blueprint-biosecurity

/*
SIMPLE PPE COST-EFFECTIVENESS MODEL

I thought that the old cost-effective model relied too much on a tower of fake assumptions, and I wanted something that was simpler and easier to intuit.

This new model assumes that there's a certain amount of x-risk that's addressable with PPE alone. This risk exists because we either 1) don't have enough effective PPE in stock to protect everyone quickly, 2) we can't produce enough effective PPE to protect everyone quickly, or 3) both.

The amount of this PPE-addressable x-risk that's *actually* addressed is a function of two things: how much we increase PPE stocks, and how much we increase PPE production capacity. To capture this, I use an OR function: 100% of the PPE-addressable risk has been reduced when:
  - We have 8B units of PPE OR
  - We're able to produce 8B units of PPE per day

// WIP: don't take these numbers literally

/*
BETS COST-EFFECTIVENESS MODEL

This model is meant to estimate the rough cost-effectiveness of different BETS interventions. Broadly, it looks at two paths for these interventions to reduce bio x-risk:
1. Lowering the chance that an x-risk-level infectious disease is able to spread sustainably (particularly before initial detection)
2. Improving humanity's ability to weather an x-risk-level infectious disease by reducing vital worker likelihood of infection, either in workplaces (wearing PPE) or safe zones (not wearing PPE)

To-do:

// This is an uncertainty analysis of the annual burden of respiratory disease from long range aeorosol transmission in the US. See this documentation for why we have chosen these parameter values: https://docs.google.com/document/d/1Tnyb5WHn727QrFKknFiMyqSm68fYG8tTlgfHedjzT98/edit

//Background parameters
@name("QALY value")
qaly = 1
@name("QALY adjustment")
qaly_adjustment = normal(0.7,0.2)

@name("Long range aerosol transmission rates")
aeorol_rates = {

// WIP: don't take these numbers literally

/*
MODELING THE COST-EFFECTIVENESS OF PPE INTERVENTIONS

Changelog from sheets version:
- assuming flex production only fully realized at the end (step func vs continuous)
- production lag and flex lag moved into one
- new lines are a multiple of flex production
*/

annual_benefit = 100

risk_window = 30

speedup = 1
probability_increase = 1%

counterfactual_probability = 1% to 10%
counterfactual_years_in_future = 10 to 100

@name("Total addressable burden of respiratory disease")
addressable_burden = 234.8B to 842.3B //  See https://squigglehub.org/models/blueprint-biosecurity/burden-of-resp-disease

@name("Total indoor public space in the US (1000 sqft)")
space = 96423000 // https://www.eia.gov/consumption/commercial/data/2018/#b1-b2

@name("Space covered")
space_covered = 10% // We fix the percentage of public indoor space covered at 10%

@name("Transmission in the space")

/*
Describe your code here
*/
// The value of work which pays off in short timelines scenarios only
p_tai_before_2030 = 1% to 24% // 1% is an assumption, 24% is the Metaculus forecast here https: https://metaculus.com/questions/19356/transformative-ai-date/
annual_biorisk = (0.1% to 10%)/30*(1 to 10) // assuming that annual biorisk is 1-10x as high in short timelines worlds, using ASB's total biorisk/30 year risk window
risk_reduction = 0.5% to 5% // assumption
years_of_risk_reduction = 0.5 to 5 // assuming for a successful intervention we get at least 0.5 years, and max 5 (2030-2025)
bps_at_stake = 10k*p_tai_before_2030*annual_biorisk*risk_reduction*years_of_risk_reduction

/*
Describe your code here
*/

fixture = 200 to 1000
installation = 50 to 500
lamp_energy_usage = 11
energy_price = 0.1 to 0.3
annual_operating_hours = 2500 to 365*24
annual_energy_costs_per_fixture = lamp_energy_usage*energy_price/1000*annual_operating_hours

/*
Describe your code here
*/
bps = 10k
bps_per_bn_bar = 0.5 to 50

risk_window = 30 // probably makes sense for this to be a distribution, e.g. TAI timelines or similar
bio_x_risk = (0.1% to 10%)/risk_window // annual biorisk during risk window

risk_reduction = lognormal({p25: 2%, p75: 5%})

/*
Describe your code here
*/
bps = 10k
bps_per_bn_bar = 0.5 to 50

risk_window = 30 // probably makes sense for this to be a distribution, e.g. TAI timelines or similar
bio_x_risk = 0.001% to 0.1% // annual biorisk during risk window, Luca (previously ASB: (0.1% to 10%)/risk_window)

risk_reduction = lognormal({p25: 2%, p75: 5%})

risk_window = 30
annual_bio_x_risk = (0.1% to 10%)/risk_window

risk_reduction = 1%

years_speedup = 1

prob_maturity_in_window = 20%

bps = 10k

// This is an uncertainty analysis of the annual burden of respiratory disease from long range aeorosol transmission in the US. See this documentation for why we have chosen these parameter values: https://docs.google.com/document/d/1Tnyb5WHn727QrFKknFiMyqSm68fYG8tTlgfHedjzT98/edit

//Background parameters
@name("QALY value")
qaly = 750k
@name("QALY adjustment")
qaly_adjustment = normal(0.7,0.2)

@name("Long range aerosol transmission rates")
aeorol_rates = {

/*
Describe your code here
*/

//Baseline
daly = 100k
pop = 8B
economic_multiplier = 2 // I think GWH uses 2 as a rule of thumb for non-existential pandemics
pandemic_duration = 2 to 4 // https://zenodo.org/records/4626111: average duration in the dataset is 4

qaly = 750k
qaly_adjustment = 71%

aeorol_rates = {
  covid: 20% to 80%,
  flu: 20% to 80%,
  lris: 12% to 50%,
  non_flu_lris: 4% to 16%
}

// Background params
bps = 10k
bps_per_bn_bar = 0.5 to 50 // OP rough bar

// How much risk is at stake?
bio_x_risk = 0.1% to 10% // biorisk during 30 year risk window, from ASB
aerosolisation_rate = 1% to 50% // placeholder - addressable transmission vectors
aware_of_transmission = 20% to 60%
infections_in_workplace = 30% to 90% // placeholder
non_redundant_infections = 30% to 70% // placeholder

// Far-UVC 

// 1. Set-up  

// 1.1 Set-up: units
m = 10^6
b = 10^9
bps = 10000

// 1.2 Set-up: biorisk

/*
Describe your code here
*/
bps = 10k
bps_per_bn_bar = 0.5 to 50

risk_window = 30 // probably makes sense for this to be a distribution, e.g. TAI timelines or similar
bio_x_risk = (0.1% to 10%)/risk_window // annual biorisk during risk window

risk_reduction = 1% // placeholder

/*
Describe your code here
*/
bps = 10k
bps_per_bn_bar = 0.5 to 50

risk_window = 30 // probably makes sense for this to be a distribution, e.g. TAI timelines or similar
bio_x_risk = (0.1% to 10%)/risk_window // annual biorisk during risk window

risk_reduction = 13% // 

/*
Describe your code here
*/

/*3.2 Bottom up theories of change */
grantmaker_budget = 5M to 100M
uvc_speed_up = 1 to 10

// Blueprint cost module
//Costs: direct costs

/*
Describe your code here
*/

// Case rate info
annual_deaths = 60k to 150k
//https://covid.cdc.gov/covid-data-tracker/#trends_totaldeaths_select_00 provisional cases in the last year: 65355
//https://ourworldindata.org/covid-cases confirmed deaths over the last year 1.18m-1.12m = 60000
//https://www.metaculus.com/questions/7546/deaths-from-covid-19-per-year-2022-2025-in-us/ 130k forecasted per year 2022 to 2025