You are on page 1of 12

Modelling projections for COVID­19 epidemic in Sudan

LSHTM CMMID COVID­19 Working Group

05 May 2020

1 Overview
This report provides simulation­based estimates for COVID­19 epidemic scenarios in Sudan. In this model, we use a population of 44M in Sudan, with
an average age of 23.7, based on [1, 2].

• We use a transmission model to project the COVID­19 epidemic in Sudan; it accounts for the local age structure, as well as uncertainty in
transmission due to uncertainty in 𝑅0 and randomness in the process.
• We consider an unmitigated epidemic (i.e., no intervention), and then interventions combining social distancing and shielding interventions to
decrease transmission.
• We estimate predicted peak counts and timing for Sudan in new symptomatic cases and deaths (incidence), as well as peaks in daily hospital
demand (prevalence). We also calculated totals for these values, and compute the impact of the interventions relative to an unmitigated
epidemic.
• The totals for hospital demand are calculated in terms of person­days, which can be translated into the total required expenditure on hospitalised
care at different levels.
• Populations are based on demographics of Sudan and estimated contact patterns, and assume a single national population
• The transmission model is based on global outbreak data, which have primarily been observed in Asia, Europe, and North America.
• Severity rates are rescaled from the global setting. We assume that due to comorbidities, lower­middle income populations will have less benefit
from age and will in general have higher rates of symptomatic disease.
• We start the epidemic simulation with 50 initial infections, proportionally distributed according to the age demography of Sudan. We refer to this
day 0 as the initial introduction. This corresponds roughly to when there were roughly 50 active infections in Sudan; due to under­reporting and
subclinical disease, this cannot be directly determined from reported cases.
• These reports will be updated as information changes, and if Sudan­specific information on hospitalisation rates become available.
• All numerical results are reported to two significant figures only, and are reported as interquartile (IQR) and 95% intervals.
• Detailed modelling methods are given at the end of the report.

Please direct correspondence regarding these reports to carl.pearson@lshtm.ac.uk. Source available at https://github.com/cmmid/covidm_reports.

2 Unmitigated COVID­19 Epidemic Trajectory


The panels in Fig. 1 show aggregate and by­age group outcomes for daily incidence of COVID­19 cases and deaths in Sudan, and daily demand for
hospital care. The ranges for peak timing and values for those outcomes is summarised in Table 2 (by 50% and 95% forecast intervals); Table 3 covers
total counts by 3, 6, 9, and 12 months from initial introduction in Sudan.

In an unmitigated epidemic, we anticipate that in Sudan, one year after the initial introduction, the total number of symptomatic cases will have been
between 12M ­ 16M (95%: 9.4M ­ 19M), the total deaths between 130K ­ 180K (95%: 84K ­ 200K), and total hospitalised person­days between 6.5M
­ 8.2M (95%: 4.1M ­ 9.5M) with between 2.2M ­ 2.9M (95%: 1.4M ­ 3.3M) of those person­days requiring critical care facilities. Table 1 shows the size
and timing of peaks for these main outcomes.

Table 1: Overview of estimated peak impacts based on the population in Sudan. Rates of symptoms and severe outcomes are based on global
outbreak data, which have primarily been observed in Asia, Europe, and North America. Timing is in days from initial introduction to Sudan.

Outcome Peak Day IQR (95% interval) Peak Number IQR (95% interval)

Incidence of Symptomatic Cases 97 ­ 130 (82 ­ 230) 400K ­ 640K (160K ­ 840K)
All Hospital Demand 110 ­ 150 (95 ­ 230) 190K ­ 290K (66K ­ 350K)
General Hospital Demand 110 ­ 150 (98 ­ 240) 120K ­ 200K (45K ­ 250K)
Critical Care Demand 110 ­ 140 (99 ­ 240) 63K ­ 110K (24K ­ 130K)
Incidence of Deaths 120 ­ 160 (100 ­ 240) 3.9K ­ 6.2K (1.4K ­ 7.6K)

1
all <14 15−29 30−44 45−59 60+
ages years years years years years Peak Values
800K 800K

600K
Symptomatic

600K
Cases

400K 400K

200K 200K

0 0

300K 300K
All Hospital
Occupancy

200K 200K

100K 100K

0 0

200K 200K
General Hospital
Occupancy

150K 150K

100K 100K

50K 50K

0 0
2

125K 125K

100K 100K
Critical Care
Occupancy

75K 75K

50K 50K

25K 25K

0 0

6K 6K
Deaths

4K 4K

2K 2K

0 0
3 6 9 12 3 6 9 12 3 6 9 12 3 6 9 12 3 6 9 12 3 6 9 12 all <14 15−29 30−44 45−59 60+
ages years years years years years
Months since initial introduction in Sudan Age Group

Figure 1: Unmitigated epidemics in Sudan. Time series and peak sizes for new symptomatic cases and deaths, general ward hospital demand, critical care hospital demand, and total hospital demand
(general ward plus critical care). Both summary and by age group results are provided; population and underlying age groups from demography of Sudan. The different line transparency denotes
different simulation quantiles; the lightest lines correspond to the 95% interval, then darker the IQR, then the solid line the median. Cases and deaths reflect new daily events (incidence), while hospital
demand reflects the number in hospital on that day (prevalence).
Table 2: Peak timing and values for main outcomes in Sudan for the unmitigated scenario, by age group. Timing is days from the date of initial
introduction.

Peaks Age Group Peak Day, IQR (95% interval) Peak Number, IQR (95% interval)
all ages 97 ­ 130 (82 ­ 230) 400K ­ 640K (160K ­ 840K)
<14 years 95 ­ 130 (89 ­ 220) 130K ­ 210K (51K ­ 270K)
15­29 years 97 ­ 130 (85 ­ 220) 100K ­ 170K (42K ­ 200K)
Incident Cases
30­44 years 93 ­ 130 (79 ­ 220) 89K ­ 130K (32K ­ 160K)
45­59 years 95 ­ 130 (81 ­ 220) 58K ­ 89K (20K ­ 110K)
60+ years 97 ­ 130 (84 ­ 230) 28K ­ 45K (8.8K ­ 58K)
all ages 110 ­ 150 (95 ­ 230) 190K ­ 290K (66K ­ 350K)
<14 years 110 ­ 140 (100 ­ 230) 8.3K ­ 13K (3.4K ­ 17K)
15­29 years 100 ­ 140 (97 ­ 230) 8.7K ­ 14K (3.4K ­ 16K)
Hospital Demand
30­44 years 100 ­ 150 (98 ­ 230) 21K ­ 31K (8.2K ­ 37K)
45­59 years 100 ­ 150 (93 ­ 230) 55K ­ 86K (21K ­ 100K)
60+ years 110 ­ 150 (100 ­ 240) 100K ­ 160K (30K ­ 200K)
all ages 110 ­ 150 (98 ­ 240) 120K ­ 200K (45K ­ 250K)
<14 years 110 ­ 140 (100 ­ 230) 5.4K ­ 8.6K (2.2K ­ 11K)
15­29 years 100 ­ 140 (97 ­ 230) 5.6K ­ 8.6K (2.2K ­ 10K)
Non­critical Demand
30­44 years 100 ­ 150 (98 ­ 230) 14K ­ 20K (5.4K ­ 25K)
45­59 years 100 ­ 150 (93 ­ 230) 36K ­ 56K (14K ­ 66K)
60+ years 110 ­ 150 (99 ­ 240) 65K ­ 110K (20K ­ 140K)
all ages 110 ­ 140 (99 ­ 240) 63K ­ 110K (24K ­ 130K)
<14 years 110 ­ 140 (100 ­ 230) 3K ­ 4.4K (1.2K ­ 5.9K)
15­29 years 110 ­ 140 (98 ­ 230) 2.9K ­ 4.6K (1.1K ­ 5.3K)
Critical Demand
30­44 years 100 ­ 150 (92 ­ 230) 7.5K ­ 11K (2.8K ­ 14K)
45­59 years 110 ­ 150 (94 ­ 230) 19K ­ 29K (7.3K ­ 35K)
60+ years 110 ­ 150 (100 ­ 240) 34K ­ 53K (10K ­ 66K)
all ages 120 ­ 160 (100 ­ 240) 3.9K ­ 6.2K (1.4K ­ 7.6K)
<14 years 110 ­ 150 (110 ­ 240) 180 ­ 280 (83 ­ 350)
15­29 years 120 ­ 150 (100 ­ 240) 170 ­ 270 (87 ­ 330)
Incident Deaths
30­44 years 110 ­ 150 (100 ­ 240) 450 ­ 640 (170 ­ 790)
45­59 years 120 ­ 160 (100 ­ 240) 1.2K ­ 1.8K (470 ­ 2.1K)
60+ years 120 ­ 160 (100 ­ 240) 2.1K ­ 3.3K (660 ­ 4.2K)

3
Table 3: Total counts for main outcomes in Sudan for the unmitigated scenario, by age group; counts are evaluated at 3, 6, 9, and 12 months from initial introduction.

Totals Age Group 3 months 6 months 9 months 12 months


all ages 300K ­ 4.9M (5.8K ­ 15M) 12M ­ 16M (1.1M ­ 19M) 12M ­ 16M (8.4M ­ 19M) 12M ­ 16M (9.4M ­ 19M)
<14 years 85K ­ 1.5M (1.7K ­ 4.6M) 4.2M ­ 5.4M (350K ­ 6.5M) 4.2M ­ 5.4M (2.8M ­ 6.5M) 4.2M ­ 5.4M (2.9M ­ 6.5M)
15­29 years 86K ­ 1.4M (1.7K ­ 3.7M) 3.1M ­ 4M (330K ­ 4.7M) 3.1M ­ 4M (2.2M ­ 4.7M) 3.1M ­ 4M (2.4M ­ 4.7M)
Cases
30­44 years 69K ­ 1.2M (1.2K ­ 3M) 2.5M ­ 3.2M (250K ­ 3.7M) 2.5M ­ 3.2M (1.7M ­ 3.7M) 2.5M ­ 3.2M (1.9M ­ 3.7M)
45­59 years 40K ­ 700K (750 ­ 2.1M) 1.8M ­ 2.2M (150K ­ 2.5M) 1.8M ­ 2.2M (1.1M ­ 2.5M) 1.8M ­ 2.2M (1.3M ­ 2.5M)
60+ years 14K ­ 260K (270 ­ 970K) 910K ­ 1.2M (54K ­ 1.4M) 920K ­ 1.2M (480K ­ 1.4M) 920K ­ 1.2M (520K ­ 1.4M)
all ages 33K ­ 470K (1K ­ 2.8M) 6.4M ­ 8.2M (230K ­ 9.5M) 6.5M ­ 8.2M (3.6M ­ 9.5M) 6.5M ­ 8.2M (4.1M ­ 9.5M)
<14 years 1.6K ­ 23K (47 ­ 130K) 280K ­ 370K (12K ­ 450K) 280K ­ 370K (180K ­ 450K) 280K ­ 370K (190K ­ 450K)
Hospital 15­29 years 1.9K ­ 27K (55 ­ 150K) 260K ­ 340K (14K ­ 390K) 260K ­ 340K (180K ­ 390K) 260K ­ 340K (200K ­ 390K)
Person­Days 30­44 years 4.6K ­ 66K (140 ­ 350K) 670K ­ 830K (32K ­ 950K) 670K ­ 830K (440K ­ 950K) 670K ­ 830K (510K ­ 950K)
45­59 years 12K ­ 170K (290 ­ 910K) 1.8M ­ 2.3M (81K ­ 2.6M) 1.8M ­ 2.3M (1.1M ­ 2.6M) 1.8M ­ 2.3M (1.3M ­ 2.6M)
60+ years 13K ­ 200K (450 ­ 1.2M) 3.3M ­ 4.4M (95K ­ 5.2M) 3.3M ­ 4.4M (1.6M ­ 5.2M) 3.3M ­ 4.4M (1.8M ­ 5.2M)
all ages 22K ­ 320K (590 ­ 1.9M) 4.1M ­ 5.4M (150K ­ 6.2M) 4.2M ­ 5.4M (2.4M ­ 6.2M) 4.2M ­ 5.4M (2.7M ­ 6.2M)
4

<14 years 1.1K ­ 16K (33 ­ 82K) 180K ­ 250K (8.3K ­ 290K) 180K ­ 250K (120K ­ 290K) 180K ­ 250K (130K ­ 290K)
Non­critical 15­29 years 1.3K ­ 19K (38 ­ 96K) 160K ­ 220K (9.3K ­ 260K) 170K ­ 220K (110K ­ 260K) 170K ­ 220K (130K ­ 260K)
Person­Days 30­44 years 3.1K ­ 45K (86 ­ 240K) 430K ­ 540K (21K ­ 620K) 440K ­ 540K (290K ­ 620K) 440K ­ 540K (330K ­ 620K)
45­59 years 8.1K ­ 110K (200 ­ 610K) 1.2M ­ 1.5M (54K ­ 1.7M) 1.2M ­ 1.5M (770K ­ 1.7M) 1.2M ­ 1.5M (880K ­ 1.7M)
60+ years 9.1K ­ 140K (270 ­ 810K) 2.1M ­ 2.9M (63K ­ 3.4M) 2.2M ­ 2.9M (1.1M ­ 3.4M) 2.2M ­ 2.9M (1.2M ­ 3.4M)
all ages 10K ­ 150K (410 ­ 890K) 2.2M ­ 2.9M (80K ­ 3.3M) 2.2M ­ 2.9M (1.2M ­ 3.3M) 2.2M ­ 2.9M (1.4M ­ 3.3M)
<14 years 520 ­ 7.3K (15 ­ 40K) 99K ­ 130K (4.4K ­ 160K) 100K ­ 130K (65K ­ 160K) 100K ­ 130K (69K ­ 160K)
Critical 15­29 years 660 ­ 8.7K (18 ­ 47K) 90K ­ 120K (4.9K ­ 140K) 92K ­ 120K (62K ­ 140K) 92K ­ 120K (71K ­ 140K)
Person­Days 30­44 years 1.5K ­ 22K (42 ­ 120K) 230K ­ 290K (10K ­ 330K) 230K ­ 290K (150K ­ 330K) 230K ­ 290K (170K ­ 330K)
45­59 years 3.8K ­ 53K (120 ­ 300K) 640K ­ 790K (27K ­ 890K) 650K ­ 790K (410K ­ 890K) 650K ­ 790K (470K ­ 890K)
60+ years 4.4K ­ 62K (160 ­ 390K) 1.1M ­ 1.6M (32K ­ 1.9M) 1.1M ­ 1.6M (580K ­ 1.9M) 1.1M ­ 1.6M (650K ­ 1.9M)
all ages 240 ­ 2.6K (13 ­ 16K) 130K ­ 180K (2.9K ­ 200K) 130K ­ 180K (73K ­ 200K) 130K ­ 180K (87K ­ 200K)
<14 years 12 ­ 130 (0 ­ 700) 5.8K ­ 7.8K (140 ­ 9.4K) 6.1K ­ 7.8K (3.8K ­ 9.4K) 6.1K ­ 7.8K (4.1K ­ 9.4K)
15­29 years 14 ­ 150 (1 ­ 910) 5.3K ­ 6.9K (170 ­ 8.3K) 5.5K ­ 7K (3.5K ­ 8.3K) 5.5K ­ 7K (4.2K ­ 8.3K)
Deaths
30­44 years 34 ­ 370 (1 ­ 2.2K) 13K ­ 18K (420 ­ 20K) 14K ­ 18K (8.9K ­ 20K) 14K ­ 18K (10K ­ 20K)
45­59 years 83 ­ 890 (4 ­ 5.6K) 38K ­ 48K (1K ­ 54K) 39K ­ 48K (24K ­ 54K) 39K ­ 48K (28K ­ 54K)
60+ years 99 ­ 1.1K (6 ­ 6.5K) 68K ­ 93K (1.1K ­ 110K) 71K ­ 93K (33K ­ 110K) 71K ­ 93K (39K ­ 110K)
3 Intervention Scenarios
The panels in Fig. 2 compare unmitigated epidemics with 5 different potential interventions in Sudan. The IQR and 95% intervals for peak timing and
values for those outcomes is summarised in Table 5. Fig. 3 shows the relative trajectories in total outcomes, with Table 6 covers cumulative reductions
due to the interventions at 3, 6, 9, and 12 months from initial introduction. Note that these reduction may rise and decline with time: as Fig. 3 shows,
some interventions have large initial impact by delaying an epidemic, but ultimately many of the cases still happen.

• We assumed these interventions are applied at the country level.


• General physical distancing was implemented as a reduction in contacts, and thus transmission, for all interactions outside the household. We
assumed no change in transmission within the household.
• Shielding was implemented by stratifying the population in one shielded and one unshielded compartment. Shielding applies to those aged 60+
years. We assume shielding of this population has a coverage fraction, and reduction fraction; we refer to shielding interventions by coverage /
reduction, so “80/80 shielding” is 80% coverage of 80% contact reduction. We show results for 40/80 and 80/80 here. Other results available
upon request. In Sudan, 80% coverage corresponds to 2M individuals or 5% of the population, with half of those values for 40% coverage.
• To be effective, shielding must be maintained until the epidemic is over and must ensure that there is not substantially increased mixing amongst
the shielded population.

The % net reduction to cases, deaths, and hospitalised person­days in these scenarios, one year after initial introduction in Sudan is summarised in
Table 4:

Table 4: Scenario Effectiveness Summary

80/80 shielding, 80/80 shielding,


Total 20% distancing 50% distancing 40/80 shielding 80/80 shielding
20% distancing 50% distancing

Cases 14­24% (9­91%) 40­72% (26­100%) 8­12% (6­36%) 10­14% (8­38%) 18­28% (13­93%) 43­75% (30­100%)
Deaths 17­27% (11­93%) 41­75% (27­100%) 23­26% (21­53%) 36­39% (34­62%) 42­49% (38­96%) 60­83% (50­100%)
Critical
17­26% (11­92%) 41­74% (27­100%) 23­26% (21­50%) 36­39% (34­59%) 42­49% (38­96%) 60­82% (50­100%)
Person­Days
Hospital
17­26% (12­92%) 41­74% (26­100%) 23­26% (21­49%) 36­39% (34­59%) 42­49% (38­96%) 60­82% (50­100%)
Person­Days
Non­critical
17­27% (11­92%) 41­74% (26­100%) 23­26% (21­49%) 36­39% (34­59%) 42­49% (38­96%) 59­82% (50­100%)
Person­Days

5
80/80 shielding, 80/80 shielding,
Unmitigated 20% distancing 50% distancing 40/80 shielding 80/80 shielding
20% distancing 50% distancing Peak Values
800K 800K

600K
Symptomatic

600K
Cases

400K 400K

200K 200K

0 0

300K 300K
All Hospital
Occupancy

200K 200K

Unmitigated
100K 100K

0 0 20% distancing

200K 200K 50% distancing


General Hospital
Occupancy

150K 150K
40/80 shielding
100K 100K

50K 50K
80/80 shielding
0
6

125K 125K 80/80 shielding,


20% distancing
100K 100K
Critical Care
Occupancy

75K 80/80 shielding,


75K 50% distancing
50K 50K

25K 25K

0 0

6K 6K
Deaths

4K 4K

2K 2K

0 0
3 6 9 12 3 6 9 12 3 6 9 12 3 6 9 12 3 6 9 12 3 6 9 12 3 6 9 12

Months since initial introduction in Sudan

Figure 2: Epidemics in Sudan with interventions for all ages. Time series and peak counts for new symptomatic cases and deaths, general ward hospital demand, critical care hospital demand, and
total hospital demand (general ward plus critical care). The different line transparency denotes different simulation quantiles; the lightest lines correspond to the 95% interval, then darker the IQR,
then the solid line the median. Cases and deaths reflect new daily events (incidence), while hospital demand reflects the number in hospital on that day (prevalence).
Cumulative Total Count Total Values at 3, 6, 9, and 12 Months

15M 15M
Total Cases

10M 10M

5M 5M

0 0

8M 8M
Person−Days
All Hospital

5M 5M

2M 2M

0 0

6M 6M
General Hospital
Person−Days

4M 4M

2M 2M

0 0
7

3M 3M
Person−Days
Critical Care

2M 2M

1M 1M

0 0

150K 150K
Total Deaths

100K 100K

50K 50K

0 0
3 6 9 12 Unmitigated 20% distancing 50% distancing 40/80 shielding 80/80 shielding 80/80 shielding, 80/80 shielding,
20% distancing 50% distancing
Months since initial introduction in Sudan

Figure 3: Trajectories for cumulative outcomes in Sudan. The ribbons correspond to the 95% intervals (lightest) and IQR (darker), with solid lines for the median. Note that some interventions have
large initial impact by delaying an epidemic, but ultimately many of the cases still happen; when comparing interventions to an unmitigated epidemic, this leads to large initial reduction in total cases,
which later declines.
Table 5: Peak timing and values for main outcomes in Sudan, by intervention scenario. Timing is from the date of ongoing community spread with 50
infections.

Scenario All ages Peak day IQR (95% interval) Peak number IQR (95% interval)
Incident Cases 97 ­ 130 (82 ­ 230) 400K ­ 640K (160K ­ 840K)
Hospital Demand 110 ­ 150 (95 ­ 230) 190K ­ 290K (66K ­ 350K)
Unmitigated Non­critical Demand 110 ­ 150 (98 ­ 240) 120K ­ 200K (45K ­ 250K)
Critical Demand 110 ­ 140 (99 ­ 240) 63K ­ 110K (24K ­ 130K)
Incident Deaths 120 ­ 160 (100 ­ 240) 3.9K ­ 6.2K (1.4K ­ 7.6K)
Incident Cases 110 ­ 170 (99 ­ 360) 200K ­ 380K (11K ­ 530K)
Hospital Demand 130 ­ 190 (110 ­ 370) 99K ­ 180K (4.1K ­ 240K)
20% distancing Non­critical Demand 130 ­ 190 (110 ­ 370) 62K ­ 120K (2.7K ­ 160K)
Critical Demand 130 ­ 180 (110 ­ 370) 33K ­ 62K (1.5K ­ 85K)
Incident Deaths 140 ­ 190 (120 ­ 360) 2K ­ 3.7K (89 ­ 5.2K)
Incident Cases 160 ­ 290 (120 ­ 150) 27K ­ 150K (470 ­ 260K)
Hospital Demand 180 ­ 320 (130 ­ 160) 14K ­ 70K (170 ­ 140K)
50% distancing Non­critical Demand 180 ­ 320 (140 ­ 160) 9.2K ­ 46K (130 ­ 91K)
Critical Demand 170 ­ 320 (130 ­ 160) 4.9K ­ 24K (74 ­ 48K)
Incident Deaths 190 ­ 340 (150 ­ 190) 310 ­ 1.5K (6 ­ 3K)
Incident Cases 100 ­ 150 (88 ­ 320) 280K ­ 460K (68K ­ 650K)
Hospital Demand 120 ­ 160 (100 ­ 340) 110K ­ 190K (23K ­ 250K)
40/80 shielding Non­critical Demand 110 ­ 170 (100 ­ 340) 77K ­ 130K (16K ­ 180K)
Critical Demand 120 ­ 170 (100 ­ 340) 41K ­ 67K (8.4K ­ 91K)
Incident Deaths 130 ­ 170 (110 ­ 350) 2.4K ­ 3.9K (540 ­ 5.6K)
Incident Cases 100 ­ 150 (90 ­ 310) 280K ­ 480K (66K ­ 670K)
Hospital Demand 110 ­ 160 (100 ­ 330) 97K ­ 160K (21K ­ 220K)
80/80 shielding Non­critical Demand 120 ­ 160 (100 ­ 330) 67K ­ 110K (13K ­ 150K)
Critical Demand 120 ­ 160 (100 ­ 330) 35K ­ 54K (7K ­ 75K)
Incident Deaths 130 ­ 170 (110 ­ 340) 2.1K ­ 3.4K (450 ­ 4.6K)
Incident Cases 110 ­ 170 (98 ­ 370) 180K ­ 350K (6K ­ 520K)
Hospital Demand 130 ­ 190 (110 ­ 370) 62K ­ 120K (1.9K ­ 170K)
80/80 shielding,
Non­critical Demand 130 ­ 190 (100 ­ 370) 41K ­ 74K (1.3K ­ 120K)
20% distancing
Critical Demand 130 ­ 190 (110 ­ 370) 21K ­ 41K (690 ­ 62K)
Incident Deaths 140 ­ 200 (110 ­ 370) 1.3K ­ 2.5K (42 ­ 3.8K)
Incident Cases 160 ­ 300 (120 ­ 160) 25K ­ 140K (500 ­ 270K)
Hospital Demand 170 ­ 300 (130 ­ 170) 7.6K ­ 48K (160 ­ 93K)
80/80 shielding,
Non­critical Demand 170 ­ 300 (130 ­ 180) 4.9K ­ 33K (130 ­ 60K)
50% distancing
Critical Demand 170 ­ 320 (130 ­ 160) 2.7K ­ 18K (63 ­ 32K)
Incident Deaths 180 ­ 330 (140 ­ 180) 180 ­ 1.1K (4 ­ 2K)

8
Table 6: Net reductions for main outcomes in Sudan by intervention scenario; reductions are evaluated at 3, 6, 9, and 12 months. Note that net reductions may decline with time; as shown in Fig. 3,
some interventions have a large initial impact, but ultimately still allow many cases.

Scenario Outcome 3 months 6 months 9 months 12 months


Cases 240K ­ 4.3M (4.3K ­ 11M) 2M ­ 4.9M (850K ­ 12M) 2M ­ 3.4M (1.2M ­ 7.6M) 2M ­ 3.4M (1.2M ­ 6.9M)
Hospital Person­Days 23K ­ 400K (740 ­ 2.2M) 1.3M ­ 3.3M (220K ­ 5.9M) 1.2M ­ 1.9M (920K ­ 3.9M) 1.2M ­ 1.9M (920K ­ 3.3M)
20% distancing Non­critical Person­Days 16K ­ 270K (430 ­ 1.5M) 860K ­ 2.2M (150K ­ 3.8M) 840K ­ 1.3M (600K ­ 2.5M) 840K ­ 1.3M (600K ­ 2.2M)
Critical Person­Days 7.8K ­ 130K (220 ­ 710K) 450K ­ 1.2M (77K ­ 2.1M) 440K ­ 650K (310K ­ 1.4M) 450K ­ 650K (310K ­ 1.2M)
Deaths 170 ­ 2K (7 ­ 13K) 28K ­ 77K (2.7K ­ 130K) 27K ­ 40K (18K ­ 84K) 27K ­ 40K (19K ­ 71K)
Cases 260K ­ 4.7M (4.3K ­ 14M) 7M ­ 13M (1.1M ­ 16M) 5.8M ­ 11M (3.6M ­ 15M) 5.8M ­ 9.5M (3.6M ­ 14M)
Hospital Person­Days 23K ­ 430K (740 ­ 2.6M) 3.9M ­ 6.6M (230K ­ 7.8M) 3.1M ­ 5.3M (2M ­ 6.9M) 3.1M ­ 4.8M (2.1M ­ 6.6M)
50% distancing Non­critical Person­Days 16K ­ 290K (430 ­ 1.7M) 2.6M ­ 4.3M (150K ­ 5.1M) 2M ­ 3.5M (1.3M ­ 4.5M) 2M ­ 3.1M (1.4M ­ 4.3M)
Critical Person­Days 7.9K ­ 140K (220 ­ 820K) 1.3M ­ 2.3M (78K ­ 2.8M) 1M ­ 1.9M (690K ­ 2.4M) 1M ­ 1.7M (750K ­ 2.3M)
Deaths 160 ­ 2.2K (7 ­ 14K) 89K ­ 150K (2.7K ­ 170K) 65K ­ 120K (42K ­ 150K) 66K ­ 110K (45K ­ 140K)
Cases 210K ­ 3.4M (3.4K ­ 7.1M) 1.1M ­ 2M (700K ­ 7.9M) 1.1M ­ 1.8M (790K ­ 4.2M) 1.1M ­ 1.8M (810K ­ 2.9M)
Hospital Person­Days 21K ­ 350K (510 ­ 1.8M) 1.7M ­ 2.2M (220K ­ 4.5M) 1.6M ­ 2M (1.3M ­ 2.4M) 1.6M ­ 2M (1.3M ­ 2.3M)
40/80 shielding Non­critical Person­Days 14K ­ 240K (340 ­ 1.2M) 1.1M ­ 1.5M (140K ­ 3M) 1M ­ 1.3M (870K ­ 1.6M) 1M ­ 1.3M (880K ­ 1.5M)
9

Critical Person­Days 7.3K ­ 120K (170 ­ 570K) 600K ­ 770K (75K ­ 1.6M) 570K ­ 690K (460K ­ 810K) 570K ­ 690K (470K ­ 780K)
Deaths 140 ­ 1.8K (7 ­ 11K) 36K ­ 50K (2.7K ­ 98K) 34K ­ 42K (28K ­ 49K) 34K ­ 42K (28K ­ 48K)
Cases 200K ­ 3.4M (4.3K ­ 7.2M) 1.4M ­ 2.3M (900K ­ 7.8M) 1.4M ­ 2M (1M ­ 4.4M) 1.4M ­ 2M (1M ­ 3M)
Hospital Person­Days 21K ­ 370K (700 ­ 2M) 2.6M ­ 3.3M (220K ­ 4.7M) 2.5M ­ 3M (2M ­ 3.5M) 2.5M ­ 3M (2M ­ 3.4M)
80/80 shielding Non­critical Person­Days 14K ­ 250K (390 ­ 1.3M) 1.7M ­ 2.1M (140K ­ 3.1M) 1.6M ­ 2M (1.3M ­ 2.3M) 1.6M ­ 2M (1.3M ­ 2.2M)
Critical Person­Days 7.3K ­ 120K (250 ­ 630K) 920K ­ 1.2M (76K ­ 1.7M) 870K ­ 1.1M (700K ­ 1.2M) 870K ­ 1.1M (690K ­ 1.2M)
Deaths 140 ­ 1.9K (7 ­ 12K) 56K ­ 70K (2.7K ­ 110K) 53K ­ 64K (42K ­ 73K) 52K ­ 64K (42K ­ 72K)
Cases 220K ­ 4.2M (4.3K ­ 9.9M) 2.5M ­ 5.3M (1M ­ 12M) 2.6M ­ 4M (1.8M ­ 7.7M) 2.6M ­ 4M (1.8M ­ 7.3M)
Hospital Person­Days 21K ­ 400K (700 ­ 2.3M) 3.2M ­ 4.2M (220K ­ 6.2M) 3.1M ­ 3.7M (2.5M ­ 4.4M) 3M ­ 3.7M (2.6M ­ 4.1M)
80/80 shielding,
Non­critical Person­Days 14K ­ 270K (390 ­ 1.6M) 2.1M ­ 2.8M (150K ­ 4M) 2M ­ 2.4M (1.6M ­ 2.9M) 2M ­ 2.4M (1.7M ­ 2.7M)
20% distancing
Critical Person­Days 7.4K ­ 130K (250 ­ 750K) 1.1M ­ 1.5M (77K ­ 2.2M) 1M ­ 1.3M (890K ­ 1.6M) 1M ­ 1.3M (930K ­ 1.4M)
Deaths 140 ­ 2K (7 ­ 14K) 69K ­ 94K (2.7K ­ 130K) 65K ­ 78K (54K ­ 96K) 64K ­ 77K (56K ­ 85K)
Cases 250K ­ 4.7M (4.3K ­ 14M) 7.4M ­ 13M (1.1M ­ 16M) 6.3M ­ 11M (4.1M ­ 15M) 6.3M ­ 9.9M (4.2M ­ 14M)
Hospital Person­Days 22K ­ 430K (700 ­ 2.6M) 5M ­ 6.8M (230K ­ 8M) 4.5M ­ 5.8M (3.4M ­ 7M) 4.5M ­ 5.6M (3.6M ­ 6.8M)
80/80 shielding,
Non­critical Person­Days 15K ­ 290K (390 ­ 1.8M) 3.2M ­ 4.5M (150K ­ 5.2M) 2.9M ­ 3.8M (2.2M ­ 4.6M) 2.9M ­ 3.7M (2.4M ­ 4.5M)
50% distancing
Critical Person­Days 7.4K ­ 140K (250 ­ 840K) 1.7M ­ 2.4M (78K ­ 2.8M) 1.5M ­ 2M (1.2M ­ 2.5M) 1.5M ­ 2M (1.2M ­ 2.4M)
Deaths 140 ­ 2.1K (7 ­ 15K) 100K ­ 150K (2.8K ­ 170K) 97K ­ 130K (72K ­ 150K) 95K ­ 120K (77K ­ 150K)
4 Methods, data and assumptions
We used the stochastic age­structured dynamic transmission model reported in [3].

4.1 Dynamic transmission model


We used a stochastic compartmental model stratified into 5­year age bands, with individuals classified according to current disease status (Fig. 4))
and transmission between groups based on social mixing patterns [4, 5]. After infection with SARS­CoV­2 in the model, susceptible individuals pass
through a latent period before becoming infectious, either with a preclinical and then clinical infection, or with a subclinical infection, before recovery
or isolation. We refer to those infections causing few or no symptoms as subclinical. We assume older individuals are more likely to show clinical
symptoms [4].

75+
70−74
60+ 65−69 Contacts
60−64
55−59 20
45−59 50−54
45−49

From
15
40−44
35−39
30−44 30−34 10
25−29
20−24 5
15−29 15−19
10−14
5−9
0−4
<14

0−4
5−9
10−14
15−19
20−24
25−29
30−34
35−39
40−44
45−49
50−54
55−59
60−64
65−69
70−74
75+
0 5000 10000 15000
Population (thousands) To

Figure 4: Population Structure; contact matrix based on Ethiopia

4.2 Key model parameters


As documented in [3], we collated multiple sources of evidence to estimate key model parameters. In a meta­analysis, we estimated that the basic
reproduction number, 𝑅0 , was 2.7 (95% credible interval: 1.6–3.9) across settings without substantial control measures in place (𝑅0 describes the
average number of secondary infections caused by a typical primary infection in a completely susceptible population). We derived age­stratified case
fatality ratios (CFR) to estimate a CFR that ranged substantially across age groups, from 0.1% in the 20–29 age group to 7.7% in the over­80 age
group.

4.3 Key parameters of the transmission model


We used a serial interval of 6.5 days based on published studies [6–8], and assumed that the length of the preclinical period was 30% of the total
period of clinical infectiousness [9]. From this, we fixed the mean of the latent period to 4 days, the mean duration of preclinical infectiousness to 1.5
days, and the mean duration of clinical infectiousness to 3.5 days. The basic reproduction number 𝑅0 was estimated by synthesizing the results of
a literature review (Fig. S1 in [3]). For each reported value of the basic reproduction number, we matched a flexible PERT distribution (a shifted beta
distribution parameterised by minimum, maximum, and mode) to the median and confidence interval reported in each study. We sampled from the
resulting distributions, weighting each study equally, to obtain estimates of 𝑅0 for our simulations. The age­specific clinical fraction, 𝑦𝑖 , was adopted
from an estimate based on case data from 6 countries [4], and the relative infectiousness of subclinical cases, 𝑓 , was assumed to be 50% relative to
clinical cases, as assumed in a previous study [4].

4.4 Hospital burden estimation


To calculate ICU and non­ICU beds in use through time, we scaled age­stratified symptomatic cases by age­specific hospitalisation and critical outcome
probability, then summed to get the total number of hospitalised and critical cases. We then distributed hospitalised cases over time based on expected
time of hospitalisation and duration admitted. We assumed gamma­distributed delays, with the shape parameter set equal to the mean, for: delay from
symptom onset to hospitalisation of mean 7 days (standard deviation 2.65) [10, 11]; delay from hospitalisation to discharge / death for non­ICU patients
of mean 8 days (s.d. 2.83) [12]; delay from hospitalisation to discharge / death for ICU patients of mean 10 days (s.d. 3.16) [11]; and delay from onset
to death of mean 22 days (s.d. 4.69) [10, 11]. We calculated the age­specific case fatality ratio based on data from the COVID­19 outbreak in China
and on the Diamond Princess cruise ship. We first calculated the naive case fatality ratio, nCFR, (i.e. deaths/cases) for each age group, then scaled
down the naive CFR based on a correction factor estimated from data from the Diamond Princess [13] to give an adjusted CFR. We then calculated
risk of hospitalisation based on the ratio of severe and critical cases to cases (18.5%) and deaths to cases (2.3%) in the early China data, which we

10
took to imply 8.04 times more hospitalisations than deaths in each age group. We assumed all age groups had a 30% risk of requiring critical care if
hospitalised [11].

4.5 Severity Shift


To account for increased prevalence of comorbidities in lower­middle income countries, we shifted the age­specific outcome probabilities by ten years.
E.g., the 0­4 year old category in our model has the same risk of disease as we estimated elsewhere for 10­14 year olds. We also applied a 1.5 relative
risk modifier to the case fatality ratio, this has the net effect of increasing the death rates relative to hospitalisation rated and infections.

5 Acknowledgements
The following authors are part of the Centre for Mathematical Modelling of Infectious Disease (CMMID) COVID­19 working group; each contributed
in processing, cleaning and interpretation of data, interpreted findings, contributed to the manuscript, and approved the work for publication: Emily
S Nightingale, James D Munday, Graham Medley, Hamish P Gibbs, Sam Abbott, Rein M G J Houben, Kathleen O’Reilly, Kiesha Prem, Akira Endo,
Samuel Clifford, Mark Jit, Simon R Procter, Nikos I Bosse, Kevin van Zandvoort, Anna M Foss, Alicia Rosello, Quentin J Leclerc, Sebastian Funk,
Stéphane Hué, Eleanor M Rees, David Simons, Christopher I Jarvis, Carl A B Pearson, Adam J Kucharski, Petra Klepac, Joel Hellewell, Arminder K
Deol, Rachel Lowe, Nicholas G. Davies, Charlie Diamond, Damien C Tully, Gwenan M Knight, Jon C Emery, Billy J Quilty, Yang Liu, W John Edmunds,
Megan Auzenbergs, C Julian Villabona­Arenas, Katherine E. Atkins, Timothy W Russell, Fiona Yueqian Sun, Stefan Flasche, Rosalind M Eggo, Thibaut
Jombart, Amy Gimma, and Sophie R Meakin.

Contributing authors gratefully acknowledge funding of the NTD Modelling Consortium by the Bill and Melinda Gates Foundation (OPP1184344) and via
other grants (INV­003174). They also acknowledge support from Elrha’s Research for Health in Humanitarian Crises (R2HC) Programme, which aims
to improve health outcomes by strengthening the evidence base for public health interventions in humanitarian crises. The R2HC programme is funded
by the UK Government (DFID), the Wellcome Trust, and the UK National Institute for Health Research (NIHR). This work was also supported by the
Department for International Development/Wellcome Epidemic Preparedness ­ Coronavirus research programme (ref. 221303/Z/20/Z). Additionally,
the authors acknowledge Global Challenges Research Fund (GCRF) project ‘RECAP’ managed through RCUK and ESRC (ES/P010873/1). Finally, this
work was also supported by grants from HDR UK (grant: MR/S003975/1), MRC (grant: MC_PC 19065), NIHR (16/137/109), NIHR: Health Protection
Research Unit for Modelling Methodology (HPRU­2012­10096), and the European Commission (101003688).

The following funding sources are acknowledged as providing funding for the working group authors: Alan Turing Institute (AE), BBSRC LIDP
(BB/M009513/1: DS), Bill & Melinda Gates Foundation (INV­003174: KP, MJ, YL; NTD Modelling Consortium OPP1184344: GM, CABP; OPP1180644:
SRP; OPP1183986: ESN; OPP1191821: KO’R, MA; ), ERC Starting Grant (#757688: JV­A, KEA; #757699: JCE, RMGJH), European Commis­
sion (101003688: KP, MJ, WJE, YL), Global Challenges Research Fund (ES/P010873/1: AG, CIJ), Nakajima Foundation (AE), NIHR (16/137/109:
CD, FYS, MJ, YL, BQ; HPRU Modelling Methodology: TJ; HPRU­2012­10096: NGD; PR­OD­1017­20002: AR), RCUK/ESRC (ES/P010873/1: TJ),
Royal Society (Dorothy Hodgkin Fellowship: RL; RP\EA\180004: PK), UK DHSC/UK Aid/NIHR (ITCRZ 03010: HPG), HDR UK Innovation Fellowship
(MR/S003975/1: RME), UK MRC (LID DTP MR/N013638/1: EMR, QJL; MR/P014658/1: GMK), UK Public Health Rapid Support Team (TJ), Wellcome
Trust (206250/Z/17/Z: AJK, TWR; 208812/Z/17/Z: SC, SF; 210758/Z/18/Z: JDM, JH, NIB, SA, SFunk, SRM), No funding (AKD, AMF, DCT, SH).

The views expressed in this publication are those of the author(s) and not necessarily those of any of the listed funding sources.

References
[1] United Nations, Department of Economic and Social Affairs, Population Division. World population prospects 2019: Data booket.
ST/ESA/SER.A/424, 2019.

[2] United Nations Population Division. wpp2019: World population prospects 2019, 2020. URL https://CRAN.R­project.org/package=wpp2019.

[3] Nicholas G Davies, Adam J Kucharski, Rosalind M Eggo, Amy Gimma, and W. John Edmunds. The effect of non­pharmaceutical interventions
on covid­19 cases, deaths and demand for hospital services in the uk: a modelling study. medRxiv, 2020. doi: 10.1101/2020.04.01.20049908.
URL https://www.medrxiv.org/content/early/2020/04/06/2020.04.01.20049908.

11
[4] Nicholas G Davies, Petra Klepac, Yang Liu, Kiesha Prem, Mark Jit, and Rosalind M Eggo. Age­dependent effects in the transmission and control
of covid­19 epidemics. medRxiv, 2020. doi: 10.1101/2020.03.24.20043018. URL https://www.medrxiv.org/content/early/2020/03/27/2020.03.24.
20043018.

[5] Joël Mossong, Niel Hens, Mark Jit, Philippe Beutels, Kari Auranen, Rafael Mikolajczyk, Marco Massari, Stefania Salmaso, Gianpaolo Scalia
Tomba, Jacco Wallinga, et al. Social contacts and mixing patterns relevant to the spread of infectious diseases. PLoS medicine, 5(3), 2008.

[6] Qun Li, Xuhua Guan, Peng Wu, Xiaoye Wang, Lei Zhou, Yeqing Tong, Ruiqi Ren, Kathy SM Leung, Eric HY Lau, Jessica Y Wong, et al. Early
transmission dynamics in wuhan, china, of novel coronavirus–infected pneumonia. New England Journal of Medicine, 2020.

[7] Qifang Bi, Yongsheng Wu, Shujiang Mei, Chenfei Ye, Xuan Zou, Zhen Zhang, Xiaojian Liu, Lan Wei, Shaun A. Truelove, Tong Zhang, Wei
Gao, Cong Cheng, Xiujuan Tang, Xiaoliang Wu, Yu Wu, Binbin Sun, Suli Huang, Yu Sun, Juncen Zhang, Ting Ma, Justin Lessler, and Tiejian
Feng. Epidemiology and transmission of COVID­19 in 391 cases and 1286 of their close contacts in shenzhen, china: a retrospective cohort
study. The Lancet Infectious Diseases, Apr 2020. ISSN 1473­3099. doi: 10.1016/S1473­3099(20)30287­5. URL https://doi.org/10.1016/S1473­
3099(20)30287­5.

[8] Hiroshi Nishiura, Natalie M. Linton, and Andrei R. Akhmetzhanov. Serial interval of novel coronavirus (COVID­19) infections. International Journal
of Infectious Diseases, 93:284–286, Apr 2020. ISSN 1201­9712. doi: 10.1016/j.ijid.2020.02.060. URL https://doi.org/10.1016/j.ijid.2020.02.060.

[9] Yang Liu, Sebastian Funk, Stefan Flasche, et al. The contribution of pre­symptomatic infection to the transmission dynamics of covid­2019.
Wellcome Open Research, 5(58):58, 2020.

[10] Natalie M Linton, Tetsuro Kobayashi, Yichi Yang, Katsuma Hayashi, Andrei R Akhmetzhanov, Sung­mok Jung, Baoyin Yuan, Ryo Kinoshita, and
Hiroshi Nishiura. Incubation period and other epidemiological characteristics of 2019 novel coronavirus infections with right truncation: a statistical
analysis of publicly available case data. Journal of clinical medicine, 9(2):538, 2020.

[11] Bin Cao, Yeming Wang, Danning Wen, Wen Liu, Jingli Wang, Guohui Fan, Lianguo Ruan, Bin Song, Yanping Cai, Ming Wei, et al. A trial of
lopinavir–ritonavir in adults hospitalized with severe covid­19. New England Journal of Medicine, 2020.

[12] NHSDigital. Hospital admitted patient care activity 2018­19, 29 March 2020 2020. URL https://digital.nhs.uk/data­and­information/publications/
statistical/hospital­admitted­patient­care­activity/2018­19.

[13] Timothy W Russell, Joel Hellewell, Christopher I Jarvis, Kevin van Zandvoort, Sam Abbott, Ruwan Ratnayake, CMMID COVID­19 working group,
Stefan Flasche, Rosalind M Eggo, W John Edmunds, and Adam J Kucharski. Estimating the infection and case fatality ratio for coronavirus
disease (covid­19) using age­adjusted data from the outbreak on the diamond princess cruise ship, february 2020. Eurosurveillance, 25(12):
2000256, 2020. doi: https://doi.org/10.2807/1560­7917.ES.2020.25.12.2000256. URL https://www.eurosurveillance.org/content/10.2807/1560­
7917.ES.2020.25.12.2000256.

12

You might also like