Integrated Multidisciplinary Modeling for Pandemic Response Policy

Integrated Multidisciplinary Modeling for Pandemic Response Policy

Revised 20 April 2020

To guide policy decisions about the COVID-19 pandemic response, there is a great need for an integrated system of models.  One set of models will provide input for another set of models in a linked structure.  Feedback loops allow candidate policy choices to be evaluated in terms of prior conditions.  The models need to be structured to support analysis of different alternatives and projected outcomes.

Policy makers cannot make good policy decisions on the basis of intuition in this circumstance.  Intuition is commonly flawed for exponential phenomena such as epidemic outbreaks.  Policy makers need to trust the predictions of good modeling.

Each run of the set of models needs to be based on a common set of assumptions and conditions in order for the outputs to be meaningful.  Isolated models are likely to diverge and create ambiguity for policy makers.

In many cases, the models will start with partial data and best guesses.  They will need to be progressively improved as more insight is gained.  When new information is obtained for a given model in the sequence, the consequences for downstream models need to be evaluated.

The starting point for the system of models is what is driving the epidemic—i.e., the characteristics of the COVID-19 disease.   Unfortunately, because this is a virus that has not been encountered before, there is considerable uncertainty about many of its characteristics until more research is performed.

  • Medical models specific to the particular infection
    • Infection transmission modes (e.g., primarily respiratory ingestion of virus particles)
    • Disease progression timing factors (e.g., time from exposure to being infectious, time from exposure to appearance of symptoms, etc.)
    • Illness severity factors (fraction of infected patients requiring hospitalization, fraction requiring intensive care, average duration of hospitalization, etc.)
    • Duration of immunity after recovery or vaccination
    • Virus mutation rate and length of protection from a vaccine

The characteristics of the disease will influence the possible timelines for developing medical responses.

  • Medical response development models
    • Projected timeline for development of a vaccine
    • Projected timeline for development of a disease treatment

Epidemiological models are driven by the medical models of the disease.

  • Models for the dynamics of the spread of the disease
    • Infection rate (particularly for direct contact personnel such as medical staff)
    • Infection doubling time for different conditions
    • Conditions required to achieve herd immunity

To control the spread of the pandemic, public health authorities institute a range of mitigation measures.  The mitigation measures need to be appropriate in form and scope to the characteristics of the disease and the epidemiological factors.

The mitigation measures have increasingly serious negative societal impacts according to the extent and duration they are imposed.  Policy makers need to justify the pain of the mitigation measures on the basis of estimates of how much illness and death will be avoided.

  • Models of mitigation measures and their effectiveness depending on levels of compliance
    • Restrictions on travel
    • Restrictions on gatherings
    • Closure of non-essential businesses and activities
    • Quarantine of those infected
    • Isolation of those known exposed
    • Stay-at-home orders for all except those engaged in essential functions
    • Physical distancing in public
    • Personal protection measures (hand washing, sanitizing, masks, gloves, etc.)
    • Infection testing
    • Contact tracking

Testing people to determine whether they are uninfected, infected and contagious, or recovered and no longer contagious is an essential element of response to the pandemic for the longer term.  However, it takes time to ramp up the necessary scope of testing.  This interacts with the models of mitigation measures to evaluate when and how much the mitigation measures can be relaxed.

  • Models of testing program logistics

The health care systems enter the pandemic period with a set of resources.  How these resources can be deployed to respond to the pandemic depend on the available capacity and the ability to re-prioritize care activities.  The health care systems will need additional resources (equipment and consumables) during the pandemic response.  A key factor is the capacity of the supply chains to respond with sufficient new supplies in a timely manner.

  • Health care system resource models
    • Hospital beds, particularly ICU facilities
    • Laboratories
    • Direct patient contact medical personnel (physicians, nurses, aides, etc.)
    • Non-direct contact personnel (lab technicians, etc.)
    • Pre-existing resource stockpiles
  • Supply chain models for producing new medical equipment and consumables
    • Personal protective equipment (PPE)
    • Ventilators, test analysis machines, etc.
    • Testing supplies, treatment drugs, etc.

The models need to project the ability of the health care systems to respond before being overwhelmed and collapse.  The models should also consider the effects on other health care services that are postponed during the pandemic.

The costs incurred by the health care systems for the pandemic response need to be assessed for the overall economic impacts picture.  Such costs are complicated by many factors, such as the negotiated arrangements with the various insuring entities and how costs are allocated between different payers.

  • Cost models, e.g.
    • Cost per test
    • Cost per hospitalization
    • Cost per vaccine
    • Cost per treatment regimen

A key factor in pandemic disease response in the United States is the degree to which patients are able to afford to access health care services.  Those who judge they can’t afford it will enter the health care system later and sicker and will increase the number of extended hospitalizations and deaths.

  • Models of patient ability to access health care (insurance, sick leave, ability to handle uninsured costs and co-pays, etc.)

The response to an extreme event such as a pandemic will cause extreme financial stress on all the components of the health care system, particularly as revenue from postponed regular activities disappears.  The ability to sustain operations during the pandemic is dependent on staying financially solvent.

  • Models of health care system financial strength and ability to adjust
    • Financial strength of hospitals
    • Financial strength of other medical resources (medical personnel, clinics, etc.)
    • Financial strength of health insurance entities

Ordering businesses and institutions, other than those providing functions considered essential, to close means that large fractions of the population lose their regular income.  It is important to model the severity of this income loss as a function of the duration of the shutdown.  The consequences of the loss of employment income need to be modeled to determine downstream effects (e.g., inability to make loan payments for housing, transportation, etc., pay rent and utility bills, and buy food and other necessities).

  • Models of employment under business shutdown conditions
    • Fraction laid off
    • Fraction working at reduced income
    • Fraction continuing to work at full rate
  • Models of population financial status (savings, monthly payment obligations, etc.)

When large fractions of the population and institutions are impacted by financial stress from the pandemic shutdown, the ability of governments to provide support needs to be modeled.  Government assistance programs need to be expanded to prevent devastating outcomes such as skyrocketing homelessness and failures of essential businesses.

  • Models of government safety net programs for individuals (unemployment insurance, loans, direct financial distributions, etc.)
  • Models of government safety net programs for businesses and institutions (loans, direct financial distributions, etc.)

How governments provide the funds for assistance programs requires models scaled to the scope of the needs.  Existing tax structures are likely to be inadequate to provide the required revenue because of the decline in both business and personal income.  Governments’ borrowing ability will depend on the existing level of indebtedness and confidence in the global markets for government financial instruments.

  • Models of government tax revenues (federal, state, local)
  • Models of government borrowing (federal, state, local)
  • Models of government debt (federal, state, local)

A factor underlying most of the other models is how populations will respond to the difficult conditions of a pandemic and the shutdown of normal activities, particularly if these persist over an extended duration.

  • Models of social response to difficult conditions, particularly depending on the messaging from authorities

2 thoughts on “Integrated Multidisciplinary Modeling for Pandemic Response Policy”

  1. Dennis:
    I agree with the model variables (current unknowns) that would need to be specified. One would likely need to run a Monte Carlo like series of simulations to get an idea of what (or if) there’s an “expected” outcome, and of the range of possible outcomes if they are spread all over the map.
    One added level of detail not on your list (or I missed it) was impacts by income and/or ethnic group, or some such distinction. Alas, policy makers seem to be more influenced by not just number of individuals effected, but by who these individuals are.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top