There are no equal opportunity infectors

Making equity a first-class concern of epidemiological model(er)s.

2023 IDM Symposium

Jon Zelner
[email protected]


Dept. of Epidemiology
Center for Social Epidemiology and Population Health
University of Michigan School of Public Health

EpiBayes Research Group
epibayes.io

Agenda

  • What is this session about?

  • How did we get here?

  • What does it mean to model for equity?

  • What challenges remain?

Inequity in COVID-19 death was easy to foresee, but most early models didn’t have much to say about it.

Effective vaccines and other interventions increased inequity even while decreasing overall risk

A motivating thought experiment.

  • If we re-ran the COVID-19 pandemic from the beginning again with all of our accumulated knowledge, would we expect local, national and global patterns of inequity in infection and death to look meaningfully different?

  • Structural inequalities that effectively guaranteed these outcomes are still firmly in place.

  • How can modelers contribute to minimizing these risks in the present and future?

A lack theory, intuition, data, and methods for anticipating and targeting inequity was a signal failure of preparedness.

Implications of COVID-19 modeling have been broad, shaping everything from intuition, to public policy and social discource (Figure from (1))

Measured outcomes of early-pandemic models showed what our public health priorities really were.

  • Jurisdictional variation in risks of infection and death.
  • Age-specific incidence and mortality.
  • Community-level hospital and ICU capacity.
  • Short-term forecasting of population incidence and mortality
  • Impacts of spatially targeted interventions on infection risk.
  • Economic, racial and geographic inequities in infection and disease risk.

Archetype of viral pandemics as equal opportunity infectors precluded socio-structural understandings of pandemic risk.

What does it mean to be an equal opportunity infector?

  • Susceptibility is uniformly distributed across the population.

  • Host and pathogen biology are the most important factors in determining infectiousness.

  • Protective health behaviors equally available to everyone.

  • Socio-spatial differences in exposure by race and wealth are dwarfed by these biological factors.

We don’t have to believe all of these things are true for them to end up in our models!

Asking who our models are for may explain some of their failures

  • Public health officials making decisions.

  • Scientists and physicians pursuing interventions.

  • Politicians implementing policy and trying to maintain power.

  • Media constructing an easily-digestible narrative.

  • These groups share a top-down view from a position of power and authority.

  • Confronting questions of local, domestic or global inequity are often not well-aligned with incentives of those with power.

Modeling for equity: A three-legged stool

  1. Ask questions that foreground macro-structural mechanisms and power differentials.

  2. Collect data that are up to the task of answering these question.

  3. Construct models that examine the potential impacts of socio-structural change on disease even if these changes don’t correspond to a formal policy or intervention.

Equity-oriented models take social mechanisms as seriously as biological ones.

COVID-19 disparities are not the fault of those who are experiencing them, but rather reflect social policies and systems that create health disparities in good times and inflate them in a crisis. The US must develop a new kind of “herd immunity,” whereby resistance to the spread of poor health in the population occurs when a sufficiently high proportion of individuals, across all racial, ethnic, and social class groups, are protected from and thus “immune” to negative social determinants.

From Williams & Cooper, “COVID-19 and Health Equity—A New Kind of “Herd Immunity”, JAMA, 6/23/2020

Equity-oriented models require an informed understanding of the connections between high-level, intermediary, and proximal causes of infection.

Flow diagram of connections between high-level, structural causes, intervening mechanisms, and downstream exposure risks, from (2)

Equity-oriented models cannot treat race and other social categorizations as causal variables.

Theoretical diagram from (3) showing the complex relationship between racism, socioeconomic inequity and racial inequity in disease outcomes.

Equity-oriented models must use comparisons that reflect the impact of structural changes on disease risk.

Proportion of SARS-CoV-2 deaths potentially averted in 2020 if all Michiganders experienced the same pacing of SARS-CoV-2 infection as Whites (From Naraharisetti et al., Under Review)

Many challenges remain

  • Models that encode biased assumptions may be worse than those that ignore equity altogether.

  • Nonexistent data on individual socioeconomic status in most public health surveillance data.

  • Datasets plagued by missingness of key socio-demographic covariates. (4)

  • Career/funding incentives still not well-aligned.

  • Modelers from marginalized backgrounds remain underrepresented.

Thanks!

  • Please get in touch with any questions/thoughts/concerns at [email protected].

  • Check out our work at epibayes.io.

References

1.
Zelner J, Eisenberg M. Rapid response modeling of SARS-CoV-2 transmission. Science [electronic article]. 2022;376(6593):579–580. (http://www.science.org/doi/full/10.1126/science.abp9498). (Accessed May 11, 2022)
2.
Noppert GA, Hegde ST, Kubale JT. Exposure, Susceptibility, and Recovery: A Framework for Examining the Intersection of the Social and Physical Environment and Infectious Disease Risk. American Journal of Epidemiology [electronic article]. 2022;kwac186. (https://doi.org/10.1093/aje/kwac186). (Accessed November 3, 2022)
3.
Phelan JC, Link BG. Is Racism a Fundamental Cause of Inequalities in Health? Annual Review of Sociology [electronic article]. 2015;41(1):311–330. (http://www.annualreviews.org/doi/10.1146/annurev-soc-073014-112305). (Accessed December 15, 2019)
4.
Trangucci R, Chen Y, Zelner J. Modeling racial/ethnic differences in COVID-19 incidence with covariates subject to non-random missingness. 2022;(http://arxiv.org/abs/2206.08161). (Accessed December 8, 2022)