key: cord-0772723-b6ovyq7z authors: Ofori, S. K.; Schwind, J. S.; Sullivan, K. L.; Cowling, B. J.; Chowell, G.; Fung, I. C.-H. title: Transmission dynamics of COVID-19 in Ghana and the impact of public health interventions date: 2021-07-06 journal: nan DOI: 10.1101/2021.07.04.21259991 sha: dd4e9fc325ba81c4963adf4f4729a5880625ca50 doc_id: 772723 cord_uid: b6ovyq7z The study characterized the transmission of COVID-19 in Ghana by estimating the time-varying reproduction number (Rt) and exploring the effect of various public health interventions at the national and regional levels. The median Rt for Ghana and six out of sixteen regions dropped from greater than 1 in March 2020 to less than 1 in September but increased above 1 in January 2021. The relaxation of movement restrictions and religious gatherings were not associated with increased Rt in the regions with lower case burdens. However, Rt increased in most regions after schools were reopened in January 2021. In a regression analysis, we estimated that the per-capita cumulative case count increased with population size. Findings indicated the public health interventions reduced the Rt at the national level while at the regional levels, the Rt fluctuated, and the extent of fluctuation varied across regions. The Rt estimated using nonoverlapping windows suggested the effect of government 85 policies and interventions on the increase and decrease of COVID-19 transmission varied across 86 regions. For example, the restriction of social gatherings and travel bans implemented on March 87 15th, 2020, may have resulted in insignificant changes in Rt at the national level and Greater Accra 88 regions. However, when all borders were closed to human traffic nationwide a week later, the Rt 89 lowered to around 1 accounting for a 12.06% (95% credible interval, CrI: 10.14%, 13.75%) 90 decrease for Ghana. In contrast, the Rt for Greater Accra increased by 27.51% (95% CrI: 24.44%, 91 31.03%). On April 27th, 2020, the mandatory wearing of masks at all businesses and organizations 92 was implemented nationwide which was followed by a decrease in Rt by 15 .34% (95% CrI: 93 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint Transmission dynamics of COVID-19 in Ghana 5 13.72%, 16 .97%) at the national level. With the relaxation of restrictions on social gatherings, 94 there was a decrease in Rt by 0.53% (95% CrI: 0.16%, 0.91%) for Ghana at the national level, 95 11.16% (95% CrI: 10.54%, 11.79%) for the Central region, 12.12% (95% CrI: 4.33%, 19.35%) 96 for the Western region, 52.86% (95% CrI: 34.95%, 65.12%) for the Upper East region and 26.49% 97 (95% CrI: 26.26%, 26.75%) for the Volta region but increased by 19.4% (95% CrI: 18.72%, 98 20.08%) in the Greater Accra region. Overall, the reopening of schools on January 9 th , 2021 was 99 followed by sustained transmission in all regions. Details of percent changes in Rt are in 100 supplementary materials. 101 The assessment of population size and cumulative incidence showed that Ghanaian regions 102 with larger populations experienced higher COVID-19 attack rates ( Figure 2 and Table S3 ). 103 Furthermore, there was a weak positive correlation between mobility changes and COVID-19 104 incidence (Tables S4 and S5 ), e.g., the daily number of new infections led to mobility changes to 105 grocery and pharmacy outlets by 2 days (+2-day-lag, r= 0.291, P value = 0.006) but synchronized 106 with mobility changes to workplaces (0-day-lag, r= 0.269, P value = 0.011). On the contrary, there 107 was a negative but insignificant correlation between daily incidence and residential mobility 108 change (+2-day-lag, r= -0.154, P value = 0.145). Rt was not found to be correlated with mobility 109 changes except for a negative correlation with residential mobility (0-day-lag, r= -0.213, P value= 110 0.043). 111 Overall, Rt >1 showed sustained transmission across the country, but the interventions 112 implemented during the study period appeared to reduce Rt in regions with relatively higher case 113 counts. Interestingly, relaxation of restrictions at social gatherings was not associated with an 114 increase in Rt. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint Transmission dynamics of COVID-19 in Ghana 6 The early estimates of Rt for Ghana were lower than the reported values for other countries 116 like Nigeria, 10 Egypt, 11 and Kenya. 12 The lower R values for Ghana suggests that public health and 117 social measures were effective in containing the epidemic at the initial stage. The estimate also 118 differs slightly from Dwomoh and colleagues (2021), but may be explained by the differences in 119 the methods including the fact that these authors used data from the first 60 days. 13 Nevertheless, 120 Rt remained greater than 1 at the national level and in most regions indicating sustained 121 transmission. It is therefore imperative that public health measures are strengthened throughout 122 the country and efforts are prioritized especially in regions with larger population sizes as the 123 disparity in the case burden across regions was reported in multiple studies. 14, 15 This disparity may 124 be explained by the difficulty in practicing social distancing due to overcrowding stemming from 125 high commercial activities, slum areas facilitating disease spread and urban residency. Hence, such 126 regions will be required to implement more stringent preventive measures in order to decrease 127 transmission. 16 The reopening of schools was associated with a surge in cases in other jurisdictions. 17 This 129 finding supports the need for routine surveillance, case investigation, better protocols for isolation 130 and quarantine, and deployment of protective personal equipment to schools. Although there are 131 limited data on disease dynamics in children in Ghana, children returning to schools would allow 132 parents to return to their workplaces. This assumption is supported by the relative increase in visits 133 to workplaces when schools were reopened according to the Google mobility report. Although the 134 correlation between changes in mobility and transmission intensity was weak in our study, 135 mobility changes in trips to grocery and pharmacy outlets and workplaces have been reported to 136 be associated with lower transmission rates in most countries before and after interventions were 137 relaxed. 18 The difference in correlation between mobility changes and transmission intensity may 138 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint be due to cultural differences, economic status, and variations in the types of interventions 139 implemented. 140 It was unexpected that the relaxation of restrictions on social gatherings was not followed 141 by an increase in the Rt estimates, although such policies make it difficult for social distancing 142 especially in enclosed places like churches or restaurants. This finding may be observed due to the 143 residual effect of the prior mask mandates and the reluctance of prominent churches to resume in-144 person religious activities. 19 In addition, mobility changes to retail and recreation centers remained 145 below baseline even after the restrictions were relaxed ( Figure S6 ). 146 Our study is not without limitations. First, we used publicly available data, which was 147 subject to underreporting or reporting delays. Second, data were only available by the date of the 148 report and unavailable by the date of symptom onset. We, therefore, accounted for this by shifting 149 the data by nine days to approximate the time of infection. Third, given the use of aggregate data, 150 individual-level assumptions cannot be made. Fourth, we cannot rule out the possibility of ceiling 151 effects during months when testing capacity was limited ( Figure S7 ). Finally, data on 152 socioeconomic variables was limited at the regional level hence further exploration of the case 153 burden by region could not be performed. 154 In conclusion, most of the interventions implemented by the Ghanaian government were 155 followed by an overall decrease in Rt estimates at the national scale, but it did not display a decline 156 across all regions. This highlights the importance of a sustained, multi-faceted response at the 157 national level to help mitigate the varying regional effects observed during this pandemic. Lastly, 158 the reopening of schools should be followed by stringent preventive and control measures targeting 159 both adults and children to counteract the increased likelihood of transmission once reopening 160 occurs. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint National 30 th March 2020 L • Lockdown of major cities in two regions Regional 3 rd April 2020 • The Director-General of the Ghana Health Service announced that nose masks were going to be manufactured locally. • Disinfection of markets in Northern, North East, Savannah, and Eastern Regions. • The president approved the construction of hospital-related facilities in the Greater Accra and Ashanti regions. Regional 6 th July 2020 E • Deployment of personnel to monitor COVID-19 cases in high schools. National 18 th July 2020 • Two-phase fumigation exercise. National 26 th July 2020 • COVID-19 related restrictions on transport operators and tourist sites were lifted. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. At the regional level, the time series began when the first cases were reported consecutively and ended after cases were consecutively reported. For example, for Central Region, the time series began on 7 th May 2020. This was done to avoid estimating the Rt too early in the pandemic and to minimize error and uncertainty around the Rt. The reproduction number, Rt, from daily incidence using the generalized growth model The daily number of new infections for the first 25 days of the epidemic and selected regional levels are calibrated using the Generalized Growth Model (GGM). 1 The GGM uses a simple generalized model for the ascending case of an epidemic using the equation: where C′(t) characterizes the epidemic curve (incident case count) over the time t and the solution C(t) which is the cumulative incidence at a given time t, r is a positive number defined as the growth rate (1/t), and p ∈ [0, 1] is the "deceleration of growth parameter". At a of zero, the equation describes the constant incidence and at =1, the equation outlines the exponential growth dynamics. Values of p between 0 and 1 describe sub-exponential growth patterns. 2 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint We first assume a gamma distribution for the generation interval of SARS-CoV-2 with a mean of 4.60 days and a standard deviation of 5.55 days, 3, 4 then estimate the growth rate parameter r, and the deceleration of growth parameter, . The GGM model simulates the growth of total cases (both imported and local cases) at Ii, and estimates the reproduction using the renewal equation: 5 The imported cases at a given time t are given as Ji; the local incidence is Ii at calendar time ti; and ρj represents the discretized probability distribution of generation interval. The factor α measures the relative contribution of imported cases to the transmission process and is assumed to be 0.15. The total number of new cases is the numerator at a given time t= Ii, and the denominator is the total number of primary cases that generated the secondary cases. Thus, the reproduction number is the average number of new cases generated by a primary case at a given calendar time. The uncertainty around the Rt is derived from the parameter estimates (r, p). The Rt is estimated from 300 simulations assuming a negative binomial structure with a mean which is assumed to be a third of the variance. 2 The Rt for the national curve is estimated from the first twenty-five days of infection (March 12th -April 5th, 2020). The Rt is used because it can be estimated easily in real-time and when control measures are implemented. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint The instantaneous reproduction number method in the R package EpiEstim developed by Cori et al and Thompson et al. was used in the analysis. 6, 7 The EpiEstim package uses a Bayesian framework assuming a gamma prior distribution for the posterior distribution for Rt. 3, 6, 8 The parametric method was used in which the serial interval was assumed to follow a gamma distribution with a mean of 4.60 days and a standard deviation of 5.55. 3, 4 Rt using 7-day sliding windows. The instantaneous reproduction number was estimated over a 7-day sliding window given the high variation in daily Rt estimates. The Rt was assumed to be constant in each time window and the average of Rt estimates over 7 days was estimated with its credible intervals. This method is used to compare the Rt before, during, and after a policy is implemented to assess its impact on transmission. 7 An average was taken of the Rt estimates over the period between two policy change time points. The data on the major interventions and newly implemented policies were obtained from the Ghana Health Service website as well as the websites of the major news media. The following data were collected: travel bans and border closures, restriction of social activities, school closure, the mandatory wearing of masks, relaxation of restrictions on social and religious activities, deployment of personnel to monitor COVID-19 in senior high schools, reopening of air borders, and reopening of schools. A detailed list of interventions and their references can be found in Table 1 . Bundled policies were implemented in close succession; hence there is difficulty in assessing the individual effects. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint The power-law relationship between cumulative case count, C, and population size, N (i.e., C~N g where g is an exponent) was transformed into a relationship between log10-transformed per-capita cumulative case count and log10-transformed population size as follows: 9 log 10 (cumulative case count) log 10 (population size) = log 10 (per capita cumulative case count × population size) log 10 (population size) = log 10 (per capita cumulative case count) log 10 (population size) + log 10 (population size) log 10 (population size) = log 10 (per capita cumulative case count) . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint The mobility data was sourced from Google to analyze changes in the number of visits to places in the following categories: (a) grocery and pharmacy, (b) parks, (c) transit stations, (d) retail and recreation, (e) residential, and (f) workplaces. 11 The data provides information for how visits and duration of staying at different places changes compared to baseline values which is the median value for the corresponding day of the week, during the five weeks of January 3 -February 6, 2020. The data is representative of users who opted for location history. The relationship between mobility and the daily number of new infections (or Rt estimates from a 7-day sliding window) is assessed using the time-lagged cross-correlation. The daily incidence of COVID-19 in Ghana, the Greater Accra Region, and the Ashanti region decreased from mid-May to mid-June and surged in mid-July, 2020. For all three locations, the daily incidence dropped again to less than 500 per day and spiked again in January 2021 ( Figure 1 ). Western and Western North regions daily cases peaked in June and July 2020 (150 and around 100 respectively) ( Figure S2 ). The daily incidence for Central, Oti, Volta, and Eastern regions fluctuated throughout the study period: the peak times were in July and December 2020 (Central=around 90 cases), August 2020 (Eastern>150), and June 2020 (> 40 for Oti, and Volta) ( Figures S1-S2) . Bono East and Ahafo cases were clustered between June and September 2020 and peaked in July at 75 and >150 cases per day respectively ( Figure S4 ). Savannah ( Figure S4 ), Upper East, and Upper West ( Figure S3 ) recorded spurious cases sparingly throughout the year. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. 1.53, 1.82) . . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint Table S3 : Linear regression analysis between the log10-transformed per capita cumulative case count and log10-transformed population size using cumulative incidence for sixteen regions. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint Table S4 : Time-lag correlation coefficients between the daily number of new COVID-19 cases, and relative mobility (percentage change from baseline) for trips to retails and recreation, grocery and pharmacy, workplace, and residential categories over the first 90 days of the pandemic in Ghana. Table S5 : Time-lag correlation coefficients between median 7-day sliding window Rt of COVID-19 and relative mobility (percentage change from baseline) for trips to retails and recreation, grocery and pharmacy, workplace, and residential categories over the first 90 days of the pandemic in Ghana. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint Fig S7: COVID-19 testing positivity rate in Ghana by date of report from April 18, 2020 to January 30, 2021. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.04.21259991 doi: medRxiv preprint COVID-19 Ghana's Outbreak Response Management Updates Coronavirus diseases 2019 (COVID-19) response: Highlights of Ghana's scientific and technological innovativeness and breakthroughs A global panel database of pandemic policies (Oxford COVID-19 Government Response Tracker) Global stability and costeffectiveness analysis of COVID-19 considering the impact of the environment: using data from Ghana A New Framework and Software to Estimate Time-Varying Reproduction Numbers During Epidemics Fitting dynamic models to epidemic outbreaks with quantified uncertainty: A primer for parameter uncertainty, identifiability, and forecasts. Infectious Disease Modelling COVID-19 Pandemic Planning Scenarios The 1918-1919 influenza pandemic in England and Wales: spatial patterns in transmissibility and mortality impact Google COVID-19 Community Mobility Reports Early Transmission Dynamics of Novel Coronavirus (COVID-19) in Nigeria Forecasting the scale of the COVID-19 epidemic in Kenya Mathematical modeling of COVID-19 infection dynamics in Ghana: Impact evaluation of integrated government and individual level interventions The sensitivity and specificity analyses of ambient temperature and population size on the transmission rate of the novel coronavirus (COVID-19) in different provinces of Iran The impact of temperature, population size and median age on COVID-19 (SARS-CoV-2) outbreak. Clinical Epidemiology and Global Health Ghana's COVID-19 response: the Black Star can do even better A generalized-growth model to characterize the early ascending phase of infectious disease outbreaks Fitting dynamic models to epidemic outbreaks with quantified uncertainty: A primer for parameter uncertainty, identifiability, and forecasts. Infectious Disease Modelling Disaster Medicine and Public Health Preparedness Estimation of the time-varying reproduction number of COVID-19 outbreak in China Early transmission dynamics of Ebola virus disease (EVD) A New Framework and Software to Estimate Time-Varying Reproduction Numbers During Epidemics Improved inference of time-varying reproduction numbers during infectious disease outbreaks Influenza transmission in households during the 1918 pandemic The 1918-1919 influenza pandemic in England and Wales: spatial patterns in transmissibility and mortality impact Ghana's population by region