key: cord-0825385-zqeu8dnq
authors: Ediriweera, Dileepa S; de Silva, Nilanthi R; Malavige, Neelika G; de Silva, H Janaka
title: AN EPIDEMIOLOGICAL MODEL TO AID DECISION-MAKING FOR COVID-19 CONTROL IN SRI LANKA
date: 2020-04-16
journal: nan
DOI: 10.1101/2020.04.11.20061481
sha: 68220aa23881ccc06387388320fd8b20d6f3a0aa
doc_id: 825385
cord_uid: zqeu8dnq

Background: Sri Lanka diagnosed its first local case of COVID-19 on 11 March 2020. The government acted swiftly to contain transmission, with extensive public health measures. At the end of 30 days, Sri Lanka had 197 cases, 54 recovered and 7 deaths; a staged relaxing of the lockdown is now underway. This paper proposes a theoretical basis for estimating the limits within which transmission should be constrained in order to ensure that the case load remains within the capacity of the health system. Methods: We used the Susceptible, Infected, Recovered (SIR) model to explore the number of new infections and estimate ICU bed requirement at different levels of R0 values after lockout. We developed a web-based application that enables visualization of cases and ICU bed requirements with time, with adjustable parameters that include: population exposed; proportion asymptomatic; number of active and recovered cases; infectious period; R0 or doubling time; proportion critically ill; available ICU beds; duration of ICU stay; and uncertainty of projection. Results: The three-day moving average of the caseload suggested two waves of transmission from Day 0 to 17 (R0=3.32, 95% CI 1.85 - 5.41) and from Day 18 - 30 (R=1.25, 95%CI: 0.93 - 1.63). We estimate that if there are 156 active cases with 91 recovered at the time of lockout, and R increases to 1.5 (doubling time 19 days), under the standard parameters for Sri Lanka, the ICU bed capacity of 300 is likely to be saturated by about 100 days days, signalled by 18 new infections (95% CI 15 - 22) on Day 14 after lockout. Conclusion: Our model suggests that to ensure that the case load remains within the available capacity of the health system after lockout, transmission should not exceed R=1.5. This model and the web-based application may be useful in other low and middle income countries which have similar constraints on health resources.

We used publicly available data and adopted a Susceptible, Infected, Recovered (SIR) model to explore the number of new infections and estimate ICU bed requirement at different levels of R0 values after a lockout period. We considered the entire population of the country exposed, with a 14-day period of infection. We assumed 50% of the infected are symptomatic, of which 5% would require critical care, and that the maximum national capacity for treatment of such patients would be 300 beds in Intensive Care Units.

Results: The cumulative case load increased exponentially during the first 8 days of the epidemic and started flattening out from day 9 in the country. In a lockout situation, if the number of infected doubles every 20 days (R=1.5), at least 300 ICU beds are likely to be required by the end of 100 days. If the number infected doubles every 14 days (R=1.7), the ICU bed capacity is likely to be exceeded in 70 days. This period is reduced to 50 days if infected cases double every 10 days (R=2.0).

Our model suggests that the desired level of control post-lockout to ensure that the case load remains within the assumed capacity of health system lies somewhere between R=1.5 and R=1.7, where the period for doubling of total infections would be 14 -20 days, and the number of new infections would be between 16 and 24 on Day 7 , and between 20 and 34 on Day 14. This model can be refined to suit other low and middle income countries which may contemplate lockout, but have similar health resource constraints.

. 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 April 16, 2020. . (Lake, 2020) .

The 1 st case of COVID-19 was diagnosed in Sri Lanka on 28 January, who was a tourist from China. The 2 nd case was detected nearly 6 weeks later, on 11 March, who was a tour guide, who probably contracted the infection from a group of Italian tourists. Since then, the spread of infection has been relatively slow, and mostly confined to returnees from countries with high transmission, and their contacts. However, it must be noted that in four of the 190 cases diagnosed in the 30 days from 11 March to 9 April 2020, it was not possible to identify the source of infection. It took nearly a week for the caseload to double from 50 (on 19 March) to . 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 April 16, 2020. The government of Sri Lanka acted swiftly to contain transmission, with very stringent measures for social distancing: complete island-wide lockdown, contact tracing and isolation, and quarantine of all inbound passengers were all adopted almost simultaneously. The airport has been closed for inbound passengers since 19 March. The national policy with regard to testing was that all symptomatic individuals clinically suspected of infection with SARS-CoV-2, should be tested in one of seven designated laboratories, using PCR as a diagnostic tool. All positive individuals (regardless of severity of illness) are managed in one of twelve state hospitals, designated for management of COVID-19. These hospitals are also equipped with intensive care units and ventilators for management of the critically ill.

However, these control measures have imposed a very heavy social and economic cost, and Sri Lanka now needs to determine a sensible exit strategy. Even if current local transmission is driven down to zero or near zero during the course of April, for economic and social reasons, the government will be forced to re-open Sri Lanka's borders in the near future, while the pandemic is still going on elsewhere. Given that a commercially available vaccine is thought to be at least 12 -18 months away, it is necessary that the population should be allowed to develop natural immunity under strictly controlled levels of transmission.

It has been suggested that most people infected with SAR-CoV-2 show no symptoms but are still able to infect others. Blanket testing of an isolated village of about 3000 individuals in northern Italy found that 50 -75% of infected individuals were asymptomatic (Day, 2020) .

Analysis of the outbreak in China found that 81% of symptomatic individuals had mild illness, whereas 14% developed severe illness (i.e., dyspnea, respiratory frequency ≥30/min, blood oxygen saturation ≤93%, partial pressure of arterial oxygen to fraction of inspired oxygen ratio . 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 April 16, 2020. . https://doi.org/10.1101/2020.04.11.20061481 doi: medRxiv preprint <300, and/or lung infiltrates >50% within 24 to 48 h) and another 5% became critically ill with respiratory failure, septic shock, and/or multiple organ dysfunction or failure (Wu et al 2020) .

It is the provision of effective care for this last group of patients, who may require ventilation for 2 -3 weeks, that is the crucial limiting factor in any health system. If the spread of infection is not controlled, the R0 of SARS-CoV-2 is such that it will sweep swiftly through the susceptible population, resulting in a large number of very ill persons within a short period of time, which overloads the health system and causes it to collapse. However, it is clearly possible to slow down transmission through social and physical distancing, as has been demonstrated in Sri Lanka. If the infection is allowed to spread slowly, a large proportion of infected persons will recover without requiring hospital care (or not even fall ill at all), some will require hospitalization, and a few, especially the elderly and others with co-morbidities, will die. The availability of beds and ventilators in hospital intensive care units (ICU), to care for critically ill patients is a major constraining factor, and Sri Lanka will need to closely monitor and control the rate of spread of infection so that the requirement for ICU beds and ventilators remains within the available capacity. This paper proposes a theoretical basis for estimating the limit within which the reproduction number should be constrained, in order to ensure that the infection spreads slowly, and the COVID-19 case load remains within the capacity of Sri Lanka's health system.

. 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 April 16, 2020. . https://doi.org/10.1101/2020.04.11.20061481 doi: medRxiv preprint 6

The cumulative case load for the period 11 March to 9 April was plotted as shown in Figure 1 . 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 April 16, 2020. . https://doi.org/10.1101/2020.04.11.20061481 doi: medRxiv preprint According to the data and figure 1, the cumulative case load increased exponentially during the first 8 days of the epidemic and started flattening out from day 9. The flattening out of the case load could be attributed to the strict social distancing and lockdown of cities in the country. Figure 2 illustrates the behavior of the caseload in these 2 phases using a log linear model with (with an exponential function of time) for the period to day 8, and after day 8 using a linear model (with a linear function of time). Therefore, we assumed that the first 8 days represented the natural epidemic curve of the country and used the initial data pertaining to the first 8 days to calculate the reproduction number (we call this 'R1').

We used moving averages of 3 days to calculate R1 ( Figure 3 ). We used R0 package in R programming language to estimate R1 using maximum likelihood method for this period.

This resulted in a R1 value of 3.05 [95%CI: 1.70 -4.98] for the initial 8 days (Figure 4) . . 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 April 16, 2020. (Table 1) . 2.0 10 days . 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 April 16, 2020. . https://doi.org/10.1101/2020.04.11.20061481 doi: medRxiv preprint These SIR models were then run for a period of 18 months using the parameters listed below to estimate the epidemiological curves from 6th of April 2020, assuming a complete lockout situation. While retaining capacity for management of patients with other illnesses, we assumed that up to 300 of these ICU beds may be made available for management of COVID-19 patients at the peak of the epidemic).

. 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 April 16, 2020. . https://doi.org/10.1101/2020.04.11.20061481 doi: medRxiv preprint Figure 5 shows the possible course of the epidemic if transmission remained at the initial level seen during the first 8 days (R=3.05). This model suggests that the epidemic would have peaked in about 3 months, with more than 5,000,000 affected individuals at the very peak. . 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 April 16, 2020. . https://doi.org/10.1101/2020.04.11.20061481 doi: medRxiv preprint Table 2 shows the scenarios that emerge at different values of R, in terms of active infections and ICU requirements after 1 month and after 2 months. If R=1.2, the total cases (symptomatic + asymptomatic) would rise to 534 (from a base of 348) after 1 month, and to 821 after 2 months. If R=1.5, the total cases would be 1,017 after one month, and 2,974 after 2 months. If R=1.7, the number of total cases would rise to 1,563 after 1 month, and 7,008 after 2 months. If R=2.0, the number of total cases would rise to 2,974 after 1 month, and 25,326 after 2 months.

. 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 April 16, 2020. 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 April 16, 2020. . https://doi.org/10.1101/2020.04.11.20061481 doi: medRxiv preprint Figure 8 shows the ICU bed requirements in these same scenarios. When R=1.2, the ICU bed requirement will rise slowly, and would not exceed 300 even after 180 days, whereas if R=1.5, the ICU bed requirement will start to rise rapidly much earlier, and 300 beds would be required by day 99. If R=1.7, this will be in 71 days and if R=2.0, this will be in 50 days. . 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 April 16, 2020. . https://doi.org/10.1101/2020.04.11.20061481 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. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 16, 2020. . https://doi.org/10.1101/2020.04.11.20061481 doi: medRxiv preprint may be considered analogous to the present situation, at least 300 ICU beds are likely to be required for treatment of critically ill patients by the end of the 3 rd month. In the third scenario, when R=1.7, the number of infections rises more rapidly and the requirement of ICU beds is likely to exceed 300 early in the 3 rd month. In the 4 th scenario, when R=2.0, the assumed ICU bed capacity of 300 is likely to be exceeded before the end of the second month.

Thus this model suggests that the desired level of control for Sri Lanka after lockout would lie somewhere between R=1.5 and R=1.7, where the number of new infections would be between 16 and 24 on Day 7 , and between 20 and 34 on Day 14.

The parameters used to set this model can be refined further as more data becomes available. The model can also be used to visualize the impact of varying levels of control in different settings, such as comparison of the 6 high risk districts with the other 19 districts categorized as of lower risk. The model may also be appropriate for other low and middle income countries that may consider lockout after having employed stringent social distancing measures to contain the epidemic, but have similar resource constraints for ICU care.

Aerosol and Surface Stability of SARS-CoV-2 as Compared with SARS-CoV-1

The reproductive number of COVID-19 is higher compared to SARS coronavirus

What we know so far: COVID-19 current clinical knowledge and research

Covid-19: identifying and isolating asymptomatic people helped eliminate virus in Italian village

Characteristics of and Important Lessons From the Coronavirus Disease 2019 (COVID-19) Outbreak in China Summary of a Report of 72 314 Cases From the Chinese Center for Disease Control and Prevention

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

We thank Deirdre Hollingsworth, Don Bundy and Rajitha Wickremasinghe for helpful guidance and comments on the draft manuscript.. 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 April 16, 2020. . https://doi.org/10.1101/2020.04.11.20061481 doi: medRxiv preprint

All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf and declare: no support from any organization for the submitted work; no financial relationships with any organizations that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have influenced the submitted work.. 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 April 16, 2020. . https://doi.org/10.1101/2020.04.11.20061481 doi: medRxiv preprint