for a preemptive megaregion governance framework and megaregion-level policy interventions in response to the pandemic, particularly in the early stages of the pandemic.In the beginning of the pandemic when there was little or no responses from state governments, the megaregion boundary had higher ICC values than did the state boundary.Results show that ICC values of the state boundary exceeded the values of megaregion models from July 2020 to March 2022.As the pandemic responses deployed by state governments, including testing, social distancing restrictions, mask mandates, and vaccination.became effective, the pattern of COVID infections started to conform within state boundaries.
Megaregions emerge as places with a great propensity for crowding with a density of population.These expanded regions are forecast to continue to contain the US increase in population.Moreover, they will remain as centres of economy and employment.With substantial amounts of employment, extensive infrastructure, and financial resources, megaregions are well-established with comprehensive collaborative planning capacity facilitated through MPOs.In a recent study, Morley et al. (2019) found that predominant themes motivating multi-MPO collaboration included highway investments, freight movement, transit services, transportation safety, congestion movement, transformative technologies, air quality, land and water resource management, extreme weather resilience, economic development, and affordable housing policy.Multi-MPO coordination platforms such as the San Joaquin Valley Regional Policy Council (SJVRPC), New York Metropolitan Area Planning Forum (MAP Forum) and Southeast Florida Transportation Council represent exemplary case studies of existing multi-jurisdictional coordination efforts that are actively collaborating on regional issues through formal and informal mechanisms.Practices that sustain collaboration include the pooling of resources, larger MPOs (anchor members) supporting smaller MPOs, and joint advocacy and collaboration events.Collaboration efforts even extend to producing coordinated policy and planning documents responsive to regional needs outside jurisdictional boundaries.
Multi-state pacts described above are similar in structure to megaregions and examples of the benefits of collaboration operating under a unified regional approach.The report Roadmap to Recovery: A public health guide for governors from the National Governors Association suggests governors consider 'regionalizing approaches to recovery by collaborating with neighbouring states' (National Governors Association, 2020, p. 22).Members of the multistate alliances agreed that reopening of their economies and easing the restrictions need to be coordinated with the other members.Such agreement stems from the mutual understanding that their regions are intricately connected with the other members of their alliances both physically and economically.The bipartisan alliance of the Midwest Regional Partnership suggests economic reliance was a strong motivation for the forming of the alliance.In addition to the coordination of easing the restrictions, the Eastern States Multi-State Council and Western States Pact actively attempted to gather resources to combat the pandemic.The Eastern States Multi-State Council planned to develop a regional supply chain for the manufacturing and distribution of essential medical supplies.The Western States Pact also joined together to develop innovative technologies to use in the effort to stop the spread of the pandemic.The member states even garnered political power to acquire necessary financial aid from the Federal government.All attempts show that the multi-state approach, namely the megaregional approach, is already being implemented in practice.These examples illustrate that 'the urban governance of the epidemic as a restructuring process of established Megaregions and COVID-19: a call for an innovative governing structure in the United States 11 REGIONAL STUDIES political hierarchies' (Wolf, 2016, p. 973) epitomized during the SARS crisis in Hong Kong (Roloff, 2007) is effective.Further efficiencies can be driven by the coordination of multi-state pacts and existing multi-MPO coordination platforms to better use communication channels, partnerships, and resources.

CONCLUSIONS AND POLICY IMPLICATIONS
Megaregion planning presents an alternative way of approaching, analysing, and managing many of the challenges confronting interconnected urban areas and cities.
In this paper, we explore how much counties and states contribute to the understanding of COVID-19 compared with multi-scale geographies (megaregions).In the early onset of COVID-19, megaregions contributed more to the understanding of disease intensity and transmission than counties and states.The review of literature, our empirical analysis, and examples of successful multi-regional collaborations suggest the value of developing megaregion administrative structures, policies, and guidance.
There is an increasing focus on the connections between health (pandemics) and coordinated regional planning to mobilize resources and places to better prepare for current and future pandemics.Connectivity across geographical scales is projected to continue, and the need for flexibility, policies targeting pandemics, and new regional planning platforms with the authority to implement and conduct operations across expanded geographies is necessary.Evidence suggests that megaregions and multi-jurisdictional planning structures can be more responsive and effective regarding emergency preparedness, disease prevention, and management of pandemics.Sagan et al. asserted that: Governance plays a critical role in health systems performance, thereby also providing the principal lever for resilience.Constructive deployment of funding and resources relies heavily on governance.Governancethe way decisions are made and implementedenables the financing, resource generation, and service delivery functions to operate as intended and in coordination with the rest of the system to achieve maximum overall system performance, and by extension, resilience.(Sagan et al., 2021, p. 10) Our research establishes the importance of the megaregion geography in explaining health outcomes.In turn, this provides the required evidence to promote new governance structures whether they are entirely new layers of government or just better coordination between existing governments within the megaregions.
Existing collaborations initiated in response to COVID-19 operated at a larger geographical scale to achieve more significant efficiencies, cost savings, greater preparedness, and improved pandemic management.Governmental regulations, policies, and mandates guide current planning practices for counties, states, and regions.However, they do not resonate sufficiently within the global economy.They are not structured and empowered to meet the challenges of climate change, natural disasters, and pandemics in the places where we live.This paper identifies contributions of different geographies and multiple scales in documenting and tracking disease levels and transmission patterns greater than counties can across selected time intervals.These urban enclaves, megaregions, require multi-jurisdictional authority to plan for and mitigate the greater exposure and transmission of disease.
To prepare for future pandemics, we suggest the establishment of an ad hoc governance structure for megaregions to institutionalize emergency response.Currently, the Federal Emergency Management Agency (FEMA) has ten regions in the continental United States and territories to raise risk awareness before disasters, coordinate federal response and apply and manage resources during disasters.However, the pandemic has brought many challenges complicated by legal issues and circumstances of local and state governments, requiring a greater capacity than FEMA has.Establishing governance systems and operations that enhance the response capacity currently confined by a myriad of jurisdictional issues and organizational silos is critical.Governance at the megaregion scale enhances our capacity to manage pandemics and natural disasters, especially when confronting highly infectious diseases.According to Kuebart and Stabler (2020), the literature on medical geography identifies three spatial diffusion processes: spatially contagious diffusion, hierarchical diffusion, and relocation diffusion.Spatially contagious diffusion is a typical spread that occurs evenly over space via everyday networks.Hierarchical diffusion 'occurs along with established patterns such as the centrality of transport networks' (p.486).Relocation diffusion is the spread of newly introduced diseases in new places.Megaregions are vulnerable to all the diffusion processes, particularly for hierarchical diffusion and relocation diffusion, due to their centralities of economy and infrastructure, making an even stronger case for multi-jurisdictional coordination and governance.
Existing literature on behavioural patterns, economic structure, infrastructure, and environment, reflects a need for a larger scale is necessary to be inclusive of today's expanded geography of human activities.Additional research on the delineation of megaregions is needed for different disciplines and spatial configurations.There is also a need for greater understanding of the delineation of megaregions, and additional data and methods to more accurately track and measure disease progression.As more data becomes available, research on identifying megaregions that capture various aspects of human activities will be possible, establishing a firm foundation for innovative governance.

Figure 1 .
Figure 1.Major transport infrastructure in 13 megaregions designated by the Federal Highway Administration (FHWA).Note: Megaregions have all of top 50 busiest airports and are connected by interstate highways, making them 'hyper-connected' domestically and internationally.

Figure 2 .
Figure 2. Map of the restriction index by state.Note: The higher the value of the index (red), the larger the number of restrictions imposed.

Figure 3 .
Figure 3. Temporal change of intraclass correlation coefficient (ICC) values of the megaregion boundary and the state boundary.Note: The red shaded area indicates the term when the megaregion boundary's ICC values were higher than the state boundary's values.

Table 1 .
Key statistics of the 13 megaregions defined by the Federal Highway Administration (FHWA) in the United States.

Table 2 .
Variables and data sources.

Table 3 .
Result of multilevel analysis (22 March 2020).Megaregions and COVID-19: a call for an innovative governing structure in the United States 9 boundaries can explain COVID-19 case rates better than the state boundaries at the beginning of the pandemic.On 22 March 2020, the ICC values associated with the state and megaregion boundaries are both 0.19.Subsequently, the ICC value of the megaregion boundary becomes slightly higher than the state models.In the 17 April 2020, model, the ICC value of the megaregion boundary reaches the peak of 0.32 indicating that it explains 32% of the variance and the state boundary also reaches its first peak having the ICC value of 0.31.However, the decrease of the ICC values of the state boundary is greater than the decrease of the ICC values of megaregion boundary.Megaregion boundary maintains greater ICC values from the peak compared with state's ICC