In this mathematical modelling study, we have assessed the potential impact of dynamic community-based NPIs, involving sixteen economically diverse countries, as a pragmatic strategy for controlling the COVID-19 pandemic in order to provide a practical illustration of interventions and strategies implemented to reduce the reproduction rate of COVID-19. Our study has several inter-related findings. First, we show that simultaneous cycles of 50-day mitigation (R value of 0.8) followed by a 30-day relaxation could provide means to reduce the effective reproduction number, however, will be insufficient to keep the number of patients requiring ICU care within manageable levels. Second, by contrast, we found that dynamic cycles of 50-day suppression (R value of 0.5) followed by a 30-day relaxation would be required, for all countries, to keep ICU demands below the national capacities. Third, significant number of new infections and deaths could be prevented if these “rolling” suppression measures can be maintained for an 18-month period, or until a suitable treatment and/or vaccination become available. Finally, a continuous, yearlong suppression strategy may also reduce overall attack rates significantly and appears effective. However, implementation (and socioeconomic sustenance) of such stringent measure could be challenged by its detrimental impacts on population well-being and livelihood.
Our findings may have several explanations. First, despite higher rates of contact across older age groups , we predict a somewhat lower incidence of ICU hospitalisation and deaths in low-income settings. This can be explained, at least partly, by the demographic differences with a relatively younger average age structure of these populations, and absence of integrated death registration system. However, given the significant inequalities in baseline health, testing capabilities and critical care infrastructure across the countries, in reality, a higher overall level of excess deaths are likely in resource-poor settings owing to health systems failure, especially in uncontrolled or mitigation intervention scenarios. Second, it was unsurprising that a more restrictive suppression strategy (R: 0.5) in our study reduced ICU hospitalisations and deaths for all countries. This is because a further reduction in the reproductive number secondary to more stringent interventions can maximally reduce the population transmissibility of the SARS-CoV-2 . Notably, implementation of such strategies also creates a policy dilemma for many low-income countries: how to address the “competing priorities” of preventing COVID-19 associated deaths and public health system failure with the long-term economic collapse and hardship. In this regard, we have observed that in contrast to a long fixed-duration social distancing, dynamic NPIs (that reduce the overall attack rates effectively) may offer a helpful balance.
Third, in our study, dynamic cycles of 50-day suppression followed by a 30-day relaxation were effective to lower the deaths significantly for all countries since both transmissibility and case severity (and by extension, critical care demands) were significantly reduced throughout the 18-month period. Notably, this intermittent combination of strict social distancing, and a relatively relaxed period (with efficient testing, case isolation, contact-tracing and shielding of the vulnerable), may allow populations and the national economies to “breathe” at intervals—a potential that might make this solution more sustainable, especially in resource-poor regions . The specific durations of these interventions can be defined by specific countries according to their needs and local facilities, what is key is to identify a combination pattern that allows to protect the health of the population not only from COVID-19 but also from economic hardship and mental health issues. Finally, these findings reinforce the value of dynamic social distancing strategies estimated by earlier studies for the UK, Canada and China [3, 25, 26], and extend these to multiple global regions under various dynamic intervention scenarios.
The strengths and limitations of our study merit careful consideration. First, as restrictive NPIs may need to be maintained worldwide for many months, we have examined the impacts of dynamic NPIs to “switch on” and “switch off” at regular intervals. These measures have shown to be largely unaffected to uncertainties in effective R estimates and in the severity of the virus . Second, NPI strategies only blunt (however prolong) the epidemic cycle, since there is lesser build-up of herd immunity while these interventions are kept in place. If these measures are, however, lifted altogether, a second (potentially more serious) outbreak could occur . Therefore, in the absence of individual-level data and more detailed country-specific parameters, our study provides an illustrative comparison of different “rolling” strategies to suggest (a) when such measures could be lifted, and (b) for how long. Third, we used the most up-to-date disease transmission parameters [4, 17, 18, 20] to construct our adaptive models, based on well-established SEIR model of epidemic dynamics for infectious diseases. Fourth, since different interventions are likely to be implemented differentially and may have a heterogeneous effect in multiple locations, we have chosen a broad illustrative target of reducing the reproduction number R rather than specific community measures that may differ significantly by context. Fifth, we employed age-standardized estimates of hospitalization and infection-fatality-ratios in countries with diverse demographic structures, and considered countries at various categories of national income, in order to provide useful “context-specific” estimates. Finally, we used rise-and-fall timescale of infections (50 days, in the absence of intervention) as the ideal intervention duration and calculated 30-day as the optimal break duration before triggering the next cycle, however specific to each country other combinations could be considered depending the specific settings and availability of resources. In this regard, triggering dynamic interventions based on a specific pre-specified mortality number or rate, as was done in earlier modelling for the UK , would not be optimal for under-developed countries since (a) the health systems are less efficient to ascertain all new cases comprehensively, and (b) a younger demographic would mean that by the time the target mortality threshold is reached for the trigger, the countries have already accrued a significantly large number of cases.
Our study also had several important limitations. In the absence of country-specific, real-time, reproduction numbers for the epidemic, we assumed a constant transmission rate during each modeled cycle. These estimates are likely to vary by a population’s adherence to the NPI and the mix of specific measures put in place. In this respect, our chosen effective R estimates of 0.8 and 0.5 reflect two scenarios of weaker and stronger reduction in transmission, respectively, which could be achieved through social distancing measures and the interruption of transmission chains (e.g., through ramping up testing, contact tracing, isolation and quarantine and other potential strategies chosen by individual countries). We anticipate that the countries will be able to introduce additional control measures with time that might counterbalance the detrimental effect of decreasing compliance. The age-standardisation analyses were based on public sector surveillance data, which may not be robust for all LMIC and LIC countries, with potentials for underestimation of cases and deaths. Furthermore, given unavailability of relevant data, we were unable to adjust for wider social and economic costs of the dynamic approaches; further studies will be needed to quantify these aspects. Additional factors such as potential seasonal variations, environmental pollutions or structural determinants may influence, at least in part, these interventions, highlighting the need of flexibility in terms of the suitable strategy and combination of interventions that can be implemented in each country. Finally, similar to all modelling studies, our analyses were based on several transmission parameter assumptions. Since some uncertainties exist around the natural history and local transmission dynamics of the SARS-CoV-2, the precise efficacy and optimal duration of the dynamic strategies may differ for other countries and will need to be tailored accordingly.
Our study may have important implications. First, we have reported several findings relevant to COVID-19 management and policy development. We provide an actionable strategy option for COVID-19 control by employing dynamic interventions that could delay the epidemic peak, while allowing time to enhance health systems capacities and efforts to develop therapies or vaccines. These dynamic measures also allow interim periods of relaxation in order to minimise socioeconomic disruptions and maximise population compliance to these stringent suppression measures. However, these should be weighed carefully against costs, any risks imposed to the society, and the social protection available in each setting. Second, these findings also stimulate further relevant research that may involve: (a) more in-depth analyses of detailed natural history of the disease (e.g., including transmissibility in asymptomatic state) based on patient-level data, when available, from various countries , (b) various spatial pathways and patterns of epidemic in different circumstances (e.g., co-morbidity, reinfection) and settings (e.g., urban vs. rural); and (c) targeted modelling studies accounting for genomic susceptibility , social behaviour  and economic diversity .
In conclusion, this multi-country analysis demonstrates that intermittent reductions of R below 1 through a potential combination of suppression interventions and relaxation can be a pragmatic strategy for COVID-19 pandemic control. Such a “schedule” of social distancing might be particularly relevant to low-income countries, where a single, prolonged suppression intervention is unsustainable. As a policy option, efficient implementation of dynamic suppression interventions worldwide, therefore, would help: (1) prevent critical care overload and deaths, (2) gain time to develop preventive and clinical measures, and (3) reduce economic hardship globally.