ABCDEFGHIJKLMNOPQRSTUVWXYZAAAB
1
Burkina FasoTanzaniaMozambiqueMalawiMadagascarEthiopiaUgandaCote d'IvoireZambiaNotesSourceLink
2
3
Cost
4
Overall funding need$2,250,000$250,000$250,000$250,000$250,000$250,000$250,000$250,000$250,000$250,000
"The costs of delivering 3 month campaigns in Burkina Faso, Côte d’Ivoire, Ethiopia, Madagascar, Malawi, Mozambique, Tanzania, Uganda and Zambia are varied, but average at approximately $250,000 USD per country."
DMI: COVID-19 Communications Response in Sub-Saharan Africa 2020
https://drive.google.com/file/d/1X8-terNgZwYc_lz8VIDI_nw8NLcEJWAH/view
5
6
Likelihood of where marginal funding will go5.0%15.0%15.0%10.0%10.0%10.0%10.0%15.0%10.0%
s a result of previous fundraising we have reduced the allocation to Burkina Faso, and as a result of DMI's preference for where to work, we have increased allocations to Tanzania, Mozambique, and Cote d'Ivoire
GiveWell's non-verbatim summary of a conversation with Development Media International, April 9, 2020
https://docs.google.com/document/d/1trPnz1v_oOMMOU1fp-B_c6J5jO-6VQnj-nRgBzFdEms/edit
7
8
Deaths prevented through social distancing
9
Imperial College model scenarios - deaths
The Imperial College model includes adjustments for age-specific mortality and different levels of social contact. I adjust for other factors below. Mortality estimates are from China but I haven't dug in deeply
Walker et al. 2020, Appendix data sources
https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-Global-unmitigated-mitigated-suppression-scenarios.xlsx
10
Unmitigated38,226117,79263,87837,25863,229282,44871,37755,13230,908
Do nothing scenario from IC model
Walker et al. 2020, Appendix data sources
https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-Global-unmitigated-mitigated-suppression-scenarios.xlsx
11
Social distancing whole population29,25488,97347,94228,25146,586205,24656,14441,22224,037
Reducing social contact by 40% uniformly
Walker et al. 2020, Appendix data sources
https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-Global-unmitigated-mitigated-suppression-scenarios.xlsx
12
Enhanced social distancing of elderly26,75380,87243,15925,57542,016182,31851,83637,37422,077
Reducing social contact by 40% uniformly and 65% for ages 70+
Walker et al. 2020, Appendix data sources
https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-Global-unmitigated-mitigated-suppression-scenarios.xlsx
13
1.6 deaths per 100,000 per week trigger17,14975,74929,79419,64624,189135,20333,82326,32817,377
Reducing social contact by 75% uniformly
Walker et al. 2020, Appendix data sources
https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-Global-unmitigated-mitigated-suppression-scenarios.xlsx
14
0.2 deaths per 100,000 per week trigger3,57018,5899,0885,1188,56325,56214,5645,3905,234
Reducing social contact by 75% uniformly
Walker et al. 2020, Appendix data sources
https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-Global-unmitigated-mitigated-suppression-scenarios.xlsx
15
16
Whole population
17
Deaths prevented by moving from 0% to 45% social distancing
8,97228,81915,9369,00716,64377,20215,23313,9106,871
(1) This assumes the effect of social distancing is the same as it would be on the margin of 0->45%. I don't think that's true but I don't know what margin of social distancing we're likely to be on (I'd subjectively guess 35%->45%) and can't back out marginal figures from the Imperial College model. My understanding from playing with this epi model (https://gabgoh.github.io/COVID/index.html) is the effect of social distancing is n-shaped so it's unclear whether this is an over or underestimate (2) The mechanism through which social distancing is expected to save lives in the Imperial College model is through achieving herd immunity at a lower level of infections.
18
Deaths prevented by each ppt of social distancing for the general population1996403542003701716339309153
19
Expressed as a percentage of unmitigated deaths
0.5%0.5%0.6%0.5%0.6%0.6%0.5%0.6%0.5%
20
21
Social distancing of elderly
22
Deaths prevented by moving from 45% to 65% social distancing
2,5018,1014,7832,6764,57022,9284,3083,8481,960
(1) This assumes the effect of social distancing is the same as it would be on the margin of 0->45%. I don't think that's true but I don't know what margin of social distancing we're likely to be on (I'd subjectively guess 35%->45%) and can't back out marginal figures from the Imperial College model. My understanding from playing with this epi model (https://gabgoh.github.io/COVID/index.html) is the effect of social distancing is n-shaped so it's unclear whether this is an over or underestimate (2) The mechanism through which social distancing is expected to save lives in the Imperial College model is (mainly) through achieving herd immunity at a lower level of infections.
23
Deaths prevented by each additonal ppt of social distancing for the elderly10032419110718391717215478
24
Expressed as a percentage of unmitigated deaths
0.3%0.3%0.3%0.3%0.3%0.3%0.2%0.3%0.3%
25
26
Best guess adjustments
27
Adjustment for comborbidities1.41.41.41.41.41.41.41.41.4
Quick guess. A lot of people we've spoken with have been concerned about this but my understanding is there isn't yet strong evidence on the effect of comorbidities. I haven't looked into it deeply.
28
Adjustment for worse healthcare at baseline
222222222
Quick guess. I'm really uncertain about this estimate. My understanding from calls is that patients treated with ventilators have >50% chance of death in any case, which means I doubt lack of ventilation will be the biggest driver of how bad this could get. I'm more concerned about lack of oxygen. My understanding is ~4x as many people need oxygen as ventilation, and a doctor told me without oxygen most of those people would die. I don't have a strong quantitative understanding of oxygen supplies in LICs. My inside view here is more like 4x because of lack of oxygen but my outside view is it won't be that big because metaculus and GJP estimates seem less pessimistic than me and I think it's likely there's something I'm missing as haven't looked deeply.
https://goodjudgment.io/covid/dashboard/
29
Adjustment for scale up of healthcare0.90.90.90.90.90.90.90.90.9
Quick guess. The Imperial College model doesn't take into account emergency increases in healthcare capacity. I haven't done research into how feasible it would be to scale up capacity but I'm skeptical this will be a large effect relative to the gap in health capacity between LICs and China where mortality estimate come from.
30
Adjustment for campaign only lasting 3 months
0.50.50.50.50.50.50.50.50.5
We're modelling the effect size of social distancing over the course of the epidemic whereas the DMI funding is enough for a 3 month campaign. I don't know how long the epidemic will last (it depends on whether these countries are able to achieve suppression (longer) or get to herd immunity (shorter) which is very uncertain). I think the first 3 months are very important, although that's only true if achieve suppression. I'm going with 0.5 adjustment here because I'd guess some form of social distancing will be important for a year but half the value is in the first 3 months. Very subjective and uncertain.
31
Internal validity adjustment0.60.60.60.60.60.60.60.60.6
Quick guess. The Imperial College model is a modelling exercise I haven't vetted deeply. (1) Metaculus estimates 2.4m people will die from covid in 2020, which is substantially lower than the 24m people IC forecasts would die if every country did a mitigation response (40% reduction in herd immunity). But I would guess HICs are more likely to be more overestimated relative to LICs (can say more) (2) Bill Gates thought the parameters in the Imperial College were too pessimistic, and I'd expect he's been well briefed (https://www.reddit.com/r/Coronavirus/comments/fksnbf/im_bill_gates_cochair_of_the_bill_melinda_gates/) (3) This paper is a great accessible overview for why the structural modelling assumptions of Imperial College could be wrong and an overestimate (can say more but in brief they assume more homegeneous social mixing than is likely true): https://www.nature.com/articles/d41586-020-01003-6
32
Overall adjustments to IC model0.80.80.80.80.80.80.80.80.8
33
34
Best guesses
35
Deaths prevented by each ppt of social distancing for the general population1514842681512801297256234115
36
Deaths prevented by each additonal ppt of social distancing for the elderly762451448113869313011659
37
38
Social distancing attributable to DMI
39
Best guess of media campaign effect size on social distancing of general population
3.0%3.0%3.0%3.0%3.0%3.0%3.0%3.0%3.0%
Very uncertain best guess based on a brief review of comparable interventions (not written up)
40
Best guess of additional media campaign effect size on social distancing of elderly
0.6%0.6%0.6%0.6%0.6%0.6%0.6%0.6%0.6%
Quick guess. One of the behaviors DMI is specifically trying to encourage is additional social distancing of the elderly, so I'd expect there to be an additional effect here. I don't feel well calibrated on what this would be but an additional 20% on top of the general population seems sensible.
41
Proportion attributable to DMI70%50%50%20%20%20%20%50%20%
Quick guess. DMI's experience in each country varies, and it will play differing roles in each country's COVID-19 campaign: "DMI believes that given sufficient funding, it can work in all nine countries it is targeting; however, its role in each country will be different. Ethiopia, for example, is a large country with powerful media, and DMI's presence in the country is smaller and newer than that of other organizations. For this reason, DMI believes it is best suited to a quality control role in Ethiopia. In Burkina Faso or Tanzania, on the other hand, DMI may take responsibility for the entire COVID-19 media campaign." GiveWell's non-verbatim summary of a conversation with Development Media International, April 9, 2020, Pg 7. Because of this, I have chosen 70% for Burkina Faso, where DMI has the greatest experience and expects to be able to substantially increase radio coverage. I have chosen 50% for Tanzania, Mozambique, and Côte D'Ivoire because DMI told me it expects to play a major role in those countries.
42
43
Best guess effect of DMI on social distancing for general population
2.1%1.5%1.5%0.6%0.6%0.6%0.6%1.5%0.6%
44
Best guess effect of DMI on additional social distancing for elderly
0.4%0.3%0.3%0.1%0.1%0.1%0.1%0.3%0.1%
45
46
Lives saved
47
Lives saved40134880044510018486116938576
48
Cost per life saved$623$718$313$562$2,488$1,356$290$1,478$649$3,275
49
50
How much more valuable is it to prevent the death of a child under the age of 5 than a death from covid?
222222222
Our moral weights are currently uncertain, but deaths from covid tend to be substantially older (can do more on this)
https://www.givewell.org/how-we-work/our-criteria/cost-effectiveness/comparing-moral-weights
51
52
Cost per outcome as good as averting the death of an individual under 5
$1,436$625$1,124$4,975$2,712$580$2,956$1,297$6,550
53
54
Cost per outcome as good as averting the death of an individual under 5 (GiveDirectly)
$29,068$29,068$29,068$29,068$29,068$29,068$29,068$29,068$29,068
https://docs.google.com/spreadsheets/d/1zLmPuddUmKsy3v55AfG_e1Quk-ngDdNzW-FDx0T-Y94/edit#gid=1034883018&range=B31
55
56
Multiples of cash20.246.525.95.810.750.19.822.44.4
57
58
Overall cost-effectiveness23.3
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100