Lessons from COVID-19 for rescalable data collection.

Novel data and analyses have had an important role in informing the public health response to the COVID-19 pandemic. Existing surveillance systems were scaled up, and in some instances new systems were developed to meet the challenges posed by the magnitude of the pandemic. We describe the routine and novel data that were used to address urgent public health questions during the pandemic, underscore the challenges in sustainability and equity in data generation, and highlight key lessons learnt for designing scalable data collection systems to support decision making during a public health crisis. As countries emerge from the acute phase of the pandemic, COVID-19 surveillance systems are being scaled down. However, SARS-CoV-2 resurgence remains a threat to global health security; therefore, a minimal cost-effective system needs to remain active that can be rapidly scaled up if necessary. We propose that a retrospective evaluation to identify the cost-benefit profile of the various data streams collected during the pandemic should be on the scientific research agenda.

Estimating the number of undetected COVID-19 cases among travellers from mainland China [version 1, peer review: awaiting peer review]

Background: Since the start of the COVID-19 epidemic in late 2019, there have been more than 152 affected regions and countries with over 110,000 confirmed cases outside mainland China. Methods: We analysed COVID-19 cases among travellers from mainland China to different regions and countries, comparing the region- and country-specific rates of detected and confirmed cases per flight volume to estimate the relative sensitivity of surveillance in different regions and countries. Results: Although travel restrictions from Wuhan City and other cities across China may have reduced the absolute number of travellers to and from China, we estimated that more than two thirds (70%, 95% CI: 54% - 80%, compared to Singapore;75%, 95% CI: 66% - 82%, compared to multiple countries) of cases exported from mainland China have remained undetected. Conclusions: These undetected cases potentially resulted in multiple chains of human-to-human transmission outside mainland China.

Quantifying the effect of delaying the second COVID-19 vaccine dose in England: a mathematical modelling study.

BACKGROUND: The UK was the first country to start national COVID-19 vaccination programmes, initially administering doses 3 weeks apart. However, early evidence of high vaccine effectiveness after the first dose and the emergence of the SARS-CoV-2 alpha variant prompted the UK to extend the interval between doses to 12 weeks. In this study, we aimed to quantify the effect of delaying the second vaccine dose in England. METHODS: We used a previously described model of SARS-CoV-2 transmission, calibrated to COVID-19 surveillance data from England, including hospital admissions, hospital occupancy, seroprevalence data, and population-level PCR testing data, using a Bayesian evidence-synthesis framework. We modelled and compared the epidemic trajectory in the counterfactual scenario in which vaccine doses were administered 3 weeks apart against the real reported vaccine roll-out schedule of 12 weeks. We estimated and compared the resulting numbers of daily infections, hospital admissions, and deaths. In sensitivity analyses, we investigated scenarios spanning a range of vaccine effectiveness and waning assumptions. FINDINGS: In the period from Dec 8, 2020, to Sept 13, 2021, the number of individuals who received a first vaccine dose was higher under the 12-week strategy than the 3-week strategy. For this period, we estimated that delaying the interval between the first and second COVID-19 vaccine doses from 3 to 12 weeks averted a median (calculated as the median of the posterior sample) of 58 000 COVID-19 hospital admissions (291 000 cumulative hospitalisations [95% credible interval 275 000-319 000] under the 3-week strategy vs 233 000 [229 000-238 000] under the 12-week strategy) and 10 100 deaths (64 800 deaths [60 200-68 900] vs 54 700 [52 800-55 600]). Similarly, we estimated that the 3-week strategy would have resulted in more infections compared with the 12-week strategy. Across all sensitivity analyses the 3-week strategy resulted in a greater number of hospital admissions. In results by age group, the 12-week strategy led to more hospitalisations and deaths in older people in spring 2021, but fewer following the emergence of the delta variant during summer 2021. INTERPRETATION: England's delayed-second-dose vaccination strategy was informed by early real-world data on vaccine effectiveness in the context of limited vaccine supplies in a growing epidemic. Our study shows that rapidly providing partial (single-dose) vaccine-induced protection to a larger proportion of the population was successful in reducing the burden of COVID-19 hospitalisations and deaths overall. FUNDING: UK National Institute for Health Research; UK Medical Research Council; Community Jameel; Wellcome Trust; UK Foreign, Commonwealth and Development Office; Australian National Health and Medical Research Council; and EU.

Comparing human and model-based forecasts of COVID-19 in Germany and Poland.

Forecasts based on epidemiological modelling have played an important role in shaping public policy throughout the COVID-19 pandemic. This modelling combines knowledge about infectious disease dynamics with the subjective opinion of the researcher who develops and refines the model and often also adjusts model outputs. Developing a forecast model is difficult, resource- and time-consuming. It is therefore worth asking what modelling is able to add beyond the subjective opinion of the researcher alone. To investigate this, we analysed different real-time forecasts of cases of and deaths from COVID-19 in Germany and Poland over a 1-4 week horizon submitted to the German and Polish Forecast Hub. We compared crowd forecasts elicited from researchers and volunteers, against a) forecasts from two semi-mechanistic models based on common epidemiological assumptions and b) the ensemble of all other models submitted to the Forecast Hub. We found crowd forecasts, despite being overconfident, to outperform all other methods across all forecast horizons when forecasting cases (weighted interval score relative to the Hub ensemble 2 weeks ahead: 0.89). Forecasts based on computational models performed comparably better when predicting deaths (rel. WIS 1.26), suggesting that epidemiological modelling and human judgement can complement each other in important ways.

Bayesian reconstruction of household transmissions to infer the serial interval of COVID-19 by variants of concern: analysis from a prospective community cohort study (Virus Watch).

BACKGROUND: The serial interval is a key epidemiological measure that quantifies the time between an infector's and an infectee's onset of symptoms. This measure helps investigate epidemiological links between cases, and is an important parameter in transmission models used to estimate transmissibility and inform control strategies. The emergence of multiple variants of concern (VOC) during the SARS-CoV-2 pandemic has led to uncertainties about potential changes in the serial interval of COVID-19. We estimated the household serial interval of multiple VOC using data collected by the Virus Watch study. This online, prospective, community cohort study followed-up entire households in England and Wales since mid-June 2020. METHODS: This analysis included 5842 symptomatic individuals with confirmed SARS-CoV-2 infection among 2579 households from Sept 1, 2020, to Aug 10, 2022. SARS-CoV-2 variant designation was based upon national surveillance data of variant prevalence by date and geographical region. We used a Bayesian framework to infer who infected whom by exploring all transmission trees compatible with the observed dates of symptoms, given assumptions on the incubation period and generation time distributions using the R package outbreaker2. FINDINGS: We characterised the serial interval of COVID-19 by VOC. The mean serial interval was shortest for omicron BA5 (2·02 days; 95% credible interval [CrI] 1·26-2·84) and longest for alpha (3·37 days; 2·52-4·04). The mean serial interval before alpha (wild-type) was 2·29 days (95% CrI 1·39-2·94), 3·11 days (2·28-3·90) for delta, 2·72 days (2·01-3·47) for omicron BA1, and 2·67 days (1·90-3·46) for omicron BA2. We estimated that 17% (95% CrI 5-26) of serial interval values are negative across all variants. INTERPRETATION: Most methods estimating the reproduction number from incidence time series do not allow for a negative serial interval by construction. Further research is needed to extend these methods and assess biases introduced by not accounting for negative serial intervals. To our knowledge, this study is the first to use a Bayesian framework to estimate the serial interval of all major SARS-CoV-2 VOC from thousands of confirmed household cases. FUNDING: UK Medical Research Council and Wellcome Trust.

Using next generation matrices to estimate the proportion of infections that are not detected in an outbreak.

Contact tracing, where exposed individuals are followed up to break ongoing transmission chains, is a key pillar of outbreak response for infectious disease outbreaks. Unfortunately, these systems are not fully effective, and infections can still go undetected as people may not remember all their contacts or contacts may not be traced successfully. A large proportion of undetected infections suggests poor contact tracing and surveillance systems, which could be a potential area of improvement for a disease response. In this paper, we present a method for estimating the proportion of infections that are not detected during an outbreak. Our method uses next generation matrices that are parameterized by linked contact tracing data and case line-lists. We validate the method using simulated data from an individual-based model and then investigate two case studies: the proportion of undetected infections in the SARS-CoV-2 outbreak in New Zealand during 2020 and the Ebola epidemic in Guinea during 2014. We estimate that only 5.26% of SARS-CoV-2 infections were not detected in New Zealand during 2020 (95% credible interval: 0.243 - 16.0%) if 80% of contacts were under active surveillance but depending on assumptions about the ratio of contacts not under active surveillance versus contacts under active surveillance 39.0% or 37.7% of Ebola infections were not detected in Guinea (95% credible intervals: 1.69 - 87.0% or 1.70 - 80.9%).

Reconstruction of transmission chains of SARS-CoV-2 amidst multiple outbreaks in a geriatric acute-care hospital: a combined retrospective epidemiological and genomic study.

Background: There is ongoing uncertainty regarding transmission chains and the respective roles of healthcare workers (HCWs) and elderly patients in nosocomial outbreaks of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in geriatric settings. Methods: We performed a retrospective cohort study including patients with nosocomial coronavirus disease 2019 (COVID-19) in four outbreak-affected wards, and all SARS-CoV-2 RT-PCR positive HCWs from a Swiss university-affiliated geriatric acute-care hospital that admitted both Covid-19 and non-Covid-19 patients during the first pandemic wave in Spring 2020. We combined epidemiological and genetic sequencing data using a Bayesian modelling framework, and reconstructed transmission dynamics of SARS-CoV-2 involving patients and HCWs, to determine who infected whom. We evaluated general transmission patterns according to case type (HCWs working in dedicated Covid-19 cohorting wards: HCWcovid; HCWs working in non-Covid-19 wards where outbreaks occurred: HCWoutbreak; patients with nosocomial Covid-19: patientnoso) by deriving the proportion of infections attributed to each case type across all posterior trees and comparing them to random expectations. Results: During the study period (1 March to 7 May 2020), we included 180 SARS-CoV-2 positive cases: 127 HCWs (91 HCWcovid, 36 HCWoutbreak) and 53 patients. The attack rates ranged from 10% to 19% for patients, and 21% for HCWs. We estimated that 16 importation events occurred with high confidence (4 patients, 12 HCWs) that jointly led to up to 41 secondary cases; in six additional cases (5 HCWs, 1 patient), importation was possible with a posterior probability between 10% and 50%. Most patient-to-patient transmission events involved patients having shared a ward (95.2%, 95% credible interval [CrI] 84.2%-100%), in contrast to those having shared a room (19.7%, 95% CrI 6.7%-33.3%). Transmission events tended to cluster by case type: patientnoso were almost twice as likely to be infected by other patientnoso than expected (observed:expected ratio 2.16, 95% CrI 1.17-4.20, p=0.006); similarly, HCWoutbreak were more than twice as likely to be infected by other HCWoutbreak than expected (2.72, 95% CrI 0.87-9.00, p=0.06). The proportion of infectors being HCWcovid was as expected as random. We found a trend towards a greater proportion of high transmitters (≥2 secondary cases) among HCWoutbreak than patientnoso in the late phases (28.6% vs. 11.8%) of the outbreak, although this was not statistically significant. Conclusions: Most importation events were linked to HCW. Unexpectedly, transmission between HCWcovid was more limited than transmission between patients and HCWoutbreak. This finding highlights gaps in infection control and suggests the possible areas of improvements to limit the extent of nosocomial transmission. Funding: This study was supported by a grant from the Swiss National Science Foundation under the NRP78 funding scheme (Grant no. 4078P0_198363).

Inferring the reproduction number using the renewal equation in heterogeneous epidemics.

Real-time estimation of the reproduction number has become the focus of modelling groups around the world as the SARS-CoV-2 pandemic unfolds. One of the most widely adopted means of inference of the reproduction number is via the renewal equation, which uses the incidence of infection and the generation time distribution. In this paper, we derive a multi-type equivalent to the renewal equation to estimate a reproduction number which accounts for heterogeneity in transmissibility including through asymptomatic transmission, symptomatic isolation and vaccination. We demonstrate how use of the renewal equation that misses these heterogeneities can result in biased estimates of the reproduction number. While the bias is small with symptomatic isolation, it can be much larger with asymptomatic transmission or transmission from vaccinated individuals if these groups exhibit substantially different generation time distributions to unvaccinated symptomatic transmitters, whose generation time distribution is often well defined. The bias in estimate becomes larger with greater population size or transmissibility of the poorly characterized group. We apply our methodology to Ebola in West Africa in 2014 and the SARS-CoV-2 in the UK in 2020-2021.

Occupational and community risk of SARS-CoV-2 infection among employees of a long-term care facility: an observational study.

BACKGROUND: We investigated the contribution of both occupational and community exposure for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection among employees of a university-affiliated long-term care facility (LTCF), during the 1st pandemic wave in Switzerland (March-June 2020). METHODS: We performed a nested analysis of a seroprevalence study among all volunteering LTCF staff to determine community and nosocomial risk factors for SARS-CoV-2 seropositivity using modified Poison regression. We also combined epidemiological and genetic sequencing data from a coronavirus disease 2019 (COVID-19) outbreak investigation in a LTCF ward to infer transmission dynamics and acquisition routes of SARS-CoV-2, and evaluated strain relatedness using a maximum likelihood phylogenetic tree. RESULTS: Among 285 LTCF employees, 176 participated in the seroprevalence study, of whom 30 (17%) were seropositive for SARS-CoV-2. Most (141/176, 80%) were healthcare workers (HCWs). Risk factors for seropositivity included exposure to a COVID-19 inpatient (adjusted prevalence ratio [aPR] 2.6; 95% CI 0.9-8.1) and community contact with a COVID-19 case (aPR 1.7; 95% CI 0.8-3.5). Among 18 employees included in the outbreak investigation, the outbreak reconstruction suggests 4 likely importation events by HCWs with secondary transmissions to other HCWs and patients. CONCLUSIONS: These two complementary epidemiologic and molecular approaches suggest a substantial contribution of both occupational and community exposures to COVID-19 risk among HCWs in LTCFs. These data may help to better assess the importance of occupational health hazards and related legal implications during the COVID-19 pandemic.

Estimating the number of undetected COVID-19 cases among travellers from mainland China.

Background: As of August 2021, every region of the world has been affected by the COVID-19 pandemic, with more than 196,000,000 cases worldwide. Methods: We analysed COVID-19 cases among travellers from mainland China to different regions and countries, comparing the region- and country-specific rates of detected and confirmed cases per flight volume to estimate the relative sensitivity of surveillance in different regions and countries. Results: Although travel restrictions from Wuhan City and other cities across China may have reduced the absolute number of travellers to and from China, we estimated that up to 70% (95% CI: 54% - 80%) of imported cases could remain undetected relative to the sensitivity of surveillance in Singapore. The percentage of undetected imported cases rises to 75% (95% CI 66% - 82%) when comparing to the surveillance sensitivity in multiple countries. Conclusions: Our analysis shows that a large number of COVID-19 cases remain undetected across the world.  These undetected cases potentially resulted in multiple chains of human-to-human transmission outside mainland China.

Characterising within-hospitalSARS-CoV-2 transmission events using epidemiological and viral genomic data across two pandemic waves.

Hospital outbreaks of COVID19 result in considerable mortality and disruption to healthcare services and yet little is known about transmission within this setting. We characterise within hospital transmission by combining viral genomic and epidemiological data using Bayesian modelling amongst 2181 patients and healthcare workers from a large UK NHS Trust. Transmission events were compared between Wave 1 (1st March to 25th J'uly 2020) and Wave 2 (30th November 2020 to 24th January 2021). We show that staff-to-staff transmissions reduced from 31.6% to 12.9% of all infections. Patient-to-patient transmissions increased from 27.1% to 52.1%. 40%-50% of hospital-onset patient cases resulted in onward transmission compared to 4% of community-acquired cases. Control measures introduced during the pandemic likely reduced transmissions between healthcare workers but were insufficient to prevent increasing numbers of patient-to-patient transmissions. As hospital-acquired cases drive most onward transmission, earlier identification of nosocomial cases will be required to break hospital transmission chains.

Reproducible parallel inference and simulation of stochastic state space models using odin, dust, and mcstate.

State space models, including compartmental models, are used to model physical, biological and social phenomena in a broad range of scientific fields. A common way of representing the underlying processes in these models is as a system of stochastic processes which can be simulated forwards in time. Inference of model parameters based on observed time-series data can then be performed using sequential Monte Carlo techniques. However, using these methods for routine inference problems can be made difficult due to various engineering considerations: allowing model design to change in response to new data and ideas, writing model code which is highly performant, and incorporating all of this with up-to-date statistical techniques. Here, we describe a suite of packages in the R programming language designed to streamline the design and deployment of state space models, targeted at infectious disease modellers but suitable for other domains. Users describe their model in a familiar domain-specific language, which is converted into parallelised C++ code. A fast, parallel, reproducible random number generator is then used to run large numbers of model simulations in an efficient manner. We also provide standard inference and prediction routines, though the model simulator can be used directly if these do not meet the user's needs. These packages provide guarantees on reproducibility and performance, allowing the user to focus on the model itself, rather than the underlying computation. The ability to automatically generate high-performance code that would be tedious and time-consuming to write and verify manually, particularly when adding further structure to compartments, is crucial for infectious disease modellers. Our packages have been critical to the development cycle of our ongoing real-time modelling efforts in the COVID-19 pandemic, and have the potential to do the same for models used in a number of different domains.

Non-pharmaceutical interventions, vaccination, and the SARS-CoV-2 delta variant in England: a mathematical modelling study.

BACKGROUND: England's COVID-19 roadmap out of lockdown policy set out the timeline and conditions for the stepwise lifting of non-pharmaceutical interventions (NPIs) as vaccination roll-out continued, with step one starting on March 8, 2021. In this study, we assess the roadmap, the impact of the delta (B.1.617.2) variant of SARS-CoV-2, and potential future epidemic trajectories. METHODS: This mathematical modelling study was done to assess the UK Government's four-step process to easing lockdown restrictions in England, UK. We extended a previously described model of SARS-CoV-2 transmission to incorporate vaccination and multi-strain dynamics to explicitly capture the emergence of the delta variant. We calibrated the model to English surveillance data, including hospital admissions, hospital occupancy, seroprevalence data, and population-level PCR testing data using a Bayesian evidence synthesis framework, then modelled the potential trajectory of the epidemic for a range of different schedules for relaxing NPIs. We estimated the resulting number of daily infections and hospital admissions, and daily and cumulative deaths. Three scenarios spanning a range of optimistic to pessimistic vaccine effectiveness, waning natural immunity, and cross-protection from previous infections were investigated. We also considered three levels of mixing after the lifting of restrictions. FINDINGS: The roadmap policy was successful in offsetting the increased transmission resulting from lifting NPIs starting on March 8, 2021, with increasing population immunity through vaccination. However, because of the emergence of the delta variant, with an estimated transmission advantage of 76% (95% credible interval [95% CrI] 69-83) over alpha, fully lifting NPIs on June 21, 2021, as originally planned might have led to 3900 (95% CrI 1500-5700) peak daily hospital admissions under our central parameter scenario. Delaying until July 19, 2021, reduced peak hospital admissions by three fold to 1400 (95% CrI 700-1700) per day. There was substantial uncertainty in the epidemic trajectory, with particular sensitivity to the transmissibility of delta, level of mixing, and estimates of vaccine effectiveness. INTERPRETATION: Our findings show that the risk of a large wave of COVID-19 hospital admissions resulting from lifting NPIs can be substantially mitigated if the timing of NPI relaxation is carefully balanced against vaccination coverage. However, with the delta variant, it might not be possible to fully lift NPIs without a third wave of hospital admissions and deaths, even if vaccination coverage is high. Variants of concern, their transmissibility, vaccine uptake, and vaccine effectiveness must be carefully monitored as countries relax pandemic control measures. FUNDING: National Institute for Health Research, UK Medical Research Council, Wellcome Trust, and UK Foreign, Commonwealth and Development Office.

Comparing the responses of the UK, Sweden and Denmark to COVID-19 using counterfactual modelling.

The UK and Sweden have among the worst per-capita COVID-19 mortality in Europe. Sweden stands out for its greater reliance on voluntary, rather than mandatory, control measures. We explore how the timing and effectiveness of control measures in the UK, Sweden and Denmark shaped COVID-19 mortality in each country, using a counterfactual assessment: what would the impact have been, had each country adopted the others' policies? Using a Bayesian semi-mechanistic model without prior assumptions on the mechanism or effectiveness of interventions, we estimate the time-varying reproduction number for the UK, Sweden and Denmark from daily mortality data. We use two approaches to evaluate counterfactuals which transpose the transmission profile from one country onto another, in each country's first wave from 13th March (when stringent interventions began) until 1st July 2020. UK mortality would have approximately doubled had Swedish policy been adopted, while Swedish mortality would have more than halved had Sweden adopted UK or Danish strategies. Danish policies were most effective, although differences between the UK and Denmark were significant for one counterfactual approach only. Our analysis shows that small changes in the timing or effectiveness of interventions have disproportionately large effects on total mortality within a rapidly growing epidemic.

A Comparative Analysis of Statistical Methods to Estimate the Reproduction Number in Emerging Epidemics, With Implications for the Current Coronavirus Disease 2019 (COVID-19) Pandemic.

BACKGROUND: As the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic continues its rapid global spread, quantification of local transmission patterns has been, and will continue to be, critical for guiding the pandemic response. Understanding the accuracy and limitations of statistical methods to estimate the basic reproduction number, R0, in the context of emerging epidemics is therefore vital to ensure appropriate interpretation of results and the subsequent implications for control efforts. METHODS: Using simulated epidemic data, we assess the performance of 7 commonly used statistical methods to estimate R0 as they would be applied in a real-time outbreak analysis scenario: fitting to an increasing number of data points over time and with varying levels of random noise in the data. Method comparison was also conducted on empirical outbreak data, using Zika surveillance data from the 2015-2016 epidemic in Latin America and the Caribbean. RESULTS: We find that most methods considered here frequently overestimate R0 in the early stages of epidemic growth on simulated data, the magnitude of which decreases when fitted to an increasing number of time points. This trend of decreasing bias over time can easily lead to incorrect conclusions about the course of the epidemic or the need for control efforts. CONCLUSIONS: We show that true changes in pathogen transmissibility can be difficult to disentangle from changes in methodological accuracy and precision in the early stages of epidemic growth, particularly for data with significant over-dispersion. As localized epidemics of SARS-CoV-2 take hold around the globe, awareness of this trend will be important for appropriately cautious interpretation of results and subsequent guidance for control efforts.

Key epidemiological drivers and impact of interventions in the 2020 SARS-CoV-2 epidemic in England.

We fitted a model of SARS-CoV-2 transmission in care homes and the community to regional surveillance data for England. Compared with other approaches, our model provides a synthesis of multiple surveillance data streams into a single coherent modeling framework, allowing transmission and severity to be disentangled from features of the surveillance system. Of the control measures implemented, only national lockdown brought the reproduction number (Rt eff) below 1 consistently; if introduced 1 week earlier, it could have reduced deaths in the first wave from an estimated 48,600 to 25,600 [95% credible interval (CrI): 15,900 to 38,400]. The infection fatality ratio decreased from 1.00% (95% CrI: 0.85 to 1.21%) to 0.79% (95% CrI: 0.63 to 0.99%), suggesting improved clinical care. The infection fatality ratio was higher in the elderly residing in care homes (23.3%, 95% CrI: 14.7 to 35.2%) than those residing in the community (7.9%, 95% CrI: 5.9 to 10.3%). On 2 December 2020, England was still far from herd immunity, with regional cumulative infection incidence between 7.6% (95% CrI: 5.4 to 10.2%) and 22.3% (95% CrI: 19.4 to 25.4%) of the population. Therefore, any vaccination campaign will need to achieve high coverage and a high degree of protection in vaccinated individuals to allow nonpharmaceutical interventions to be lifted without a resurgence of transmission.

Genetic evidence for the association between COVID-19 epidemic severity and timing of non-pharmaceutical interventions.

Unprecedented public health interventions including travel restrictions and national lockdowns have been implemented to stem the COVID-19 epidemic, but the effectiveness of non-pharmaceutical interventions is still debated. We carried out a phylogenetic analysis of more than 29,000 publicly available whole genome SARS-CoV-2 sequences from 57 locations to estimate the time that the epidemic originated in different places. These estimates were examined in relation to the dates of the most stringent interventions in each location as well as to the number of cumulative COVID-19 deaths and phylodynamic estimates of epidemic size. Here we report that the time elapsed between epidemic origin and maximum intervention is associated with different measures of epidemic severity and explains 11% of the variance in reported deaths one month after the most stringent intervention. Locations where strong non-pharmaceutical interventions were implemented earlier experienced much less severe COVID-19 morbidity and mortality during the period of study.

Reduction in mobility and COVID-19 transmission.

In response to the COVID-19 pandemic, countries have sought to control SARS-CoV-2 transmission by restricting population movement through social distancing interventions, thus reducing the number of contacts. Mobility data represent an important proxy measure of social distancing, and here, we characterise the relationship between transmission and mobility for 52 countries around the world. Transmission significantly decreased with the initial reduction in mobility in 73% of the countries analysed, but we found evidence of decoupling of transmission and mobility following the relaxation of strict control measures for 80% of countries. For the majority of countries, mobility explained a substantial proportion of the variation in transmissibility (median adjusted R-squared: 48%, interquartile range - IQR - across countries [27-77%]). Where a change in the relationship occurred, predictive ability decreased after the relaxation; from a median adjusted R-squared of 74% (IQR across countries [49-91%]) pre-relaxation, to a median adjusted R-squared of 30% (IQR across countries [12-48%]) post-relaxation. In countries with a clear relationship between mobility and transmission both before and after strict control measures were relaxed, mobility was associated with lower transmission rates after control measures were relaxed indicating that the beneficial effects of ongoing social distancing behaviours were substantial.